{"pop2piano/modeling_pop2piano.py:Pop2PianoLayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "pop2piano/modeling_pop2piano.py:Pop2PianoDenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "pop2piano/modeling_pop2piano.py:Pop2PianoDenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseGatedActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "pop2piano/modeling_pop2piano.py:Pop2PianoLayerFF": ["DenseReluDense", "Dropout", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelLayerFF", "ModelLayerNorm", "Module", "__init__", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is_gated_act", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "pop2piano/modeling_pop2piano.py:Pop2PianoAttention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "if", "index", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "len", "list", "log", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "union", "unsqueeze", "update", "use_cache", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "pop2piano/modeling_pop2piano.py:Pop2PianoLayerSelfAttention": ["Dropout", "False", "ModelAttention", "ModelLayerNorm", "ModelLayerSelfAttention", "Module", "None", "Optional", "SelfAttention", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "pop2piano/modeling_pop2piano.py:Pop2PianoLayerCrossAttention": ["Dropout", "EncDecAttention", "False", "ModelAttention", "ModelLayerCrossAttention", "ModelLayerNorm", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "pop2piano/modeling_pop2piano.py:Pop2PianoBlock": ["False", "GradientCheckpointingLayer", "ModelBlock", "ModelLayerCrossAttention", "ModelLayerFF", "ModelLayerSelfAttention", "ModuleList", "None", "Optional", "True", "__init__", "and", "any", "append", "attention_mask", "attention_outputs", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "encoder_attention_mask", "encoder_decoder_position_bias", "encoder_hidden_states", "finfo", "float16", "forward", "has_relative_attention_bias", "hidden_states", "if", "int", "is", "is_decoder", "isinf", "key_value_states", "layer", "layer_head_mask", "layer_idx", "max", "min", "new_name", "nn", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "return_dict", "self", "self_attention_outputs", "super", "torch", "use_cache", "version", "where"], "pop2piano/modeling_pop2piano.py:Pop2PianoPreTrainedModel": ["False", "In", "Model", "ModelAttention", "ModelBlock", "ModelConcatEmbeddingToMel", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelForConditionalGeneration", "ModelLayerNorm", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "ValueError", "_can_compile_fullgraph", "_init_weights", "_keep_in_fp32_modules", "_no_split_modules", "_shift_right", "and", "base_model_prefix", "be", "bias", "cat", "class", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_start_token_id", "def", "defined", "dim", "elif", "else", "embedding", "factor", "fill_", "full", "has", "has_relative_attention_bias", "hasattr", "if", "initializer_factor", "input_ids", "is", "is_parallelizable", "is_torch_fx_proxy", "isinstance", "it", "k", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "n_heads", "new_zeros", "normal_", "not", "num_heads", "o", "pad_token_id", "q", "raise", "relative_attention_bias", "return", "self", "set", "shape", "shared", "shifted_input_ids", "std", "supports_gradient_checkpointing", "the", "tie_word_embeddings", "to", "torch", "transformer", "usually", "v", "weight", "wi", "wi_0", "wi_1", "wo", "zero_"], "pop2piano/modeling_pop2piano.py:Pop2PianoStack": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "Dropout", "DynamicCache", "EncoderDecoderCache", "False", "ModelBlock", "ModelLayerNorm", "ModelPreTrainedModel", "ModelStack", "ModuleList", "None", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "a", "all_attentions", "all_cross_attentions", "all_hidden_states", "and", "any", "arange", "as", "at", "attention_mask", "attentions", "batch_size", "be", "block", "bool", "both", "cache_position", "can", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "d_model", "decoder", "decoder_", "def", "device", "device_map", "diagonal", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder_attention_mask", "encoder_batch_size", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_relative_attention_bias", "have", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_torchdynamo_compiling", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "logger", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "min", "min_dtype", "model", "model_parallel", "new_embeddings", "nn", "not", "npu", "num_layers", "ones", "only", "or", "output_attentions", "output_hidden_states", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_bias", "post_init", "raise", "range", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "used", "using_compilable_cache", "v", "valid", "view", "warning_once", "with", "xpu"], "pop2piano/modeling_pop2piano.py:Pop2PianoConcatEmbeddingToMel": ["Embedding", "ModelConcatEmbeddingToMel", "Module", "__init__", "cat", "class", "composer_embedding", "composer_vocab_size", "config", "d_model", "def", "dim", "embedding", "embedding_dim", "embedding_offset", "feature", "forward", "index_shifted", "index_value", "inputs_embeds", "nn", "num_embeddings", "return", "self", "super", "torch", "unsqueeze"], "pop2piano/modeling_pop2piano.py:Pop2PianoForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "Both", "Cache", "Composer", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "Found", "GenerationConfig", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelConcatEmbeddingToMel", "ModelConfig", "ModelForConditionalGeneration", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Please", "Seq2SeqLMOutput", "Tensor", "True", "Union", "ValueError", "__init__", "_shift_right", "_tied_weights_keys", "a", "and", "as", "attention_mask", "attentions", "auto_docstring", "axis", "be", "bias", "blob", "bool", "cache_position", "choose", "class", "co", "composer", "composer1", "composer_to_feature_token", "composer_value", "composer_vocab_size", "concatenate", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "device", "dict", "elif", "else", "embed_tokens", "embedding_offset", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "feature", "forward", "found", "from", "generate", "generation_config", "get_encoder", "get_input_embeddings", "get_mel_conditioner_outputs", "hasattr", "head_mask", "hidden_states", "https", "huggingface", "if", "ignore_index", "in", "index_value", "input_features", "input_ids", "inputs", "inputs_embeds", "is", "is_decoder", "isinstance", "json", "keys", "kwargs", "labels", "last_hidden_state", "len", "like", "list", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "main", "mel_conditioner", "min", "model_dim", "must", "new_embeddings", "nn", "no_grad", "not", "num_decoder_layers", "num_layers", "number", "of", "one", "only", "output", "output_attentions", "output_hidden_states", "parse", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provide", "r", "raise", "received", "refer", "repeat", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "shape", "shared", "size", "str", "super", "sweetcocoa", "tensor", "that", "the", "them", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "tuple", "update", "use_cache", "use_return_dict", "values", "view", "vocab_size", "vs", "was", "weight"], "blt/modeling_blt.py:BltMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "blt/modeling_blt.py:BltRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "blt/modeling_blt.py:BltRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "repeat_interleave", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "blt/modeling_blt.py:BltTransformerLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelMLP", "ModelRMSNorm", "ModelSelfAttention", "ModelTransformerLayer", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "cross_attention_mask", "cross_attention_states", "def", "deprecate_kwarg", "eps", "forward", "full_text_row_masked_out_mask", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "blt/modeling_blt.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "blt/modeling_blt.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "blt/modeling_blt.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "rot_x", "stack", "torch", "x", "x1", "x2"], "blt/modeling_blt.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "blt/modeling_blt.py:BltSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_len", "q_proj", "query_states", "reshape", "return", "rope_theta", "scaling", "self", "sin", "size", "super", "torch", "training", "transpose", "update", "use_cache", "v_proj", "value_states", "version", "view"], "blt/modeling_blt.py:BltCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Cross", "False", "Linear", "LongTensor", "ModelConfig", "ModelCrossAttention", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "_", "__init__", "_attn_implementation", "attention", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bsz", "cache_position", "cached", "can", "class", "config", "contiguous", "cross_attention_states", "cross_attn_states", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "elif", "else", "eps", "find", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_norm", "k_proj", "key", "key_states", "keys", "kwargs", "layer", "layer_idx", "layers", "neither", "new_name", "nn", "nor", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "q_len", "q_norm", "q_proj", "query_states", "raise", "reshape", "return", "rms_norm_eps", "scaling", "self", "size", "super", "t", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "blt/modeling_blt.py:BltPreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "ModelSelfAttention", "ModelTransformerLayer", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "index", "layer_name", "local_decoder", "supports_gradient_checkpointing"], "blt/modeling_blt.py:BltLocalEncoder": ["Cache", "Embedding", "F", "False", "Linear", "LongTensor", "ModelCrossAttention", "ModelLocalEncoder", "ModelLocalEncoderConfig", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModelSelfAttention", "ModelTransformerLayer", "ModuleList", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_can_record_outputs", "amax", "append", "arange", "attention_mask", "batch_size", "bias", "cache_position", "class", "config", "cross_attention_output", "cross_attention_states", "cross_attn_all_layers", "cross_attn_k", "cross_attn_layers", "def", "device", "dim", "dropout", "dtype", "else", "embed_tokens", "embedding_dim", "encoder_attention_mask", "encoder_attentions", "encoder_cross_states", "enumerate", "expand", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "idx", "if", "in", "in_features", "include_self", "index", "input_ids", "inputs_embeds", "int", "is", "kwargs", "layer", "layer_idx", "layer_name", "layers", "layers_to_add", "len", "local_encoder", "max_num_patches", "nn", "num_hidden_layers", "num_patches", "or", "out_features", "p", "past_key_values", "patch_embedding_projection", "patch_embeds", "patch_ids", "patch_reduce", "position_embeddings", "position_ids", "post_init", "range", "reduce", "reduced_embeddings", "reshape", "return", "rotary_emb", "scatter_reduce", "self", "shape", "src", "super", "torch", "training", "unsqueeze", "vocab_size", "zeros"], "blt/modeling_blt.py:BltLocalDecoder": ["Cache", "F", "False", "Linear", "LongTensor", "ModelCrossAttention", "ModelLocalDecoder", "ModelLocalDecoderConfig", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTransformerLayer", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "_", "__init__", "and", "append", "arange", "attention_mask", "batch_size", "bias", "cache_position", "check_model_inputs", "class", "config", "cross_attention_output", "cross_attention_states", "cross_attn_all_layers", "cross_attn_decoder", "cross_attn_k", "cross_attn_layers", "def", "device", "dropout", "else", "encoder_attention_mask", "enumerate", "eps", "expand", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_size_global", "hidden_states", "i", "if", "in", "in_features", "input_ids", "inputs_embeds", "is", "kwargs", "layer", "layer_idx", "layers", "layers_to_add", "logits", "nn", "norm", "not", "num_hidden_layers", "or", "out_features", "p", "past_key_values", "patch_embedding_projection", "patch_embeds", "position_embeddings", "position_ids", "post_init", "range", "reshape", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "super", "torch", "training", "unsqueeze"], "blt/modeling_blt.py:BltGlobalTransformer": ["Cache", "F", "False", "Identity", "Linear", "LongTensor", "ModelGlobalTransformer", "ModelGlobalTransformerConfig", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModelSelfAttention", "ModelTransformerLayer", "ModuleList", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_can_record_outputs", "append", "arange", "attention_mask", "batch_size", "bias", "cache_position", "class", "config", "def", "device", "dropout", "else", "encoder_cross_output_size", "enumerate", "expand", "for", "forward", "getattr", "global_attentions", "global_transformer", "hidden_size", "hidden_states", "i", "if", "in", "index", "input_embeds", "is", "kwargs", "layer", "layer_idx", "layer_name", "layers", "nn", "not", "num_hidden_layers", "p", "past_key_values", "position_embeddings", "position_ids", "post_init", "range", "return", "rotary_emb", "self", "seq_len", "shape", "super", "token_embedding_projection", "torch", "training", "unsqueeze"], "blt/modeling_blt.py:process_patch_lengths": ["Model_patch_lengths", "Modeled", "None", "Optional", "Tensor", "any", "append", "batch_size", "def", "device", "dim", "divmod", "dtype", "enumerate", "extend", "for", "full_chunks", "i", "if", "in", "int", "is", "item", "last_nonzero", "len", "length", "max", "max_len", "max_patch_length", "nonzero", "padded", "patch_lengths", "remainder", "return", "seq", "shape", "size", "splits", "sum", "tensor", "torch", "zeros"], "blt/modeling_blt.py:BltPatcher": ["Cache", "Categorical", "DynamicCache", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "ModelPatcher", "ModelPatcherConfig", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTransformerLayer", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "append", "arange", "attention_mask", "batch_size", "bias", "bool", "cache_position", "cat", "causal_mask", "class", "config", "create_causal_mask", "def", "device", "dim", "distributions", "dtype", "else", "embed_tokens", "entropies", "entropy", "eps", "exactly", "expand", "float", "for", "forward", "full_like", "get_seq_length", "hidden_size", "hidden_states", "if", "in", "init_tokens", "input_embeds", "input_ids", "inputs_embeds", "int", "is", "kwargs", "last_token", "layer", "layer_idx", "layers", "lm_head", "logits", "long", "max", "max_patch_length", "max_valid_patches", "must", "nn", "norm", "not", "num_hidden_layers", "of", "offset", "one", "ones", "or", "padded_indices", "padded_mask", "past_key_values", "past_seen_tokens", "patch_ends", "patch_lengths", "patch_lengths_from_entropies", "patch_mask", "patch_size", "patch_start_ids", "patch_starts", "position_embeddings", "position_ids", "prediction_entropies", "process_patch_lengths", "raise", "range", "repeat", "reshape", "return", "rms_norm_eps", "rotary_emb", "self", "sentinel", "seq_len", "sequence_length", "shape", "specify", "staticmethod", "sum", "super", "tensor", "threshold", "token_indices", "torch", "unsqueeze", "use_cache", "vocab_size"], "blt/modeling_blt.py:rolling_polynomial_hash": ["Model_polynomial_hash", "arange", "def", "device", "dim", "dtype", "int", "int64", "powers", "prime", "prime_powers", "prime_tensor", "return", "shape", "sum", "tensor", "token_tensor", "torch"], "blt/modeling_blt.py:byte_group_hash_function": ["Model_group_hash_function", "Tensor", "batch_size", "cat", "def", "device", "dim", "dtype", "group_size", "hash_values", "hashes", "int", "int64", "max_hash", "no_grad", "padded_tokens", "padding", "prime", "return", "rolling_polynomial_hash", "seq_len", "shape", "token_ids", "torch", "unfold", "windows", "with", "zeros"], "blt/modeling_blt.py:compute_hash_embeddings": ["Embedding", "Model_hash_embeddings", "Tensor", "byte_group_hash_function", "def", "embed_tokens", "embedding_idx", "embeddings", "encoder_hash_byte_group_nb_functions", "encoder_hash_byte_group_size", "encoder_hash_byte_group_vocab", "encoder_hash_tok_embedding", "for", "func_nb", "group_size", "hash_ids", "in", "int", "len", "list", "local_encoder", "local_encoder_tokens", "nn", "offset_hash_ids", "prime", "primes", "range", "return", "torch"], "blt/modeling_blt.py:_prepare_patch_cross_attention_mask": ["Cross", "False", "Tensor", "ValueError", "_prepare_patch_cross_attention_mask", "arange", "attention", "batch_size", "bool", "cross_attention_mask", "cross_attn_k", "def", "device", "dim", "doesn", "dtype", "else", "expand", "expected", "expected_shape", "f", "finfo", "float32", "if", "int", "inverted_cross_attn_mask", "kv_len", "kv_patch_ids", "mask", "masked_fill", "match", "min", "num_patches", "patch_ids", "patches_as_queries", "q_len", "q_patch_ids", "raise", "repeat_dim", "repeat_interleave", "return", "seq_len", "sequence_length", "shape", "t", "to", "torch", "tuple", "unsqueeze"], "blt/modeling_blt.py:BltModel": ["BaseModelOutputWithPast", "Cache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelGlobalTransformer", "ModelLocalDecoder", "ModelLocalEncoder", "ModelModel", "ModelPatcher", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "_", "__init__", "_patch_ids_from_lengths", "_prepare_patch_cross_attention_mask", "and", "arange", "attention_mask", "based", "batch_size", "bool", "cache_position", "cat", "causal_mask", "check_model_inputs", "class", "compute_hash_embeddings", "config", "create_causal_mask", "cross_attn_k", "cross_attn_mask_dec", "cross_attn_mask_enc", "cumsum", "decoder_config", "decoder_patch_ids", "def", "device", "dim", "dtype", "else", "embed_tokens", "encoder_attention_mask", "encoder_config", "encoder_cross_states", "encoder_embeds", "encoder_hash_byte_group_nb_functions", "encoder_hash_byte_group_size", "encoder_hash_byte_group_vocab", "encoder_hash_tok_embedding", "encoder_hidden_states", "entropy", "eval", "exactly", "for", "forward", "get_input_embeddings", "get_seq_length", "global_cache_position", "global_causal_mask", "global_config", "global_hidden_states", "global_position_ids", "global_transformer", "gradient_checkpointing", "hidden_size", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "int", "is", "kwargs", "last_hidden_state", "len", "local_decoder", "local_encoder", "max_patch_length", "must", "nn", "not", "num_embeddings", "num_patches", "of", "one", "ones", "or", "output", "param", "parameters", "past_key_values", "past_seen_tokens", "patch_embeds", "patch_ids", "patch_in_forward", "patch_lengths", "patch_size", "patch_starts", "patcher", "patcher_config", "patches_as_queries", "patching", "patching_batch_size", "patching_mode", "patching_threshold", "position_ids", "post_init", "process_patch_lengths", "raise", "required", "requires_grad", "return", "self", "seq_len", "sequence_length", "set_input_embeddings", "shape", "specify", "sum", "super", "threshold", "token_positions", "torch", "total_vocab_size", "unsqueeze", "use_cache", "value", "view", "zeros"], "blt/modeling_blt.py:BltForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attention_mask", "cross_attention_states", "decoder_config", "def", "else", "float", "forward", "full_text_row_masked_out_mask", "get_text_config", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "vocab_size", "weight"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "codevector_perplexity", "contrastive_loss", "diversity_loss", "hidden_states", "loss", "projected_quantized_states", "projected_states", "r", "torch", "tuple"], "wav2vec2/modeling_wav2vec2.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "wav2vec2/modeling_wav2vec2.py:_sample_negative_indices": ["None", "Optional", "_sample_negative_indices", "arange", "astype", "batch_idx", "batch_size", "bool", "broadcast_to", "def", "dtype", "else", "feature_indices", "features_shape", "for", "high", "if", "in", "int", "int32", "is", "mapped_masked_indices", "mask_time_indices", "ndarray", "not", "np", "num_negatives", "ones", "randint", "random", "range", "return", "sampled_indices", "sampled_negative_indices", "sequence_length", "sequence_length_range", "shape", "size", "sum", "tuple", "zeros"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2NoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2LayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2GroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2PositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2SamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2FeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2FeatureExtractor": ["FutureWarning", "ModelFeatureEncoder", "ModelFeatureExtractor", "The", "Transformers", "Use", "__bases__", "__class__", "__init__", "__name__", "and", "be", "been", "class", "config", "def", "depreciated", "f", "has", "in", "instead", "removed", "self", "super", "v5", "warn", "warnings", "will"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2FeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "wav2vec2/modeling_wav2vec2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2FeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2EncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelFeedForward", "None", "_", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2EncoderLayerStableLayerNorm": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelAttnAdapterLayer", "ModelEncoderLayerStableLayerNorm", "ModelFeedForward", "None", "Optional", "Tensor", "_", "__init__", "adapter_attn_dim", "adapter_layer", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "bool", "class", "config", "def", "dropout", "else", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "getattr", "hidden_dropout", "hidden_size", "hidden_states", "if", "is", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super", "torch"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2Encoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalConvEmbedding", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "tensor", "torch", "training", "tuple", "unsqueeze", "v"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2EncoderStableLayerNorm": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoderLayerStableLayerNorm", "ModelEncoderStableLayerNorm", "ModelPositionalConvEmbedding", "Module", "ModuleList", "None", "Tensor", "True", "Union", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "unsqueeze", "v"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2GumbelVectorQuantizer": ["FloatTensor", "Linear", "ModelGumbelVectorQuantizer", "Module", "None", "Parameter", "True", "ValueError", "__init__", "_compute_perplexity", "argmax", "batch_size", "be", "by", "class", "codevector_dim", "codevector_idx", "codevector_probs", "codevector_soft_dist", "codevectors", "codevectors_per_group", "concatenation", "config", "conv_dim", "def", "dim", "divisible", "else", "exp", "expand", "f", "flatten", "float", "for", "forward", "functional", "gumbel_softmax", "hard", "hidden_size", "hidden_states", "if", "is", "log", "marginal_probs", "mask", "mask_extended", "mask_time_indices", "mean", "must", "new_zeros", "nn", "not", "num_codevector_groups", "num_codevectors_per_group", "num_groups", "num_vars", "perplexity", "probs", "raise", "return", "scatter_", "self", "sequence_length", "shape", "softmax", "staticmethod", "sum", "super", "tau", "temperature", "torch", "training", "type_as", "unsqueeze", "view", "weight_proj", "where", "zeros_like"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2Adapter": ["LayerNorm", "Linear", "ModelAdapter", "ModelAdapterLayer", "Module", "ModuleList", "None", "_", "__init__", "and", "class", "config", "def", "else", "for", "forward", "hidden_size", "hidden_states", "if", "in", "is", "layer", "layerdrop", "layerdrop_prob", "layers", "nn", "not", "np", "num_adapter_layers", "or", "output_hidden_size", "proj", "proj_layer_norm", "random", "range", "return", "self", "super", "training", "transpose"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2AdapterLayer": ["Conv1d", "ModelAdapterLayer", "Module", "__init__", "adapter_kernel_size", "adapter_stride", "class", "config", "conv", "def", "dim", "forward", "functional", "glu", "hidden_states", "nn", "output_hidden_size", "padding", "return", "self", "stride", "super"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2AttnAdapterLayer": ["FloatTensor", "LayerNorm", "Linear", "ModelAttnAdapterLayer", "Module", "ReLU", "__init__", "act_fn", "adapter_attn_dim", "class", "config", "def", "forward", "hidden_dim", "hidden_size", "hidden_states", "input_dim", "linear_1", "linear_2", "nn", "norm", "return", "self", "super", "torch"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2PreTrainedModel": ["Adapter", "Can", "Cannot", "Conv1d", "Exception", "False", "FutureWarning", "GroupNorm", "If", "LayerNorm", "Linear", "LongTensor", "Make", "Model", "ModelAttnAdapterLayer", "ModelConfig", "ModelFeatureProjection", "ModelForCTC", "ModelForPreTraining", "ModelGumbelVectorQuantizer", "ModelPositionalConvEmbedding", "ModelPreTrainedModel", "Model_ADAPTER_PT_FILE", "Model_ADAPTER_SAFE_FILE", "None", "OSError", "Optional", "Otherwise", "Please", "PreTrainedModel", "The", "Transformers", "True", "Union", "ValueError", "_", "__class__", "_conv_out_length", "_get_adapters", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_is_hf_initialized", "_name_or_path", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "adapter", "adapter_attn_dim", "adapter_stride", "adapter_weights", "add_adapter", "already", "and", "arange", "are", "argument", "attention_mask", "b", "base_model_prefix", "batch_size", "be", "bias", "bool", "both", "cache_dir", "cached_file", "check_torch_load_is_safe", "class", "co", "codevectors", "config", "constant_", "containing", "conv", "conv_kernel", "conv_stride", "correct", "cpu", "cumsum", "data", "def", "define", "defined", "deprecated", "device", "dim", "directory", "div", "don", "dtype", "elif", "else", "except", "f", "feature_vector_length", "file", "filename", "filepath", "fill_", "flip", "floor", "for", "force_download", "force_load", "format", "from", "groups", "has", "have", "https", "huggingface", "if", "in", "in_channels", "in_features", "init", "init_adapter_layers", "initializer_range", "input_length", "input_lengths", "input_values", "instead", "int", "is", "is_safetensors_available", "isinstance", "it", "items", "join", "k", "kaiming_normal_", "kernel_size", "keys", "kwargs", "layers", "len", "lm_head", "load", "load_adapter", "load_state_dict", "local", "local_files_only", "logger", "long", "main_input_name", "make", "map_location", "math", "mean", "missing", "missing_keys", "model", "model_path_or_id", "models", "module", "modules", "name", "named", "named_modules", "named_parameters", "nn", "no", "non_padded_lengths", "normal_", "not", "num_adapter_layers", "of", "only", "output_hidden_size", "output_lengths", "param", "param_name", "path", "pop", "project_hid", "project_q", "projection", "proxies", "r", "raise", "range", "removed", "reset_parameters", "resume_download", "return", "revision", "rounding_mode", "safe_load_file", "same", "self", "set", "shape", "specified", "sqrt", "state_dict", "std", "str", "strict", "stride", "supports_gradient_checkpointing", "sure", "t", "target_lang", "target_vocab_size", "the", "to", "token", "torch", "try", "trying", "unexpected", "unexpected_keys", "uniform_", "use", "use_auth_token", "use_safetensors", "v", "v5", "vocab_size", "warn", "warning", "warnings", "weight", "weight_path", "weight_proj", "weights", "weights_only", "were", "will", "with", "you", "zero_", "zeros", "zip"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2Model": ["False", "FloatTensor", "FutureWarning", "LongTensor", "ModelAdapter", "ModelBaseModelOutput", "ModelConfig", "ModelEncoder", "ModelEncoderStableLayerNorm", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Tensor", "The", "Transformers", "True", "Union", "__init__", "_compute_mask_indices", "_freeze_parameters", "_get_feature_vector_attention_mask", "_mask_hidden_states", "adapter", "add_adapter", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "class", "config", "def", "deprecated", "device", "do_stable_layer_norm", "dtype", "elif", "else", "encoder", "encoder_outputs", "equivalent", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "freeze_feature_encoder", "freeze_feature_extractor", "getattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "method", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "removed", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "the", "to", "torch", "training", "transpose", "tuple", "uniform_", "use", "use_return_dict", "v5", "warn", "warnings", "will"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForPreTraining": ["BoolTensor", "Dropout", "False", "FloatTensor", "FutureWarning", "Linear", "Model", "ModelConfig", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelGumbelVectorQuantizer", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "Union", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "add_adapter", "all", "and", "any", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "cat", "class", "codevector_dim", "codevector_perplexity", "compute_contrastive_logits", "config", "contrastive_logits_temperature", "contrastive_loss", "cosine_similarity", "cross_entropy", "def", "deprecated", "dim", "diversity_loss", "diversity_loss_weight", "dropout_features", "dtype", "else", "equivalent", "extract_features", "feat_quantizer_dropout", "feature_extractor", "flatten", "float", "forward", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hidden_size", "hidden_states", "if", "in", "inf", "input_values", "instead", "int", "is", "logits", "long", "loss", "mask_time_indices", "method", "neg_is_pos", "negative_features", "negative_quantized_features", "nn", "not", "num_codevector_groups", "num_codevectors", "num_codevectors_per_group", "output_attentions", "output_hidden_states", "outputs", "permute", "post_init", "predicted_features", "proj_codevector_dim", "project_hid", "project_q", "projected_quantized_states", "projected_states", "quantized_features", "quantizer", "r", "reduction", "removed", "reshape", "return", "return_dict", "sampled_negative_indices", "self", "sequence_length", "set_gumbel_temperature", "shape", "size", "staticmethod", "sum", "super", "target", "target_features", "temperature", "the", "to", "torch", "transformer_features", "transpose", "tuple", "type_as", "use", "use_return_dict", "v5", "view", "warn", "warnings", "weight", "will"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForMaskedLM": ["Dropout", "FloatTensor", "FutureWarning", "Linear", "LongTensor", "MaskedLMOutput", "Model", "ModelForCTC", "ModelForMaskedLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "deprecated", "dropout", "else", "final_dropout", "forward", "hidden_size", "hidden_states", "if", "input_values", "instead", "is", "labels", "lm_head", "logits", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "return", "return_dict", "self", "super", "torch", "tuple", "use", "use_return_dict", "vocab_size", "warn", "warnings"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForAudioFrameClassification": ["Audio", "CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForAudioFrameClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Tensor", "The", "TokenClassifierOutput", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "adapters", "add_adapter", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "axis", "be", "bool", "class", "classification", "classifier", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "feature_extractor", "for", "forward", "frame", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "init_weights", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "param", "parameters", "r", "raise", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "wav2vec2/modeling_wav2vec2.py:AMSoftmaxLoss": ["CrossEntropyLoss", "Model", "Module", "Parameter", "True", "__init__", "bool", "class", "cos_theta", "def", "dim", "flatten", "forward", "functional", "hidden_states", "input_dim", "labels", "logits", "loss", "margin", "mm", "nn", "normalize", "num_labels", "one_hot", "onehot", "psi", "randn", "requires_grad", "return", "scale", "self", "super", "torch", "weight", "where"], "wav2vec2/modeling_wav2vec2.py:TDNNLayer": ["Detected", "Linear", "LoRA", "LoraLayer", "Model", "Module", "ReLU", "Tensor", "You", "__init__", "activation", "applied", "be", "bias", "class", "config", "conv1d", "def", "dilation", "due", "else", "exclude", "forward", "from", "functional", "hidden_states", "if", "in_conv_dim", "is_peft_available", "isinstance", "kernel", "kernel_size", "layer_id", "modules", "nn", "on", "optimization", "out_conv_dim", "return", "s", "self", "should", "super", "t", "target", "tdnn_dilation", "tdnn_dim", "tdnn_kernel", "to", "torch", "transpose", "view", "warn", "warnings", "weight", "weights", "won"], "wav2vec2/modeling_wav2vec2.py:Wav2Vec2ForXVector": ["AMSoftmaxLoss", "False", "FutureWarning", "Linear", "LongTensor", "Model", "ModelForXVector", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "Please", "TDNNLayer", "Tensor", "The", "Transformers", "True", "Union", "XVectorOutput", "_HIDDEN_STATES_START_POSITION", "__init__", "_conv_out_length", "_freeze_parameters", "_get_feat_extract_output_lengths", "_get_tdnn_output_lengths", "and", "append", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cat", "class", "classifier", "config", "def", "deprecated", "dim", "else", "embeddings", "enumerate", "equivalent", "feat_extract_output_lengths", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hidden_size", "hidden_states", "i", "if", "in", "init_weights", "input_length", "input_lengths", "input_values", "instead", "int", "is", "kernel_size", "labels", "layer_weights", "len", "length", "logits", "loss", "mean", "mean_features", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "objective", "ones", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "parameters", "projector", "r", "range", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "statistic_pooling", "std", "std_features", "stride", "sum", "super", "tdnn", "tdnn_dim", "tdnn_kernel", "tdnn_layer", "tdnn_layers", "tdnn_output_lengths", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will", "xvector_output_dim"], "prophetnet/modeling_prophetnet.py:softmax": ["False", "Model", "def", "dim", "dtype", "else", "float", "float32", "functional", "hidden_state", "if", "nn", "onnx_trace", "return", "torch"], "prophetnet/modeling_prophetnet.py:ngram_attention_bias": ["False", "Model", "Model_attention_bias", "cat", "clone", "def", "detach", "device", "dim", "dtype", "fill_diagonal_", "finfo", "for", "in", "left_block", "min", "ones", "range", "return", "right_block", "sequence_length", "stream_idx", "torch", "triu_", "wrap"], "prophetnet/modeling_prophetnet.py:compute_relative_buckets": ["False", "Model_relative_buckets", "abs", "def", "else", "float", "if", "int", "inv_relative_positions", "is_bidirectional", "is_small", "log", "lt", "math", "max", "max_distance", "max_exact", "min", "num_buckets", "ones_like", "rel_positions_bucket", "relative_positions", "return", "torch", "val_if_large", "where", "zeros_like"], "prophetnet/modeling_prophetnet.py:compute_all_stream_relative_buckets": ["False", "Model_all_stream_relative_buckets", "Model_relative_buckets", "cat", "def", "dim", "is_bidirectional", "main_relative_position_buckets", "main_stream_relative_positions", "max_distance", "num_buckets", "position_ids", "predict_relative_position_buckets", "predicting_stream_relative_positions", "repeat", "return", "size", "torch", "unsqueeze"], "prophetnet/modeling_prophetnet.py:ProphetNetSeq2SeqLMOutput": ["Cache", "FloatTensor", "FutureWarning", "ModelOutput", "ModelSeq2SeqLMOutput", "None", "Optional", "Please", "and", "be", "class", "cross_attentions", "decoder_attentions", "decoder_cross_attentions", "decoder_hidden_states", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "def", "deprecated", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "instead", "is", "logits", "logits_ngram", "loss", "past_key_values", "property", "r", "removed", "return", "self", "soon", "torch", "tuple", "use", "warn", "warnings", "will"], "prophetnet/modeling_prophetnet.py:ProphetNetSeq2SeqModelOutput": ["Cache", "FloatTensor", "FutureWarning", "ModelOutput", "ModelSeq2SeqModelOutput", "None", "Optional", "Please", "and", "be", "class", "cross_attentions", "decoder_attentions", "decoder_cross_attentions", "decoder_hidden_states", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "def", "deprecated", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "instead", "is", "last_hidden_state", "last_hidden_state_ngram", "past_key_values", "property", "r", "removed", "return", "self", "soon", "torch", "tuple", "use", "warn", "warnings", "will"], "prophetnet/modeling_prophetnet.py:ProphetNetDecoderModelOutput": ["Cache", "FloatTensor", "ModelDecoderModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "hidden_states_ngram", "last_hidden_state", "last_hidden_state_ngram", "ngram_attentions", "past_key_values", "r", "torch", "tuple"], "prophetnet/modeling_prophetnet.py:ProphetNetDecoderLMOutput": ["Cache", "FloatTensor", "ModelDecoderLMOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "hidden_states_ngram", "logits", "logits_ngram", "loss", "ngram_attentions", "past_key_values", "r", "torch", "tuple"], "prophetnet/modeling_prophetnet.py:ProphetNetPreTrainedModel": ["Embedding", "In", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "See", "True", "Verify", "_init_weights", "_shift_right", "all", "assert", "base_model_prefix", "be", "bias", "class", "clone", "config", "data", "decoder_start_token_id", "def", "defined", "docs", "elif", "for", "has", "if", "information", "init_std", "input_ids", "is", "isinstance", "it", "item", "masked_fill_", "mean", "model", "module", "more", "new_zeros", "nn", "normal_", "not", "only", "pad_token_id", "padding_idx", "positive", "return", "self", "set", "shape", "shifted_input_ids", "std", "supports_gradient_checkpointing", "that", "the", "to", "torch", "usually", "values", "weight", "zero_"], "prophetnet/modeling_prophetnet.py:ProphetNetPositionalEmbeddings": ["Embedding", "If", "ModelConfig", "ModelPositionalEmbeddings", "None", "__init__", "_forward", "and", "assert", "attention_mask", "be", "clamp", "class", "computed", "config", "cumsum", "def", "device", "dim", "dtype", "else", "forward", "get_seq_length", "hidden_size", "if", "inputs_shape", "int", "is", "long", "max_length", "max_position_embeddings", "nn", "not", "num_input_ids", "ones", "or", "pad_token_id", "padding_idx", "past_key_values", "position_ids", "pre", "prev_num_input_ids", "return", "self", "set", "should", "super", "then", "torch", "type_as"], "prophetnet/modeling_prophetnet.py:ProphetNetAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Size", "Tensor", "True", "ValueError", "__init__", "a", "and", "assert", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bool", "bsij", "bsik", "bsjk", "but", "by", "cache_position", "class", "config", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "einsum", "else", "expected_shape", "f", "for", "forward", "functional", "get", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_updated", "isinstance", "key_proj", "key_states", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "list", "mask", "must", "new_name", "nn", "not", "num_attn_heads", "num_decoder_attention_heads", "num_encoder_attention_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "query_proj", "query_states", "raise", "reshape", "return", "self", "self_attention_cache", "shape", "should", "single", "size", "softmax", "src_len", "states", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "value_proj", "value_states", "values", "version", "view", "weights"], "prophetnet/modeling_prophetnet.py:ProphetNetFeedForward": ["ACT2FN", "Linear", "ModelConfig", "ModelFeedForward", "Module", "__init__", "activation_dropout", "activation_fn", "activation_function", "class", "config", "def", "dropout", "ffn_dim", "forward", "functional", "hidden_size", "hidden_states", "int", "intermediate", "nn", "output", "p", "return", "self", "super", "training"], "prophetnet/modeling_prophetnet.py:ProphetNetNgramSelfAttention": ["Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelConfig", "ModelNgramSelfAttention", "Module", "None", "Optional", "They", "True", "__init__", "_shape", "a", "arange", "are", "assert", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bncs", "bnhsc", "bnhtc", "bnhts", "bntc", "bnts", "but", "by", "cache_position", "cat", "chunk", "class", "compute_relative_buckets", "config", "contiguous", "curr_past_key_value", "def", "deprecate_kwarg", "device", "dim", "divisible", "dropout", "dtype", "einsum", "else", "extended_predict_attention_mask", "f", "for", "format", "forward", "functional", "gather", "get_main_relative_pos_embeddings", "get_predict_relative_pos_embeddings", "head_dim", "hidden_size", "hidden_states", "hidden_states_list", "hidden_states_predict_list", "if", "in", "incorrect", "index", "is", "isinstance", "key", "key_proj", "key_sequence_length", "key_states", "key_states_list", "layer", "layer_head_mask", "layer_idx", "list", "long", "main_attn_output", "main_attn_probs", "main_attn_weights", "main_hidden_states", "main_key_states", "main_query_states", "main_relative_pos_embeddings", "main_relative_position_buckets", "main_value_states", "mask", "must", "new_name", "ngram", "ngram_sequence_length", "nn", "not", "num_attn_heads", "num_buckets", "num_decoder_attention_heads", "of", "onnx_trace", "out_proj", "p", "past_key_value", "past_key_values", "permute", "position_ids", "predict_attn_output", "predict_attn_probs", "predict_attn_weights", "predict_hidden_states", "predict_key_states", "predict_key_states_list", "predict_query_states", "predict_query_states_list", "predict_relative_pos_embeddings", "predict_relative_position_buckets", "predict_value_states", "predict_value_states_list", "prepare_for_onnx_export_", "proj_shape", "query_proj", "query_states", "query_states_list", "rel_pos_embeddings", "relative_max_distance", "relative_pos_embeddings", "relative_positions", "repeat", "reshape", "return", "self", "self_attention_cache", "seq_len", "sequence_length", "shape", "should", "single", "size", "softmax", "src_len", "stack", "super", "tensor", "tgt_len", "the", "to", "torch", "training", "transpose", "type_as", "unsqueeze", "update", "v_p", "value_proj", "value_states", "value_states_list", "version", "view"], "prophetnet/modeling_prophetnet.py:ProphetNetEncoderLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelFeedForward", "__init__", "attention_mask", "attention_output", "attn_weights", "bool", "class", "config", "def", "encoder_ffn_dim", "feed_forward", "feed_forward_layer_norm", "feed_forward_output", "forward", "hidden_size", "hidden_states", "if", "layer_head_mask", "num_encoder_attention_heads", "output_attentions", "outputs", "return", "self", "self_attn", "self_attn_layer_norm", "super"], "prophetnet/modeling_prophetnet.py:ProphetNetDecoderLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFeedForward", "ModelNgramSelfAttention", "None", "Optional", "Tensor", "True", "__init__", "add_cross_attention", "attention_mask", "attention_output", "bool", "cache_position", "class", "config", "cross_attn", "cross_attn_layer_head_mask", "cross_attn_layer_norm", "cross_attn_weights", "decoder_ffn_dim", "def", "deprecate_kwarg", "encoder_attn_mask", "encoder_hidden_states", "extended_predict_attention_mask", "feed_forward", "feed_forward_layer_norm", "feed_forward_output", "forward", "hidden_size", "hidden_states", "if", "is", "key_value_states", "layer_head_mask", "layer_idx", "main_relative_position_buckets", "new_name", "ngram_attention_output", "not", "num_decoder_attention_heads", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "predict_relative_position_buckets", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "self_attn_weights_ngram", "super", "torch", "use_cache", "version"], "prophetnet/modeling_prophetnet.py:ProphetNetEncoder": ["BaseModelOutput", "Either", "Embedding", "False", "LayerNorm", "Make", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalEmbeddings", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "The", "Union", "ValueError", "_", "__init__", "all_attentions", "and", "assert", "attention_mask", "attentions", "auto_docstring", "be", "bool", "but", "class", "config", "def", "device", "dropout", "dtype", "elif", "else", "embeddings_layer_norm", "encoder_hidden_states", "encoder_layer", "enumerate", "extended_attention_mask", "f", "finfo", "for", "forward", "functional", "get_input_embeddings", "gradient_checkpointing", "has", "head_mask", "hidden_size", "hidden_states", "idx", "if", "in", "input_ids", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layers", "len", "min", "nn", "not", "num_encoder_attention_heads", "num_encoder_layers", "only", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "passed", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "repeat", "return", "return_dict", "self", "set_input_embeddings", "shape", "should", "size", "specified", "super", "sure", "to", "torch", "training", "tuple", "use_return_dict", "v", "value", "vocab_size", "word_embeddings"], "prophetnet/modeling_prophetnet.py:ProphetNetDecoder": ["At", "Cache", "DynamicCache", "Either", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "Make", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderModelOutput", "ModelPositionalEmbeddings", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_forward", "a", "add_cross_attention", "all_cross_attns", "all_main_stream_attns", "all_main_stream_hidden_states", "all_ngram_stream_attns", "all_ngram_stream_hidden_states", "an", "and", "arange", "assert", "attention_mask", "attentions", "attn_mask", "auto_docstring", "batch_size", "be", "bool", "but", "cache_position", "cat", "causal_mask", "checkpointing", "class", "compute_all_stream_relative_buckets", "compute_buffered_relative_buckets", "config", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "def", "deprecated", "device", "dim", "dropout", "dtype", "e", "elif", "else", "embeddings_layer_norm", "encoder_attention_mask", "encoder_attn_mask", "encoder_hidden_states", "enumerate", "expand", "extended_attention_mask", "extended_causal_mask", "extended_encoder_attention_mask", "extended_predict_attention_mask", "extended_predict_causal_mask", "f", "finfo", "for", "forward", "from_legacy_cache", "full", "functional", "g", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "has", "head_mask", "hidden_size", "hidden_states", "hidden_states_ngram", "i", "idx", "if", "in", "incompatible", "input_ids", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "last_hidden_state_ngram", "layer_head_mask", "layer_idx", "layer_outputs", "layers", "len", "length", "logger", "main_relative_buckets", "main_relative_position_buckets", "main_stream_pos_embed", "mask_name", "max_position_embeddings", "max_target_positions", "min", "moment", "ngram", "ngram_attention_bias", "ngram_attentions", "ngram_embeddings", "ngram_hidden_states", "nn", "not", "num_buckets", "num_decoder_attention_heads", "num_decoder_layers", "of", "only", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "passed", "past_key_values", "past_key_values_length", "position_embeddings", "position_ids", "post_init", "predict_causal_mask", "predict_relative_buckets", "predict_relative_position_buckets", "predicting_stream_pos_embed", "prepare_attention_mask", "prepare_predict_attention_mask", "r", "raise", "range", "relative_max_distance", "removed", "repeat", "return", "return_dict", "self", "seq_length", "sequence_length", "set_input_embeddings", "shape", "should", "size", "specified", "super", "supported", "sure", "the", "to", "torch", "training", "triu", "tuple", "use_cache", "use_return_dict", "v", "v4", "value", "vocab_size", "warning_once", "weight", "will", "with", "word_embeddings", "zeros_like", "zip"], "prophetnet/modeling_prophetnet.py:ProphetNetModel": ["BoolTensor", "Cache", "Embedding", "False", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelSeq2SeqModelOutput", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "decoder_outputs", "deepcopy", "def", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "hidden_states_ngram", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "last_hidden_state", "last_hidden_state_ngram", "ngram_attentions", "nn", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "tie_encoder_decoder", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight", "word_embeddings"], "prophetnet/modeling_prophetnet.py:ProphetNetForConditionalGeneration": ["BoolTensor", "Cache", "False", "GenerationMixin", "Linear", "Model", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "ModelSeq2SeqLMOutput", "None", "Optional", "Tensor", "True", "Union", "__init__", "_compute_loss", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "all_logits", "and", "attention_mask", "auto_docstring", "batch_size", "bias", "bool", "break", "cache_position", "class", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "def", "dim", "disable_ngram_loss", "dtype", "else", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eps", "eps_i", "expend_targets", "fill_", "float32", "for", "forward", "functional", "get_decoder", "get_encoder", "get_input_embeddings", "head_mask", "hidden_size", "i", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "is", "is_contiguous", "keepdim", "kwargs", "labels", "lm_head", "log_softmax", "logits", "logits_ngram", "loss", "lprobs", "mean", "ne", "new_zeros", "ngram", "nll_loss", "nn", "non_masked_tokens", "not", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "padding_idx", "past_key_values", "post_init", "predict_logits", "predicting_streams", "prepare_decoder_input_ids_from_labels", "r", "range", "reduction", "return", "return_dict", "self", "sequence_length", "shape", "size", "smooth_loss", "sum", "super", "tie_word_embeddings", "torch", "transpose", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "weight", "word_embeddings"], "prophetnet/modeling_prophetnet.py:ProphetNetForCausalLM": ["Cache", "False", "GenerationMixin", "Linear", "Model", "ModelConfig", "ModelDecoderLMOutput", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_compute_loss", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "all_logits", "and", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "break", "cache_position", "class", "config", "contiguous", "copy", "cross_attentions", "cross_attn_head_mask", "decoder", "deepcopy", "def", "dim", "disable_ngram_loss", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "eps_i", "expend_targets", "fill_", "float32", "for", "forward", "functional", "get_decoder", "get_input_embeddings", "get_seq_length", "head_mask", "hidden_size", "hidden_states", "hidden_states_ngram", "i", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "items", "keepdim", "key", "kwargs", "labels", "lm_head", "log_softmax", "logits", "logits_ngram", "loss", "lprobs", "mean", "model_inputs", "ne", "new_ones", "new_zeros", "ngram", "ngram_attentions", "nll_loss", "nn", "non_masked_tokens", "not", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "padding_idx", "past_key_values", "pop", "post_init", "predict_logits", "predicting_streams", "prepare_inputs_for_generation", "r", "range", "reduction", "return", "return_dict", "self", "sequence_length", "set_decoder", "set_input_embeddings", "shape", "size", "smooth_loss", "sum", "super", "tie_word_embeddings", "torch", "transpose", "tuple", "use_cache", "use_return_dict", "v", "value", "view", "vocab_size", "weight", "word_embeddings"], "prophetnet/modeling_prophetnet.py:ProphetNetDecoderWrapper": ["Embedding", "ModelConfig", "ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "_tie_or_clone_weights", "_tie_weights", "args", "class", "config", "decoder", "def", "forward", "get_input_embeddings", "hidden_size", "kwargs", "nn", "pad_token_id", "padding_idx", "post_init", "return", "self", "super", "vocab_size", "word_embeddings"], "qwen2_moe/modeling_qwen2_moe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen2_moe/modeling_qwen2_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen2_moe/modeling_qwen2_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen2_moe/modeling_qwen2_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "errors", "f", "float32", "forward", "functional", "getattr", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "logger", "make", "math", "matmul", "max_position_embeddings", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "qkv_bias", "query_states", "raise", "recommended", "repeat_kv", "reshape", "return", "rope_theta", "rotary_emb", "self", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "The", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "and", "apply_rotary_pos_emb", "args", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache_kwargs", "cache_position", "cast", "casted", "class", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "hasattr", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "max_window_layers", "might", "mps", "new_name", "norm", "not", "num_key_value_groups", "o_proj", "or", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "related", "repeat_kv", "reshape", "return", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "super", "target_dtype", "the", "this", "to", "torch", "training", "transpose", "tuple", "type", "upcasted", "update", "use_cache", "use_sliding_window", "use_top_left_mask", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "you"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_size", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_key_value_groups", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeSparseMoeBlock": ["F", "False", "Linear", "ModelMLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "keepdim", "moe_intermediate_size", "nn", "nonzero", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "shared_expert", "shared_expert_gate", "shared_expert_intermediate_size", "shared_expert_output", "sigmoid", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSparseMoeBlock", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "and", "attention_mask", "bool", "cache_position", "class", "config", "decoder_sparse_step", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "int", "intermediate_size", "isinstance", "kwargs", "layer_idx", "mlp", "mlp_only_layers", "new_name", "not", "num_experts", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoePreTrainedModel": ["Embedding", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeModel": ["Attention", "AttentionMaskConverter", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "Flash", "FloatTensor", "LongTensor", "Make", "Model", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "arange", "are", "attempting", "attention_mask", "attentions", "auto_docstring", "batch_size", "batched", "before", "behaviour", "bitwise_or_", "bool", "cache_position", "call", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal_attend_mask", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "generation", "get_max_cache_shape", "get_seq_length", "get_text_config", "getattr", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_padding_right", "is_sliding", "is_static_sliding_cache", "is_training", "isinstance", "item", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "lead", "left", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "may", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "padding_side", "past_key_values", "past_key_values_length", "past_seen_tokens", "perform", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "right", "rms_norm_eps", "rotary_emb", "router_logits", "sdpa", "self", "sequence_length", "shape", "size", "sliding_attend_mask", "sliding_window", "specify", "staticmethod", "sum", "super", "sure", "target_length", "text_config", "the", "this", "to", "tokenizer", "tokenizing", "torch", "training", "type", "unexpected", "unsqueeze", "use_cache", "use_sliding_window", "using_static_cache", "version", "vocab_size", "warning_once", "with", "xpu"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class"], "qwen2_moe/modeling_qwen2_moe.py:Qwen2MoeForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "class"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackbonePatchEmbeddings": ["Conv2d", "Input", "Iterable", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "class", "collections", "config", "def", "doesn", "else", "embed_dim", "embeddings", "f", "flatten", "forward", "height", "hidden_size", "if", "image", "image_size", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "or", "padding", "patch_size", "pixel_values", "projection", "raise", "return", "self", "shape", "size", "stride", "super", "t", "torch", "transpose", "width"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneEmbeddings": ["Dropout", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "Parameter", "Tensor", "__init__", "class", "config", "def", "dropout", "embeddings", "forward", "hidden_dropout_prob", "hidden_size", "nn", "num_patches", "patch_embeddings", "pixel_values", "position_embeddings", "return", "self", "super", "torch", "zeros"], "vitpose_backbone/modeling_vitpose_backbone.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneMoeMLP": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelMoeMLP", "Module", "ModuleList", "Tensor", "_", "__init__", "act", "cat", "class", "config", "current_hidden_state", "def", "dim", "drop", "expert_hidden_state", "experts", "fc1", "fc2", "for", "forward", "hidden_act", "hidden_dropout_prob", "hidden_features", "hidden_size", "hidden_state", "i", "in", "in_features", "indices", "int", "mlp_ratio", "nn", "num_experts", "out_features", "part_features", "range", "return", "selected_index", "self", "shared_hidden_state", "super", "torch", "view", "zeros_like"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneMLP": ["ACT2FN", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "True", "__init__", "activation", "bias", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "hidden_size", "hidden_state", "in_features", "int", "mlp_ratio", "nn", "out_features", "return", "self", "super", "torch"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelLayer", "ModelMLP", "ModelMoeMLP", "None", "Optional", "Please", "Tensor", "ValueError", "__init__", "and", "attention", "attention_output", "be", "class", "config", "dataset_index", "def", "else", "eps", "experts", "f", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "if", "indices", "is", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "mlp", "multiple", "must", "nn", "num_experts", "pass", "provide", "provided", "raise", "return", "self", "super", "the", "to", "torch", "using", "when"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackboneEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_hidden_states", "append", "bool", "class", "config", "dataset_index", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "output_hidden_states", "range", "return", "self", "super", "torch", "tuple"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackbonePreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "dtype", "elif", "fill_", "float32", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "vitpose_backbone/modeling_vitpose_backbone.py:VitPoseBackbone": ["BackboneMixin", "BackboneOutput", "BaseModelOutput", "LayerNorm", "Model", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "_", "__init__", "_init_backbone", "append", "auto_docstring", "bool", "check_model_inputs", "class", "config", "dataset_index", "def", "else", "embedding_output", "embeddings", "encoder", "eps", "feature_maps", "for", "forward", "get_head_mask", "head_mask", "hidden_size", "hidden_state", "hidden_states", "if", "in", "is", "kwargs", "layer_norm_eps", "layernorm", "nn", "num_features", "num_hidden_layers", "out_features", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "range", "return", "self", "stage", "stage_names", "super", "torch", "tuple", "zip"], "sam2_video/modeling_sam2_video.py:Sam2VideoInferenceCache": ["ModelInferenceCache", "None", "Optional", "Tensor", "True", "Union", "__init__", "_vision_features", "and", "cache_vision_features", "cached", "class", "clear", "clear_all", "cpu", "def", "device", "dict", "elif", "else", "features", "for", "frame_idx", "get_vision_features", "if", "in", "inference_device", "inference_state_device", "int", "isinstance", "items", "key", "keys", "len", "list", "max_vision_features_cache_size", "min", "moved", "non_blocking", "not", "pop", "return", "self", "str", "to", "torch", "tuple", "v", "value"], "sam2_video/modeling_sam2_video.py:Sam2VideoInferenceSession": ["FloatTensor", "ModelInferenceCache", "ModelInferenceSession", "None", "Optional", "OrderedDict", "Tensor", "True", "Union", "__init__", "_obj_id_to_idx", "_obj_idx_to_id", "add_mask_inputs", "add_new_frame", "add_point_inputs", "and", "append", "bool", "cache", "class", "clear", "clear_all", "cond_frame_outputs", "cpu", "def", "device", "device_inputs", "dict", "dim", "dtype", "elif", "else", "float32", "for", "frame_idx", "frames_tracked_per_obj", "get", "get_frame", "get_obj_num", "get_output", "if", "in", "inference_device", "inference_state_device", "inputs", "int", "is", "is_conditioning_frame", "isinstance", "items", "key", "len", "list", "mask_inputs_per_obj", "max_vision_features_cache_size", "non_blocking", "non_cond_frame_outputs", "not", "num_frames", "obj_id", "obj_id_to_idx", "obj_ids", "obj_idx", "obj_idx_to_id", "obj_with_new_inputs", "object_pointer", "object_score_logits", "out", "output_dict_per_obj", "output_key", "output_value", "pixel_values", "point_inputs_per_obj", "pop", "processed_frames", "property", "r", "remove_mask_inputs", "remove_point_inputs", "reset_inference_session", "reset_tracking_data", "return", "self", "squeeze", "storage_key", "store_output", "str", "to", "torch", "value", "video", "video_height", "video_storage_device", "video_width"], "sam2_video/modeling_sam2_video.py:Sam2VideoLayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "sam2_video/modeling_sam2_video.py:Sam2VideoPositionEmbeddingSine": ["False", "ModelPositionEmbeddingSine", "Module", "None", "Optional", "Size", "Tensor", "True", "Union", "ValueError", "__init__", "and", "arange", "be", "bool", "cat", "class", "compile_compatible_method_lru_cache", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "else", "eps", "flatten", "float", "floor", "forward", "if", "int", "int64", "is", "mask", "math", "maxsize", "nn", "normalize", "not", "not_mask", "num_pos_feats", "passed", "permute", "pi", "pos", "pos_x", "pos_y", "raise", "return", "rounding_mode", "scale", "self", "shape", "should", "sin", "stack", "str", "super", "temperature", "to", "torch", "x_embed", "y_embed", "zeros"], "sam2_video/modeling_sam2_video.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "sam2_video/modeling_sam2_video.py:Sam2VideoAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "attention_downModelple_rate", "attention_interface", "attention_mask", "attention_similarity", "attn_output", "attn_weights", "batch_size", "class", "config", "contiguous", "def", "downModelple_rate", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_size", "if", "internal_dim", "is", "is_causal", "k_proj", "key", "kwargs", "new_shape", "nn", "num_attention_heads", "o_proj", "point_batch_size", "q_proj", "query", "reshape", "return", "scaling", "self", "shape", "super", "torch", "transpose", "tuple", "v_proj", "value", "view"], "sam2_video/modeling_sam2_video.py:Sam2VideoTwoWayAttentionBlock": ["False", "LayerNorm", "ModelAttention", "ModelFeedForward", "ModelMaskDecoderConfig", "ModelTwoWayAttentionBlock", "Module", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_similarity", "attn_out", "bool", "class", "config", "cross_attn_image_to_token", "cross_attn_token_to_image", "def", "downModelple_rate", "else", "forward", "hidden_size", "if", "key", "key_point_embedding", "keys", "kwargs", "layer_norm1", "layer_norm2", "layer_norm3", "layer_norm4", "mlp", "mlp_dim", "mlp_out", "nn", "num_hidden_layers", "num_layers", "queries", "query", "query_point_embedding", "return", "self", "self_attn", "skip_first_layer_pe", "super", "value"], "sam2_video/modeling_sam2_video.py:Sam2VideoFeedForward": ["ACT2FN", "F", "False", "Linear", "ModelFeedForward", "Module", "ModuleList", "_", "__init__", "activation", "bool", "class", "def", "for", "forward", "hidden_dim", "hidden_states", "if", "in", "input_dim", "int", "layer", "layers", "nn", "num_layers", "output_dim", "proj_in", "proj_out", "range", "relu", "return", "self", "sigmoid", "sigmoid_output", "str", "super"], "sam2_video/modeling_sam2_video.py:Sam2VideoImageSegmentationOutput": ["FloatTensor", "ModelImageSegmentationOutput", "ModelOutput", "None", "Optional", "class", "high_res_masks", "image_embeddings", "iou_scores", "mask_decoder_attentions", "object_pointer", "object_score_logits", "pred_masks", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states"], "sam2_video/modeling_sam2_video.py:Sam2VideoSegmentationOutput": ["FloatTensor", "ModelOutput", "ModelSegmentationOutput", "None", "Optional", "class", "frame_idx", "int", "pred_masks", "r", "torch"], "sam2_video/modeling_sam2_video.py:Sam2VideoPreTrainedModel": ["Conv2d", "ConvTranspose2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLayerNorm", "ModelMemoryFuserCXBlock", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_supports_attention_backend", "_supports_flash_attn_2", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "config_class", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "memory_temporal_positional_encoding", "module", "nn", "no_memory_positional_encoding", "no_object_pointer", "normal_", "not", "occlusion_spatial_embedding_parameter", "padding_idx", "pixel_values", "scale", "self", "std", "weight", "zero_"], "sam2_video/modeling_sam2_video.py:Sam2VideoVisionRotaryEmbedding": ["Dimension", "False", "ModelConfig", "ModelVisionRotaryEmbedding", "Module", "RoPE", "Tensor", "ValueError", "__init__", "arange", "axial", "be", "by", "cat", "class", "config", "cos", "def", "dim", "div", "divisible", "dtype", "end_x", "end_y", "flattened_indices", "float", "floor", "for", "forward", "freqs", "freqs_x", "freqs_y", "if", "inv_freq", "long", "memory_attention_downModelple_rate", "memory_attention_hidden_size", "memory_attention_num_attention_heads", "memory_attention_rope_feat_sizes", "memory_attention_rope_theta", "must", "nn", "no_grad", "outer", "persistent", "raise", "register_buffer", "repeat_interleave", "return", "rope_embeddings_cos", "rope_embeddings_sin", "rounding_mode", "self", "sin", "super", "torch", "tuple", "x_positions", "y_positions"], "sam2_video/modeling_sam2_video.py:rotate_pairwise": ["Model_pairwise", "def", "dim", "flatten", "return", "shape", "stack", "start_dim", "torch", "unbind", "view", "x", "x1", "x2"], "sam2_video/modeling_sam2_video.py:apply_rotary_pos_emb_2d": ["False", "Model_rotary_pos_emb_2d", "Tensor", "and", "bool", "cat", "cos", "cos_k", "def", "dim", "else", "float", "if", "int", "k", "k_embed", "k_pass", "k_rot", "num_k_exclude_rope", "q", "q_embed", "repeat", "repeat_factor", "repeat_freqs_k", "return", "rotate_pairwise", "shape", "sin", "sin_k", "torch", "tuple", "type_as"], "sam2_video/modeling_sam2_video.py:Sam2VideoRoPEAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelConfig", "ModelRoPEAttention", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb_2d", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "class", "config", "contiguous", "cos", "def", "dropout", "dropout_p", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_size", "if", "int", "internal_dim", "is", "is_causal", "k_proj", "key", "kv_in_dim", "kwargs", "memory_attention_downModelple_rate", "memory_attention_hidden_size", "memory_attention_num_attention_heads", "memory_attention_rope_dropout", "new_shape", "nn", "not", "num_attention_heads", "num_k_exclude_rope", "o_proj", "point_batch_size", "position_embeddings", "q_proj", "query", "repeat_freqs_k", "reshape", "return", "rope_k_repeat", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "v_proj", "value", "view"], "sam2_video/modeling_sam2_video.py:Sam2VideoMemoryAttentionLayer": ["ACT2FN", "Dropout", "LayerNorm", "Linear", "ModelConfig", "ModelMemoryAttentionLayer", "ModelRoPEAttention", "Module", "Tensor", "True", "_", "__init__", "activation", "class", "config", "cross_attn_image", "def", "dropout", "dropout1", "dropout2", "dropout3", "forward", "hidden_size", "int", "key", "key_point_embedding", "keys", "kv_in_dim", "layer_norm1", "layer_norm2", "layer_norm3", "linear1", "linear2", "memory_attention_dropout", "memory_attention_feed_forward_hidden_act", "memory_attention_feed_forward_hidden_size", "memory_attention_hidden_size", "nn", "num_k_exclude_rope", "position_embeddings", "queries", "query", "return", "rope_k_repeat", "rope_position_embeddings", "self", "self_attn", "super", "torch", "tuple", "value"], "sam2_video/modeling_sam2_video.py:Sam2VideoMemoryAttention": ["LayerNorm", "ModelConfig", "ModelMemoryAttention", "ModelMemoryAttentionLayer", "ModelVisionRotaryEmbedding", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "current_vision_features", "current_vision_position_embeddings", "def", "else", "for", "forward", "if", "in", "int", "is", "key_point_embedding", "keys", "layer", "layer_norm", "layers", "memory", "memory_attention_hidden_size", "memory_attention_num_layers", "memory_posision_embeddings", "ndim", "nn", "normed_output", "not", "num_k_exclude_rope", "num_object_pointer_tokens", "output", "queries", "range", "return", "rope_position_embeddings", "rotary_emb", "self", "super", "torch", "transpose", "unsqueeze"], "sam2_video/modeling_sam2_video.py:Sam2VideoMemoryFuserCXBlock": ["ACT2FN", "Conv2d", "GradientCheckpointingLayer", "Linear", "ModelConfig", "ModelLayerNorm", "ModelMemoryFuserCXBlock", "Parameter", "True", "__init__", "activation", "channels_first", "class", "config", "data_format", "def", "depthwise_conv", "eps", "forward", "groups", "hidden_states", "input", "kernel_size", "layer_norm", "memory_fuser_embed_dim", "memory_fuser_hidden_act", "memory_fuser_intermediate_dim", "memory_fuser_kernel_size", "memory_fuser_layer_scale_init_value", "memory_fuser_padding", "nn", "ones", "padding", "permute", "pointwise_conv1", "pointwise_conv2", "requires_grad", "return", "scale", "self", "super", "torch"], "sam2_video/modeling_sam2_video.py:Sam2VideoMemoryFuser": ["ModelConfig", "ModelMemoryFuser", "ModelMemoryFuserCXBlock", "Module", "ModuleList", "_", "__init__", "class", "config", "def", "for", "forward", "hidden_states", "in", "layer", "layers", "memory_fuser_num_layers", "nn", "range", "return", "self", "super"], "sam2_video/modeling_sam2_video.py:Sam2VideoMaskDownSamplerLayer": ["ACT2FN", "Conv2d", "ModelConfig", "ModelLayerNorm", "ModelMaskDownModelplerLayer", "Module", "__init__", "activation", "channels_first", "class", "config", "conv", "data_format", "def", "eps", "forward", "in_channels", "int", "kernel_size", "layer_norm", "mask_downModelpler_hidden_act", "mask_downModelpler_kernel_size", "mask_downModelpler_padding", "mask_downModelpler_stride", "nn", "out_channels", "padding", "return", "self", "stride", "super", "x"], "sam2_video/modeling_sam2_video.py:Sam2VideoMaskDownSampler": ["ACT2FN", "Conv2d", "ModelConfig", "ModelMaskDownModelpler", "ModelMaskDownModelplerLayer", "Module", "ModuleList", "_", "__init__", "activation", "append", "class", "config", "def", "final_conv", "for", "forward", "in", "int", "kernel_size", "layer", "layers", "log2", "mask_downModelpler_embed_dim", "mask_downModelpler_hidden_act", "mask_downModelpler_stride", "mask_downModelpler_total_stride", "mask_in_chans", "mask_out_chans", "math", "nn", "num_layers", "range", "return", "self", "super", "x"], "sam2_video/modeling_sam2_video.py:Sam2VideoMemoryEncoder": ["Conv2d", "ModelConfig", "ModelMaskDownModelpler", "ModelMemoryEncoder", "ModelMemoryFuser", "ModelPositionEmbeddingSine", "Module", "Tensor", "True", "__init__", "class", "config", "def", "device", "dtype", "feature_projection", "forward", "hidden_size", "kernel_size", "mask_downModelpler", "masks", "memory_encoder_hidden_size", "memory_encoder_output_channels", "memory_fuser", "nn", "normalize", "num_pos_feats", "output_channels", "position_encoding", "projection", "return", "self", "shape", "super", "torch", "tuple", "vision_features", "vision_pos_enc"], "sam2_video/modeling_sam2_video.py:Sam2VideoVisionEncoderOutput": ["FloatTensor", "ModelOutput", "ModelVisionEncoderOutput", "None", "Optional", "attentions", "class", "fpn_hidden_states", "fpn_position_encoding", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "sam2_video/modeling_sam2_video.py:Sam2VideoPositionalEmbedding": ["ModelPositionalEmbedding", "ModelPromptEncoderConfig", "Module", "None", "__init__", "cat", "class", "clone", "config", "coordinates", "cos", "def", "dim", "dtype", "float32", "forward", "hidden_size", "if", "input_coords", "input_shape", "is", "nn", "not", "np", "pi", "positional_embedding", "randn", "register_buffer", "return", "scale", "self", "sin", "super", "to", "torch"], "sam2_video/modeling_sam2_video.py:Sam2VideoMaskEmbedding": ["ACT2FN", "Conv2d", "ModelLayerNorm", "ModelMaskEmbedding", "ModelPromptEncoderConfig", "Module", "__init__", "activation", "channels_first", "class", "config", "conv1", "conv2", "conv3", "data_format", "def", "dense_embeddings", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "layer_norm_eps", "mask_input_channels", "masks", "nn", "return", "self", "stride", "super"], "sam2_video/modeling_sam2_video.py:Sam2VideoPromptEncoder": ["Embedding", "If", "ModelMaskEmbedding", "ModelPositionalEmbedding", "ModelPromptEncoder", "ModelPromptEncoderConfig", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_embed_boxes", "_embed_points", "also", "are", "batch_size", "be", "bool", "box_embeddings", "boxes", "cat", "clamp", "class", "config", "constant", "coords", "corner_embedding", "def", "dense_embeddings", "dim", "else", "expand", "expand_as", "forward", "functional", "hidden_size", "if", "image_embedding_size", "image_size", "input_boxes", "input_image_size", "input_labels", "input_masks", "input_points", "input_shape", "is", "labels", "mask_embed", "mask_input_size", "min", "mode", "must", "nn", "no_mask_embed", "not", "not_a_point_embed", "num_point_embeddings", "pad", "patch_size", "point_embed", "point_embedding", "point_embeddings", "points", "provided", "raise", "reshape", "return", "self", "shape", "shared_embedding", "sparse_embeddings", "super", "torch", "tuple", "unsqueeze", "value", "view", "weight", "where", "zeros_like"], "sam2_video/modeling_sam2_video.py:Sam2VideoTwoWayTransformer": ["BaseModelOutput", "LayerNorm", "ModelAttention", "ModelMaskDecoderConfig", "ModelTwoWayAttentionBlock", "ModelTwoWayTransformer", "Module", "ModuleList", "None", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "_", "__init__", "an", "append", "attention_similarity", "attn_out", "class", "config", "def", "final_attn_token_to_image", "flatten", "for", "forward", "have", "hidden_size", "i", "if", "image_embedding", "image_embeddings", "image_positional_embeddings", "in", "is", "key", "key_point_embedding", "keys", "kwargs", "layer", "layer_norm_final_attn", "layers", "nn", "not", "num_hidden_layers", "permute", "point_embeddings", "queries", "query", "query_point_embedding", "raise", "range", "return", "self", "skip_first_layer_pe", "specify", "super", "target_embedding", "to", "tuple", "unsqueeze", "value"], "sam2_video/modeling_sam2_video.py:Sam2VideoMaskDecoder": ["Conv2d", "ConvTranspose2d", "Embedding", "GELU", "ModelFeedForward", "ModelLayerNorm", "ModelMaskDecoder", "ModelMaskDecoderConfig", "ModelTwoWayTransformer", "Model_tokens_out", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "_", "__init__", "_dynamic_multimask_via_stability", "_get_stability_scores", "activation", "all_iou_scores", "all_mask_logits", "and", "area_i", "area_u", "argmax", "attention_similarity", "batch_size", "best_multimask_iou_scores", "best_multimask_logits", "best_scores_inds", "best_scores_inds_expanded", "bool", "cat", "channels_first", "class", "config", "conv_s0", "conv_s1", "current_mlp", "data_format", "def", "dense_prompt_embeddings", "dim", "dtype", "dynamic_multimask_stability_delta", "dynamic_multimask_stability_thresh", "dynamic_multimask_via_stability", "elif", "else", "expand", "expand_as", "feat_s0", "feat_s1", "flatten", "float", "for", "forward", "gather", "height", "hidden_size", "high_resolution_features", "hyper_in", "hyper_in_list", "i", "if", "image_embeddings", "image_positional_embeddings", "in", "iou_head_depth", "iou_head_hidden_dim", "iou_pred", "iou_prediction_head", "iou_scores_out", "iou_token", "iou_token_out", "is_stable", "kernel_size", "kwargs", "list", "mask_logits", "mask_logits_out", "mask_slice", "mask_tokens", "mask_tokens_out", "masks", "mlps_list", "multimask_iou_scores", "multimask_logits", "multimask_output", "nn", "not", "num_channels", "num_mask_tokens", "num_multimask_outputs", "obj_score_token", "object_score_logits", "output_hypernetworks_mlps", "output_tokens", "point_batch_size", "point_embeddings", "pred_obj_score_head", "range", "repeat", "repeat_interleave", "return", "self", "shape", "sigmoid_output", "singlemask_iou_scores", "singlemask_logits", "size", "slice", "sparse_prompt_embeddings", "stability_delta", "stability_scores", "stack", "stride", "sum", "super", "target_embedding", "to", "tokens", "torch", "training", "transformer", "transpose", "tuple", "unsqueeze", "upscale_conv1", "upscale_conv2", "upscale_layer_norm", "upscaled_embedding", "view", "weight", "where", "width"], "sam2_video/modeling_sam2_video.py:get_1d_sine_pe": ["Model_1d_sine_pe", "arange", "cat", "cos", "def", "device", "dim", "dim_t", "dtype", "float32", "pe_dim", "pos_embed", "pos_inds", "return", "sin", "temperature", "torch", "unsqueeze"], "sam2_video/modeling_sam2_video.py:Sam2VideoModel": ["Any", "AutoModel", "Cannot", "Conv2d", "Exactly", "F", "False", "FloatTensor", "Got", "Identity", "Iterator", "Linear", "LongTensor", "ModelConfig", "ModelFeedForward", "ModelImageSegmentationOutput", "ModelInferenceSession", "ModelMaskDecoder", "ModelMemoryAttention", "ModelMemoryEncoder", "ModelModel", "ModelPositionalEmbedding", "ModelPreTrainedModel", "ModelPromptEncoder", "ModelSegmentationOutput", "ModelTwoWayAttentionBlock", "ModelVisionEncoderOutput", "Model_output", "Model_output_token", "Model_output_tokens", "Model_outputs", "Modele", "NO_OBJ_SCORE", "No", "None", "Optional", "OutputRecorder", "Parameter", "Propagate", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "_", "__init__", "_attn_implementation", "_can_record_outputs", "_encode_new_memory", "_keys_to_ignore_on_load_missing", "_keys_to_ignore_on_load_unexpected", "_prepare_memory_conditioned_features", "_prepare_vision_features", "_run_single_frame_inference", "_single_frame_forward", "_tie_weights", "_tied_weights_keys", "_use_mask_as_output", "_use_multimask", "a", "add", "add_new_frame", "align_corners", "all_pred_masks", "and", "antialias", "any", "appear", "append", "arange", "are", "argmax", "as", "attention_similarity", "attentions", "auto_docstring", "backbone_feature_sizes", "backbone_features", "batch_inds", "batch_size", "be", "best_iou_inds", "bfloat16", "bilinear", "bool", "bounding", "box", "boxes", "break", "cache", "cache_vision_features", "cached_features", "cannot", "cat", "channels", "class", "combined_memory", "combined_memory_pos_embed", "combined_memory_positional_embeddings", "cond_frame_outputs", "conditioned_feature_map", "conditioned_feature_map_flat", "conditioning", "conditioning_outputs", "config", "continue", "conv_s0", "conv_s1", "cumsum", "current_feature_map", "current_out", "current_vision_feats", "current_vision_features", "current_vision_pos_embeds", "current_vision_position_embeddings", "current_vision_positional_embeddings", "custom_intro", "data", "def", "dense_embeddings", "dense_prompt_embeddings", "desc", "determine", "device", "dict", "dim", "dtype", "elif", "eligible_conditioning_outputs", "else", "empty", "enable_occlusion_spatial_embedding", "enable_temporal_pos_encoding_for_object_pointers", "end_frame_idx", "expand", "f", "feat", "feat_size", "feature_map", "feature_map_position_embedding", "feature_maps", "feature_maps_position_embeddings", "first", "flatten", "float", "float32", "for", "forward", "fpn_hidden_size", "fpn_hidden_states", "fpn_position_encoding", "frame", "frame_idx", "frames_tracked_per_obj", "frames_with_inputs", "from_config", "get", "get_1d_sine_pe", "get_frame", "get_image_embeddings", "get_image_features", "get_image_wide_positional_embeddings", "get_input_embeddings", "get_obj_num", "get_output", "get_prompt_embeddings", "get_vision_features", "grid", "has_cond_output", "has_new_inputs", "height", "hidden_dim", "hidden_states", "high_res_features", "high_res_masks", "high_res_multimasks", "high_resolution_features", "if", "image_batch", "image_embedding_size", "image_embeddings", "image_positional_embeddings", "image_size", "in", "index", "inference", "inference_mode", "inference_session", "input", "input_boxes", "input_labels", "input_masks", "input_points", "inputs", "int", "int32", "interpolate", "iou_scores", "is", "is_conditioning_frame", "is_init_cond_frame", "is_initial_conditioning_frame", "is_mask_from_pts", "is_obj_appearing", "it", "items", "kernel_size", "kwargs", "lambda_is_obj_appearing", "len", "list", "low_res_masks", "low_res_multimasks", "manually", "many", "mask_decoder", "mask_decoder_attentions", "mask_decoder_config", "mask_downModelple", "mask_for_mem", "mask_input_size", "mask_inputs", "mask_inputs_float", "mask_inputs_per_obj", "maskmem_features", "maskmem_pos_enc", "max", "max_frame_num_to_track", "max_object_pointers_in_encoder", "max_object_pointers_to_use", "max_temporal_diff", "mem_dim", "memories_to_concatenate", "memory", "memory_attention", "memory_encoder", "memory_encoder_output_channels", "memory_features", "memory_posision_embeddings", "memory_positional_embeddings_to_concatenate", "memory_temporal_positional_encoding", "min", "mode", "multimask_max_pt_num", "multimask_min_pt_num", "multimask_output", "multimask_output_for_tracking", "multimask_output_in_Model", "must", "new_ones", "new_zeros", "nn", "no_grad", "no_memory_embedding", "no_memory_positional_encoding", "no_object_pointer", "non_blocking", "non_cond_frame_outputs", "normalized_temporal_diffs", "not", "num_channels", "num_feature_levels", "num_frames", "num_maskmem", "num_object_pointer_tokens", "num_objects", "num_pts", "num_splits", "num_total_frames", "obj_id", "obj_idx", "obj_idx_to_id", "obj_output_dict", "obj_with_new_inputs", "object_batch_inds", "object_pointer", "object_pointer_proj", "object_pointers", "object_pointers_list", "object_pointers_pos_embed", "object_score_logits", "objects", "occlusion_spatial_embedding_parameter", "of", "on", "one", "ones", "ones_like", "or", "out", "out_bias", "out_data", "out_scale", "output_data", "output_dict_per_obj", "output_value", "outputs", "pe", "per", "permute", "pix_feat", "pixel_values", "please", "point_coords", "point_inputs", "point_inputs_per_obj", "point_labels", "pointer_tpos_dim", "points", "positional_embedding", "post_init", "pred_masks", "pred_masks_high_res", "pred_masks_per_obj", "prev_Model_mask_logits", "prev_output_data", "previous_frame_idx", "processing_order", "projected_sine_pe", "prompt_encoder", "prompt_encoder_config", "prompt_output", "propagate", "propagate_in_video_iterator", "provide", "provided", "r", "raise", "range", "ref_frame_idx", "relative_temporal_offset", "remove", "repeat", "repeat_interleave", "reshape", "return", "reverse", "run", "run_mem_encoder", "s", "self", "shape", "shared_embedding", "shared_image_embedding", "should", "sigmoid", "sigmoid_bias_for_mem_enc", "sigmoid_scale_for_mem_enc", "simultaneously", "sine_pe", "size", "sparse_embeddings", "sparse_prompt_embeddings", "spatial_memory_pos_embed", "specify", "squeeze", "stack", "start_frame_idx", "starting", "store_output", "str", "streamed", "streaming", "stride", "super", "t_diff_offset", "target_device", "target_dtype", "target_embedding", "temporal_diff_and_pointers", "temporal_difference", "temporal_differences", "temporal_idx", "temporal_position_sign_multiplier", "temporal_positional_encoding_projection_layer", "temporal_positions_and_previous_outputs", "tensor", "the", "through", "to", "torch", "tqdm", "track_in_reverse_time", "tracking", "training", "tuple", "unsqueeze", "values", "video", "view", "vision_attentions", "vision_config", "vision_encoder", "vision_feats", "vision_hidden_states", "vision_outputs", "vision_pos_embeds", "when", "where", "width", "with", "x", "x_embed", "y_embed", "yield", "zeros", "zip"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerGatedAttention": ["Linear", "ModelGatedAttention", "Module", "Softmax", "__init__", "attn_layer", "attn_softmax", "attn_weight", "class", "def", "dim", "forward", "in_size", "inputs", "int", "nn", "out_size", "return", "self", "super"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerBatchNorm": ["BatchNorm1d", "ModelBatchNorm", "ModelConfig", "Module", "Tensor", "__init__", "batchnorm", "class", "config", "d_model", "def", "eps", "forward", "inputs", "nn", "norm_eps", "output", "return", "self", "super", "torch", "transpose"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerPositionalEncoding": ["Available", "False", "ModelConfig", "ModelPositionalEncoding", "Model_input", "Module", "Parameter", "Tensor", "True", "ValueError", "__init__", "_init_pe", "a", "and", "arange", "are", "class", "config", "cos", "d_model", "def", "div_term", "elif", "else", "encoder", "exp", "f", "forward", "hidden_state", "if", "is", "log", "math", "mean", "nn", "not", "num_Modeles", "position", "position_enc", "positional", "positional_encoding_type", "raise", "randn", "random", "requires_grad", "return", "self", "sin", "sincos", "staticmethod", "std", "super", "torch", "types", "unsqueeze", "use_positional_encoding", "valid", "zeros"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerNormLayer": ["LayerNorm", "ModelBatchNorm", "ModelConfig", "ModelNormLayer", "Module", "Tensor", "__init__", "batch", "class", "config", "d_model", "def", "else", "eps", "forward", "if", "in", "inputs", "inputs_reshaped", "lower", "nn", "norm", "norm_eps", "norm_mlp", "reshape", "return", "self", "shape", "super", "torch"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerMLP": ["Dropout", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "class", "config", "def", "dropout", "dropout1", "dropout2", "expansion_factor", "fc1", "fc2", "forward", "functional", "gelu", "in_features", "inputs", "nn", "num_hidden", "out_features", "return", "self", "super", "torch"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerChannelFeatureMixerBlock": ["ModelChannelFeatureMixerBlock", "ModelConfig", "ModelGatedAttention", "ModelMLP", "ModelNormLayer", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "gated_attn", "gating_block", "if", "in_features", "in_size", "inputs", "mlp", "nn", "norm", "num_input_channels", "out", "out_features", "out_size", "permute", "residual", "return", "self", "super", "torch"], "patchtsmixer/modeling_patchtsmixer.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "patchtsmixer/modeling_patchtsmixer.py:PatchMixerBlock": ["False", "ModelAttention", "ModelConfig", "ModelGatedAttention", "ModelMLP", "ModelMixerBlock", "ModelNormLayer", "Module", "_", "__init__", "batch_size", "class", "config", "d_model", "def", "dropout", "embed_dim", "forward", "gated_attn", "gating_block", "hidden_state", "hidden_state_reshaped", "if", "in_features", "in_size", "mlp", "n_vars", "nn", "norm", "norm_attn", "num_Modeles", "num_heads", "out", "out_features", "out_size", "output_attentions", "reshape", "residual", "return", "self", "self_attn", "self_attn_heads", "self_attn_layer", "shape", "super", "transpose", "x_attn"], "patchtsmixer/modeling_patchtsmixer.py:FeatureMixerBlock": ["ModelConfig", "ModelGatedAttention", "ModelMLP", "ModelMixerBlock", "ModelNormLayer", "Module", "Tensor", "__init__", "class", "config", "d_model", "def", "forward", "gated_attn", "gating_block", "hidden", "if", "in_Models", "in_size", "mlp", "nn", "norm", "out", "out_Models", "out_size", "residual", "return", "self", "super", "torch"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerLayer": ["FeatureMixerBlock", "ModelChannelFeatureMixerBlock", "ModelConfig", "ModelLayer", "ModelMixerBlock", "Model_mixer", "Module", "Tensor", "__init__", "channel_feature_mixer", "class", "config", "def", "feature_mixer", "forward", "hidden", "if", "mix_channel", "mode", "nn", "return", "self", "super", "torch"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerBlock": ["False", "ModelBlock", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "_", "__init__", "all_hidden_states", "append", "bool", "class", "config", "def", "else", "embedding", "for", "forward", "hidden_state", "if", "in", "mixers", "mod", "nn", "num_layers", "output_hidden_states", "range", "return", "self", "super"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForPredictionHead": ["Dropout", "Flatten", "Linear", "ModelConfig", "ModelForPredictionHead", "Module", "None", "__init__", "base_forecast_block", "class", "config", "d_model", "def", "distribution_output", "dropout_layer", "else", "flatten", "for", "forecast", "forward", "get_parameter_projection", "head_dropout", "hidden_features", "if", "in", "is", "isinstance", "nn", "not", "num_Modeles", "prediction_channel_indices", "prediction_length", "return", "self", "sort", "start_dim", "super", "transpose", "tuple", "z"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerLinearHead": ["Dropout", "Flatten", "Linear", "ModelConfig", "ModelLinearHead", "Module", "None", "__init__", "and", "avg_pool", "class", "config", "d_model", "def", "dim", "distribution_output", "dropout", "elif", "else", "flatten", "forward", "get_parameter_projection", "head_aggregation", "head_dropout", "hidden_features", "if", "is", "max", "max_pool", "mean", "mul_factor", "nn", "not", "num_Modeles", "num_input_channels", "num_targets", "output_range", "projection", "return", "self", "sigmoid", "start_dim", "super", "torch", "transpose", "use_last", "values"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerPreTrainedModel": ["BatchNorm1d", "False", "LayerNorm", "Linear", "ModelBatchNorm", "ModelConfig", "ModelPositionalEncoding", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "batchnorm", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "init", "init_std", "is", "isinstance", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "past_values", "position_enc", "positional_encoding_type", "random", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerPretrainHead": ["Dropout", "Linear", "ModelConfig", "ModelPretrainHead", "Model_length", "Module", "__init__", "base_pt_block", "class", "config", "d_model", "def", "dropout_layer", "forecast", "forward", "head_dropout", "hidden_features", "nn", "return", "self", "super"], "patchtsmixer/modeling_patchtsmixer.py:random_masking": ["False", "Mask", "Model_masking", "None", "Optional", "Tensor", "ValueError", "and", "argsort", "batch_size", "be", "between", "bool", "channel_consistent_masking", "def", "device", "dim", "else", "f", "float", "gather", "has", "ids_restore", "ids_shuffle", "if", "index", "inputs", "inputs_mask", "int", "is", "len_keep", "list", "mask", "mask_ratio", "mask_value", "masked_fill", "noise", "not", "num_channels", "num_features", "ones", "or", "raise", "rand", "ratio", "repeat", "return", "sequence_length", "shape", "to", "torch", "unmasked_channel_indices", "unsqueeze"], "patchtsmixer/modeling_patchtsmixer.py:forecast_masking": ["Model_mask_ratios", "Model_masking", "None", "Optional", "Tensor", "Union", "ValueError", "_", "and", "append", "batch1", "batch2", "batch_size", "be", "bool", "def", "device", "elif", "f", "for", "greater", "if", "in", "inputs", "inputs_mask", "int", "is", "isinstance", "key", "lambda", "less", "list", "mask", "mask_value", "masked_fill", "not", "num_Model_mask_patches", "num_channels", "num_features", "or", "patch_len", "patch_length", "patches", "perm", "raise", "randperm", "ratio", "repeat", "return", "sequence_length", "shape", "should", "sorted", "sum", "t_list", "temp_len", "than", "torch", "total", "total_length", "total_ratio", "unmasked_channel_indices", "unsqueeze", "x", "zeros", "zip"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerPatchify": ["Input", "Model", "ModelConfig", "ModelModelify", "Model_length", "Model_stride", "Module", "Sequence", "Tensor", "ValueError", "__init__", "be", "class", "config", "configuration", "context_length", "contiguous", "def", "dimension", "doesn", "f", "forward", "greater", "has", "if", "length", "match", "max", "model", "new_sequence_length", "nn", "num_Modeles", "output", "past_values", "raise", "return", "self", "sequence", "sequence_length", "sequence_start", "shape", "size", "step", "super", "t", "than", "the", "to", "torch", "transpose", "unfold"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerMasking": ["Invalid", "ModelConfig", "ModelMasking", "Model_input", "Module", "None", "Tensor", "ValueError", "__init__", "bool", "channel_consistent_masking", "class", "config", "def", "elif", "else", "f", "forecast", "forecast_masking", "forward", "if", "inputs", "is", "mask", "mask_ratio", "mask_type", "mask_value", "masked_input", "nn", "not", "num_forecast_mask_Modeles", "raise", "random", "random_mask_ratio", "random_masking", "return", "self", "sorted", "super", "torch", "type", "unmasked_channel_indices"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerStdScaler": ["ModelConfig", "ModelStdScaler", "Module", "Tensor", "True", "__init__", "clamp_min", "class", "config", "data", "def", "denominator", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "minimum_scale", "nn", "observed_indicator", "return", "scale", "scaling_dim", "self", "sqrt", "sum", "super", "torch", "tuple", "variance"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerMeanScaler": ["ModelConfig", "ModelMeanScaler", "Module", "None", "Tensor", "True", "__init__", "abs", "batch_observations", "batch_sum", "clamp", "class", "config", "data", "def", "default_scale", "dim", "else", "forward", "hasattr", "if", "is", "keepdim", "min", "minimum_scale", "nn", "not", "num_observed", "observed_indicator", "ones_like", "return", "scale", "scaled_data", "scaling_dim", "self", "squeeze", "sum", "super", "torch", "ts_sum", "tuple", "where", "zeros_like"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerNOPScaler": ["False", "ModelConfig", "ModelNOPScaler", "Module", "None", "Optional", "Tensor", "True", "__init__", "class", "config", "data", "def", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "mean", "nn", "observed_indicator", "ones_like", "requires_grad", "return", "scale", "scaling_dim", "self", "super", "torch", "tuple", "zeros_like"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerEncoder": ["False", "Linear", "ModelBlock", "ModelConfig", "ModelEncoder", "ModelEncoderOutput", "ModelPositionalEncoding", "ModelPreTrainedModel", "Model_length", "Modeler", "Modeles", "None", "Optional", "Tensor", "Union", "__init__", "auto_docstring", "bool", "class", "config", "d_model", "def", "else", "for", "forward", "hidden_states", "if", "in", "is", "last_hidden_state", "mlp_mixer_encoder", "nn", "not", "output_hidden_states", "past_values", "positional_encoder", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_positional_encoding", "use_return_dict", "v"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "Model_input", "None", "Optional", "class", "hidden_states", "last_hidden_state", "loc", "mask", "r", "scale", "torch", "tuple"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerModel": ["False", "ModelConfig", "ModelEncoder", "ModelEncoderOutput", "ModelMasking", "ModelMeanScaler", "ModelModel", "ModelModelOutput", "ModelModelify", "ModelNOPScaler", "ModelPreTrainedModel", "ModelStdScaler", "Model_input", "Modeled_x", "Modeling", "None", "Optional", "Tensor", "True", "__init__", "auto_docstring", "bool", "class", "config", "def", "elif", "else", "enc_input", "encoder", "encoder_output", "for", "forward", "hidden_states", "if", "in", "is", "isinstance", "last_hidden_state", "loc", "mask", "mask_input", "masking", "mean", "not", "observed_mask", "ones_like", "or", "output_hidden_states", "past_values", "post_init", "r", "return", "return_dict", "scale", "scaled_past_values", "scaler", "scaling", "self", "std", "super", "torch", "tuple", "use_return_dict", "v"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "loss", "prediction_outputs", "r", "torch", "tuple"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForPretraining": ["False", "MSELoss", "ModelConfig", "ModelForPreTrainingOutput", "ModelForPretraining", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModelPretrainHead", "Model_input", "None", "Optional", "Tensor", "True", "__init__", "and", "auto_docstring", "bool", "class", "config", "def", "dim", "else", "for", "forward", "head", "hidden_states", "if", "in", "is", "isinstance", "last_hidden_state", "loss", "loss_val", "mask", "mask_input", "masked_loss", "mean", "model", "model_output", "nn", "none", "not", "observed_mask", "output_hidden_states", "past_values", "post_init", "prediction_outputs", "r", "reduction", "return", "return_dict", "return_loss", "self", "sum", "super", "torch", "tuple", "use_return_dict", "v", "x_hat"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForPredictionOutput": ["FloatTensor", "ModelForPredictionOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "loc", "loss", "prediction_outputs", "r", "scale", "torch", "tuple"], "patchtsmixer/modeling_patchtsmixer.py:SamplePatchTSMixerPredictionOutput": ["FloatTensor", "ModelModelPredictionOutput", "ModelOutput", "None", "Optional", "class", "r", "sequences", "torch"], "patchtsmixer/modeling_patchtsmixer.py:SamplePatchTSMixerRegressionOutput": ["FloatTensor", "ModelModelRegressionOutput", "ModelOutput", "None", "Optional", "class", "r", "sequences", "torch"], "patchtsmixer/modeling_patchtsmixer.py:nll": ["Distribution", "Model", "Tensor", "def", "distributions", "input", "log_prob", "return", "target", "torch"], "patchtsmixer/modeling_patchtsmixer.py:weighted_average": ["Model_average", "Model_tensor", "None", "Optional", "Tensor", "clamp", "def", "dim", "else", "if", "input_tensor", "is", "mean", "min", "not", "return", "sum", "sum_weights", "torch", "weights", "where", "zeros_like"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForPrediction": ["Allowed", "False", "Invalid", "MSELoss", "ModelConfig", "ModelForPrediction", "ModelForPredictionHead", "ModelForPredictionOutput", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleModelPredictionOutput", "StudentTOutput", "Tensor", "True", "Unknown", "ValueError", "_", "__init__", "and", "auto_docstring", "bool", "class", "config", "def", "dim", "distribution", "distribution_output", "distribution_output_map", "elif", "else", "f", "for", "forward", "function", "future_values", "generate", "get", "head", "hidden_states", "if", "in", "is", "isinstance", "last_hidden_state", "loc", "loss", "loss_val", "mean", "model", "model_output", "mse", "negative_binomial", "nll", "nn", "no_grad", "normal", "not", "num_parallel_samples", "observed_mask", "output", "output_class", "output_hidden_states", "outputs", "past_values", "post_init", "prediction_channel_indices", "prediction_length", "prediction_outputs", "r", "raise", "range", "reduction", "return", "return_dict", "return_loss", "sample", "samples", "scale", "self", "sequences", "stack", "student_t", "super", "torch", "tuple", "use_return_dict", "v", "values", "weighted_average", "y_hat"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForTimeSeriesClassificationOutput": ["FloatTensor", "ModelForTimeSeriesClassificationOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "loss", "prediction_outputs", "r", "torch", "tuple"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForTimeSeriesClassification": ["CrossEntropyLoss", "False", "InjectScalerStatistics4D", "ModelConfig", "ModelForTimeSeriesClassification", "ModelForTimeSeriesClassificationOutput", "ModelLinearHead", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "and", "auto_docstring", "bool", "class", "config", "d_model", "def", "else", "for", "forward", "head", "hidden_states", "if", "in", "inject_scale", "is", "isinstance", "last_hidden_state", "loc", "loss", "loss_val", "mean", "model", "model_output", "nn", "not", "num_Modeles", "output_hidden_states", "past_values", "post_init", "prediction_outputs", "r", "return", "return_dict", "return_loss", "scale", "scaling", "self", "std", "super", "target_values", "torch", "tuple", "use_return_dict", "v", "y_hat"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForRegressionOutput": ["FloatTensor", "ModelForRegressionOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "loss", "r", "regression_outputs", "torch", "tuple"], "patchtsmixer/modeling_patchtsmixer.py:InjectScalerStatistics4D": ["Linear", "ModelScalerStatistics4D", "Module", "Tensor", "__init__", "cat", "class", "concat_stats", "d_model", "def", "dim", "expansion", "forward", "inputs", "int", "inverse_trans_compression", "inverse_trans_expansion", "loc", "map_scale_compression", "map_scale_expansion", "mean", "nn", "num_patches", "repeat", "return", "scale", "self", "stdev", "super", "torch", "transpose", "unsqueeze"], "patchtsmixer/modeling_patchtsmixer.py:PatchTSMixerForRegression": ["Allowed", "Exception", "False", "InjectScalerStatistics4D", "Invalid", "MSELoss", "ModelConfig", "ModelForRegression", "ModelForRegressionOutput", "ModelLinearHead", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleModelRegressionOutput", "StudentTOutput", "Tensor", "True", "Unknown", "ValueError", "_", "__init__", "and", "any", "auto_docstring", "be", "bool", "cannot", "class", "config", "d_model", "def", "dim", "distribution", "distribution_output", "distribution_output_map", "elif", "else", "f", "for", "forward", "function", "generate", "get", "head", "hidden_states", "if", "in", "inject_scale", "is", "isinstance", "item", "last_hidden_state", "loc", "loss", "loss_val", "mean", "model", "model_output", "mse", "negative", "negative_binomial", "nll", "nn", "no_grad", "normal", "not", "num_Modeles", "num_parallel_samples", "num_targets", "output", "output_class", "output_hidden_states", "outputs", "past_values", "post_init", "r", "raise", "range", "reduction", "regression_outputs", "return", "return_dict", "return_loss", "sample", "samples", "scale", "scaling", "self", "sequences", "stack", "std", "student_t", "super", "target_values", "torch", "tuple", "use_return_dict", "v", "values", "view", "weighted_average", "y_hat"], "doge/modeling_doge.py:DogeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "doge/modeling_doge.py:DogeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "doge/modeling_doge.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "doge/modeling_doge.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "doge/modeling_doge.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "doge/modeling_doge.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "doge/modeling_doge.py:flex_attention_forward": ["BlockMask", "Model_attention_forward", "Module", "None", "Optional", "Tensor", "True", "Union", "attention_mask", "attention_weights", "attn_output", "batch_idx", "block_mask", "causal_mask", "compile_friendly_Model_attention", "contiguous", "def", "dtype", "else", "enable_gqa", "float", "head_idx", "head_mask", "if", "is", "isinstance", "key", "kv_idx", "kwargs", "module", "nn", "not", "q_idx", "query", "return", "return_lse", "scale", "scaling", "score", "score_mod", "shape", "softcap", "tanh", "to", "torch", "transpose", "tuple", "value"], "doge/modeling_doge.py:DogeAttention": ["A", "ALL_ATTENTION_FUNCTIONS", "BlockMask", "Cache", "Callable", "F", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Parameter", "Tensor", "True", "__init__", "_attn_implementation", "active_mask", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_mask", "attn_output", "attn_weights", "bias", "bool", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "device", "dim", "dropout", "dt_proj", "dt_states", "dtype", "eager", "eager_attention_forward", "else", "eps", "exp", "expand", "finfo", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "indices", "input_shape", "int", "is", "isinstance", "k_norm", "k_proj", "keep_window_size", "key_states", "kwargs", "largest", "layer_idx", "masked_fill", "min", "min_dtype", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "prepare_dynamic_mask", "q_norm", "q_proj", "query_states", "repeat_kv", "reshape", "return", "rms_norm_eps", "scaling", "scatter", "self", "shape", "sin", "softplus", "sorted", "super", "tensor", "topk", "topk_indices", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view", "where", "zeros", "zeros_like"], "doge/modeling_doge.py:DogeMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "doge/modeling_doge.py:DogeCDMoE": ["ACT2FN", "Embedding", "F", "False", "Linear", "ModelCDMoE", "ModelConfig", "Module", "Tensor", "True", "_", "__init__", "act_fn", "all_indices", "all_scores", "bias", "bsz", "class", "config", "def", "dim", "down_embed", "down_proj", "experts_states", "experts_weights", "floor", "forward", "gate_proj", "gather", "hidden_act", "hidden_size", "hidden_states", "if", "indices", "indices_x", "indices_y", "intermediate_size", "keepdim", "kwargs", "math", "matmul", "mlp_bias", "nn", "norm_topk_prob", "num_experts", "num_experts_per_tok", "num_keys", "position_indices", "return", "router_gate", "router_logits", "routing_weights", "scores", "scores_x", "scores_y", "self", "seq_len", "shape", "softmax", "sqrt", "sum", "super", "top_k", "topk", "torch", "unsqueeze", "up_embed", "up_proj", "view"], "doge/modeling_doge.py:DogeDecoderLayer": ["Cache", "F", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelCDMoE", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "else", "eps", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "input_residual", "int", "is_moe", "kwargs", "layer_idx", "mlp", "new_name", "nn", "not", "ones", "p", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_attention_residual", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "doge/modeling_doge.py:DogePreTrainedModel": ["A", "False", "ModelAttention", "ModelCDMoE", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "elif", "fill_", "hasattr", "hidden_states", "if", "index", "input_residual", "isinstance", "model", "module", "past_key_values", "post_attention_residual", "router_logits", "self", "super", "supports_gradient_checkpointing", "zero_"], "doge/modeling_doge.py:DogeModel": ["Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "doge/modeling_doge.py:load_balancing_loss_func": ["F", "Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "all_expert_indices", "all_indices", "all_routing_weights", "all_scores", "append", "attention_mask", "batch_size", "bool", "cat", "compute_device", "compute_dtype", "def", "device", "dim", "dtype", "else", "expand", "expert_attention_mask", "expert_indices", "for", "gate_logits", "gather", "if", "in", "indices_x", "indices_y", "int", "is", "isinstance", "layer_gate_logits", "len", "mean", "not", "num_experts", "num_hidden_layers", "num_keys", "ones_like", "or", "overall_loss", "pad", "position_indices", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "scatter_add_", "scores_x", "scores_y", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze", "view", "zeros"], "doge/modeling_doge.py:DogeForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "floor", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "math", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "sqrt", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "doge/modeling_doge.py:DogeForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "dac/modeling_dac.py:DacOutput": ["FloatTensor", "LongTensor", "ModelOutput", "None", "Optional", "audio_codes", "audio_values", "class", "loss", "projected_latents", "quantized_representation", "r", "torch"], "dac/modeling_dac.py:DacEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "audio_codes", "class", "loss", "projected_latents", "quantized_representation", "r", "torch"], "dac/modeling_dac.py:DacDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "audio_values", "class", "r", "torch"], "dac/modeling_dac.py:Snake1d": ["Model", "Module", "Parameter", "__init__", "alpha", "class", "def", "forward", "hidden_dim", "hidden_states", "nn", "ones", "pow", "reciprocal", "reshape", "return", "self", "shape", "sin", "super", "torch"], "dac/modeling_dac.py:DacVectorQuantize": ["Conv1d", "Embedding", "F", "ModelConfig", "ModelVectorQuantize", "Module", "True", "__init__", "audio_codes", "batch_size", "class", "codebook", "codebook_dim", "codebook_loss", "codebook_size", "commitment_loss", "config", "decode_latents", "def", "detach", "dist", "encodings", "forward", "hidden_dim", "hidden_size", "hidden_state", "hidden_states", "in_proj", "indices", "keepdim", "kernel_size", "l2_norm", "max", "mean", "mse_loss", "nn", "normalize", "out_proj", "permute", "pow", "projected_latents", "quantized_representation", "reduction", "reshape", "return", "self", "sequence_length", "shape", "size", "sum", "super", "t", "transpose", "weight"], "dac/modeling_dac.py:DacResidualUnit": ["Conv1d", "ModelResidualUnit", "Module", "Snake1d", "__init__", "class", "conv1", "conv2", "def", "dilation", "dimension", "forward", "hidden_state", "if", "int", "kernel_size", "nn", "output_tensor", "pad", "padding", "return", "self", "shape", "snake1", "snake2", "super"], "dac/modeling_dac.py:DacEncoderBlock": ["Conv1d", "ModelConfig", "ModelEncoderBlock", "ModelResidualUnit", "Module", "Snake1d", "__init__", "ceil", "class", "config", "conv1", "def", "dilation", "dimension", "encoder_hidden_size", "forward", "hidden_state", "int", "kernel_size", "math", "nn", "padding", "res_unit1", "res_unit2", "res_unit3", "return", "self", "snake1", "stride", "stride_index", "super"], "dac/modeling_dac.py:DacDecoderBlock": ["ConvTranspose1d", "ModelConfig", "ModelDecoderBlock", "ModelResidualUnit", "Module", "Snake1d", "__init__", "ceil", "class", "config", "conv_t1", "decoder_hidden_size", "def", "dilation", "forward", "hidden_state", "input_dim", "int", "kernel_size", "math", "nn", "output_dim", "padding", "res_unit1", "res_unit2", "res_unit3", "return", "self", "snake1", "stride", "stride_index", "super"], "dac/modeling_dac.py:DacResidualVectorQuantize": ["False", "ModelConfig", "ModelResidualVectorQuantize", "ModelVectorQuantize", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "__init__", "and", "append", "audio_codes", "axis", "break", "cat", "class", "codebook", "codebook_dim", "codebook_dims_tensor", "codebook_loss", "codebook_loss_i", "codes", "codes_i", "commitment_loss", "commitment_loss_i", "config", "cumsum", "decode_latents", "def", "device", "dim", "dims", "dropout", "else", "enumerate", "fill_value", "for", "forward", "from_codes", "from_latents", "full", "hidden_dim_j", "hidden_dim_k", "hidden_state", "i", "if", "in", "indices_i", "int", "is", "keepdims", "latents", "mask", "max", "n_codebooks", "n_dropout", "n_quantizers", "nn", "not", "np", "ones", "out_proj", "projected_latents", "projected_latents_i", "q", "quantized_latents", "quantized_latents_i", "quantized_representation", "quantized_representation_i", "quantizer", "quantizer_dropout", "quantizers", "randint", "range", "residual", "return", "self", "shape", "stack", "super", "tensor", "to", "torch", "training", "transpose", "where"], "dac/modeling_dac.py:DacDecoder": ["Conv1d", "ModelConfig", "ModelDecoder", "ModelDecoderBlock", "Module", "ModuleList", "Snake1d", "Tanh", "__init__", "block", "channels", "class", "config", "conv1", "conv2", "decoder_hidden_size", "def", "enumerate", "for", "forward", "hidden_size", "hidden_state", "in", "input_channel", "kernel_size", "layer", "nn", "output_dim", "padding", "return", "self", "snake1", "stride", "stride_index", "strides", "super", "tanh", "upsampling_ratios"], "dac/modeling_dac.py:DacEncoder": ["Conv1d", "ModelConfig", "ModelEncoder", "ModelEncoderBlock", "Module", "ModuleList", "Snake1d", "__init__", "block", "class", "config", "conv1", "conv2", "d_model", "def", "downsampling_ratios", "encoder_hidden_size", "enumerate", "for", "forward", "hidden_size", "hidden_state", "in", "kernel_size", "module", "nn", "padding", "return", "self", "snake1", "stride", "stride_index", "strides", "super"], "dac/modeling_dac.py:DacPreTrainedModel": ["Conv1d", "ConvTranspose1d", "Embedding", "Model", "ModelConfig", "ModelPreTrainedModel", "PreTrainedAudioTokenizerBase", "Snake1d", "_init_weights", "alpha", "apply_weight_norm", "base_model_prefix", "bias", "block", "class", "config", "constant_", "conv1", "conv2", "conv_t1", "data", "decoder", "def", "elif", "encoder", "fill_", "for", "hasattr", "if", "in", "in_proj", "init", "input_values", "isinstance", "layer", "main_input_name", "mean", "module", "nn", "normal_", "out_proj", "parametrizations", "quantizer", "quantizers", "remove_weight_norm", "res_unit1", "res_unit2", "res_unit3", "reset_parameters", "self", "std", "trunc_normal_", "utils", "weight", "weight_norm"], "dac/modeling_dac.py:DacModel": ["Either", "False", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelEncoder", "ModelEncoderOutput", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelResidualVectorQuantize", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "a", "and", "audio_codes", "audio_values", "auto_docstring", "be", "bits_per_codebook", "bool", "class", "codebook_loss", "codebook_loss_weight", "codebook_size", "commitment_loss", "commitment_loss_weight", "config", "decode", "decoder", "def", "else", "encode", "encoder", "forward", "from_codes", "if", "input_values", "int", "is", "length", "log2", "loss", "math", "must", "n_quantizers", "not", "of", "or", "post_init", "power", "projected_latents", "provided", "quantized_representation", "quantizer", "r", "raise", "return", "return_dict", "self", "shape", "squeeze", "super", "torch"], "chinese_clip/modeling_chinese_clip.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "chinese_clip/modeling_chinese_clip.py:chinese_clip_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPOutput": ["Any", "BaseModelOutputWithPoolingAndCrossAttentions", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "chinese_clip/modeling_chinese_clip.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelTextSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "bool", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_states", "kwargs", "multiple", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "query", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_states", "view"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextAttention": ["False", "FloatTensor", "ModelTextAttention", "ModelTextSelfAttention", "ModelTextSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attn_output", "attn_weights", "be", "bool", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "k_proj", "key_states", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "query_states", "raise", "reshape", "return", "scale", "scaling", "self", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextIntermediate": ["ACT2FN", "Linear", "ModelTextIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelTextAttention", "ModelTextIntermediate", "ModelTextLayer", "ModelTextOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelConfig", "ModelVisionAttention", "ModelVisionLayer", "ModelVisionMLP", "Optional", "Tensor", "__init__", "attn_weights", "bool", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextPooler": ["Linear", "ModelTextPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPPreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionAttention", "ModelVisionEmbeddings", "ModelVisionMLP", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "embed_dim", "embedding", "factor", "fc1", "fc2", "fc_std", "fill_", "for", "hidden_size", "if", "in", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "padding_idx", "patch_embedding", "position_embedding", "position_embeddings", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_type_embeddings", "v_proj", "vision_embed_dim", "visual_projection", "weight", "word_embeddings", "zero_"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelTextEncoder", "ModelTextLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelVisionEncoder", "ModelVisionLayer", "Module", "ModuleList", "None", "Optional", "Union", "_", "__init__", "all_attentions", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "tuple", "use_return_dict"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionTransformer": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionTransformer", "Module", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layrnorm", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPTextModel": ["BaseModelOutputWithPooling", "Cache", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextModel", "ModelTextPooler", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_no_split_modules", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "can_return_tuple", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPVisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "_no_split_modules", "auto_docstring", "bool", "class", "config", "def", "else", "embeddings", "forward", "get_input_embeddings", "if", "interpolate_pos_encoding", "is", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "vision_model"], "chinese_clip/modeling_chinese_clip.py:ChineseCLIPModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelVisionConfig", "ModelVisionTransformer", "Model_loss", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "Union", "__init__", "_attn_implementation", "add_pooling_layer", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "can_return_tuple", "class", "config", "def", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "output_attentions", "output_hidden_states", "p", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "return", "return_dict", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "token_type_ids", "torch", "tuple", "type", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "convbert/modeling_convbert.py:ConvBertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "__init__", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embedding_size", "embeddings", "eps", "expand", "forward", "hasattr", "hidden_dropout_prob", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "convbert/modeling_convbert.py:ConvBertPreTrainedModel": ["Embedding", "GroupedLinearLayer", "LayerNorm", "Linear", "Model", "Model1d", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "SeparableModel1D", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "convbert/modeling_convbert.py:SeparableConv1D": ["Conv1d", "False", "ModelConv1D", "Module", "Parameter", "Tensor", "__init__", "bias", "class", "config", "data", "def", "depthwise", "forward", "groups", "hidden_states", "initializer_range", "input_filters", "kernel_size", "kwargs", "mean", "nn", "normal_", "output_filters", "padding", "pointwise", "return", "self", "std", "super", "torch", "weight", "x", "zeros"], "convbert/modeling_convbert.py:ConvBertSelfAttention": ["Dropout", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Model_attn_layer", "Model_kernel_layer", "Model_kernel_size", "Model_out", "Model_out_layer", "Module", "None", "Optional", "SeparableModel1D", "Tensor", "The", "Unfold", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "be", "bool", "by", "cat", "class", "config", "context_layer", "contiguous", "def", "dilation", "dim", "divisible", "dropout", "else", "embedding_size", "encoder_hidden_states", "f", "forward", "functional", "hasattr", "head_mask", "head_ratio", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "kernel_size", "key", "key_Model_attn_layer", "key_layer", "math", "matmul", "mixed_key_Model_attn_layer", "mixed_key_layer", "mixed_query_layer", "mixed_value_layer", "multiple", "multiply", "new_context_layer_shape", "new_num_attention_heads", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "padding", "permute", "query", "query_layer", "raise", "reshape", "return", "self", "seq_length", "shape", "should", "size", "softmax", "sqrt", "stride", "super", "the", "torch", "transpose", "tuple", "unfold", "unsqueeze", "value", "value_layer", "view"], "convbert/modeling_convbert.py:ConvBertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "convbert/modeling_convbert.py:ConvBertAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "convbert/modeling_convbert.py:GroupedLinearLayer": ["ModelLinearLayer", "Module", "Parameter", "Tensor", "__init__", "batch_size", "bias", "class", "def", "empty", "forward", "group_in_dim", "group_out_dim", "hidden_states", "input_size", "list", "matmul", "nn", "num_groups", "output_size", "permute", "reshape", "return", "self", "size", "super", "torch", "weight", "x"], "convbert/modeling_convbert.py:ConvBertIntermediate": ["ACT2FN", "GroupedLinearLayer", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "input_size", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "num_groups", "output_size", "return", "self", "str", "super", "torch"], "convbert/modeling_convbert.py:ConvBertOutput": ["Dropout", "GroupedLinearLayer", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "else", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_size", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "num_groups", "output_size", "return", "self", "super", "torch"], "convbert/modeling_convbert.py:ConvBertLayer": ["AttributeError", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "TypeError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "bool", "by", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_output", "layers", "model", "not", "output", "output_attentions", "outputs", "passed", "raise", "return", "self", "self_attention_outputs", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "convbert/modeling_convbert.py:ConvBertEncoder": ["BaseModelOutputWithCrossAttentions", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "convbert/modeling_convbert.py:ConvBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "convbert/modeling_convbert.py:ConvBertSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "ModelConfig", "ModelSequenceSummary", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "convbert/modeling_convbert.py:ConvBertModel": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "Linear", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_size", "embeddings", "embeddings_project", "encoder", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "layer", "long", "nn", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "seq_length", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "convbert/modeling_convbert.py:ConvBertGeneratorPredictions": ["FloatTensor", "LayerNorm", "Linear", "ModelGeneratorPredictions", "Module", "__init__", "activation", "class", "config", "def", "dense", "embedding_size", "eps", "forward", "gelu", "generator_hidden_states", "get_activation", "hidden_size", "hidden_states", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "convbert/modeling_convbert.py:ConvBertForMaskedLM": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelGeneratorPredictions", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedding_size", "forward", "generator", "generator_hidden_states", "generator_lm_head", "generator_predictions", "generator_sequence_output", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_head", "logits", "loss", "loss_fct", "nn", "not", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight", "word_embeddings"], "convbert/modeling_convbert.py:ConvBertClassificationHead": ["ACT2FN", "Dropout", "Linear", "ModelClassificationHead", "Module", "None", "Tensor", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "torch", "x"], "convbert/modeling_convbert.py:ConvBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "convbert/modeling_convbert.py:ConvBertForMultipleChoice": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "MultipleChoiceModelOutput", "None", "Optional", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_output", "sequence_summary", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "convbert/modeling_convbert.py:ConvBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "convbert/modeling_convbert.py:ConvBertForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "xlnet/modeling_xlnet.py:XLNetRelativeAttention": ["Dropout", "False", "FloatTensor", "LayerNorm", "Model", "Module", "None", "NotImplementedError", "Parameter", "The", "True", "ValueError", "__init__", "a", "ac", "and", "arange", "attention", "attn_mask", "attn_mask_g", "attn_mask_h", "attn_out", "attn_prob", "attn_prob_g", "attn_prob_h", "attn_score", "attn_vec", "attn_vec_g", "attn_vec_h", "bd", "bnij", "cat", "class", "config", "d_head", "d_model", "def", "device", "dim", "dropout", "dtype", "ef", "einsum", "else", "eps", "f", "float16", "forward", "functional", "g", "h", "head_mask", "heads", "hidden", "hnd", "ibh", "ibnd", "ibns", "if", "ijbn", "ijbs", "index_select", "is", "jbnd", "k", "k_head_h", "k_head_r", "klen", "layer_norm", "layer_norm_eps", "lbnd", "long", "mbnd", "mems", "mlb", "multiple", "n_head", "nn", "not", "number", "o", "of", "output", "output_attentions", "output_g", "output_h", "outputs", "post_attention", "prune_heads", "q", "q_head", "q_head_g", "q_head_h", "r", "r_r_bias", "r_s_bias", "r_w_bias", "raise", "rel_attn_core", "rel_shift", "rel_shift_bnij", "reshape", "residual", "return", "scale", "seg_embed", "seg_mat", "self", "shape", "size", "snd", "softmax", "staticmethod", "super", "target_mapping", "the", "torch", "type", "v", "v_head_h", "x", "x_size"], "xlnet/modeling_xlnet.py:XLNetFeedForward": ["ACT2FN", "Dropout", "LayerNorm", "Linear", "Model", "Module", "__init__", "activation_function", "class", "config", "d_inner", "d_model", "def", "dropout", "else", "eps", "ff_activation", "forward", "if", "inp", "isinstance", "layer_1", "layer_2", "layer_norm", "layer_norm_eps", "nn", "output", "return", "self", "str", "super"], "xlnet/modeling_xlnet.py:XLNetLayer": ["Dropout", "False", "Model", "ModelFeedForward", "ModelRelativeAttention", "Module", "None", "__init__", "apply_chunking_to_forward", "attn_mask_g", "attn_mask_h", "chunk_size_feed_forward", "class", "config", "def", "dropout", "ff", "ff_chunk", "forward", "head_mask", "if", "is", "mems", "nn", "not", "output_attentions", "output_g", "output_h", "output_x", "outputs", "r", "rel_attn", "return", "seg_mat", "self", "seq_len_dim", "super", "target_mapping"], "xlnet/modeling_xlnet.py:XLNetPoolerStartLogits": ["FloatTensor", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "__init__", "class", "config", "def", "dense", "dtype", "else", "float16", "forward", "hidden_size", "hidden_states", "if", "is", "nn", "not", "p_mask", "return", "self", "squeeze", "super", "torch", "x"], "xlnet/modeling_xlnet.py:XLNetPoolerEndLogits": ["FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "One", "Optional", "Tanh", "__init__", "activation", "assert", "be", "cat", "class", "config", "def", "dense_0", "dense_1", "dim", "dtype", "else", "eps", "expand", "float16", "forward", "gather", "hidden_size", "hidden_states", "hsz", "if", "is", "layer_norm_eps", "nn", "not", "of", "or", "p_mask", "return", "self", "shape", "should", "slen", "squeeze", "start_positions", "start_states", "super", "torch", "x"], "xlnet/modeling_xlnet.py:XLNetPoolerAnswerClass": ["False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "One", "Optional", "Tanh", "__init__", "activation", "assert", "be", "bias", "cat", "class", "cls_index", "cls_token_state", "config", "def", "dense_0", "dense_1", "dim", "else", "expand", "forward", "gather", "hidden_size", "hidden_states", "hsz", "if", "is", "nn", "not", "of", "or", "return", "self", "shape", "should", "squeeze", "start_positions", "start_states", "super", "torch", "x"], "xlnet/modeling_xlnet.py:XLNetSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "xlnet/modeling_xlnet.py:XLNetPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "ModelRelativeAttention", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "for", "if", "in", "initializer_range", "is", "isinstance", "k", "mask_emb", "mean", "module", "nn", "normal_", "not", "o", "padding_idx", "param", "q", "r", "r_r_bias", "r_s_bias", "r_w_bias", "seg_embed", "self", "std", "transformer", "v", "weight", "zero_"], "xlnet/modeling_xlnet.py:XLNetModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "list", "mems", "r", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetLMHeadModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "mems", "r", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetForSequenceClassificationOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "mems", "r", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetForTokenClassificationOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "mems", "r", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetForMultipleChoiceOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "mems", "r", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetForQuestionAnsweringSimpleOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "end_logits", "hidden_states", "list", "loss", "mems", "r", "start_logits", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetForQuestionAnsweringOutput": ["FloatTensor", "LongTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cls_logits", "end_top_index", "end_top_log_probs", "hidden_states", "list", "loss", "mems", "r", "start_top_index", "start_top_log_probs", "torch", "tuple"], "xlnet/modeling_xlnet.py:XLNetModel": ["BERT", "Dropout", "Embedding", "FloatTensor", "FutureWarning", "Model", "ModelLayer", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "NotImplementedError", "Optional", "Parameter", "Please", "Tensor", "The", "Union", "Unknown", "Unsupported", "ValueError", "You", "_", "__init__", "_prune_heads", "a", "added", "and", "append", "arange", "argument", "assert", "at", "att_stream", "attention", "attention_mask", "attentions", "attn_mask", "attn_mask_g", "attn_mask_h", "attn_type", "auto_docstring", "be", "beg", "bi", "bi_data", "bool", "both", "bsz", "bwd_pos_emb", "bwd_pos_seq", "cache_mem", "can", "cannot", "cat", "cat_ids", "choose", "clamp", "clamp_len", "class", "compatibility", "config", "contiguous", "cos", "create_mask", "curr_out", "cutoff", "d", "d_model", "data_mask", "def", "deprecated", "detach", "device", "dim", "dropout", "dtype", "dtype_float", "einsum", "either", "elif", "else", "end", "enumerate", "expand", "eye", "f", "float", "for", "forward", "freq_seq", "functional", "future", "fwd_pos_emb", "fwd_pos_seq", "get_input_embeddings", "h", "have", "head_mask", "heads_to_prune", "hidden_states", "hs", "i", "id", "if", "in", "input_ids", "input_mask", "inputs_embeds", "instead", "int64", "inv_freq", "is", "klen", "kwargs", "last_hidden_state", "layer", "layer_module", "len", "long", "mask", "mask_emb", "mask_lo", "mem_len", "mem_pad", "mems", "mems_mask", "mlen", "n_layer", "new_embeddings", "new_mem", "new_mems", "next", "nn", "non_tgt_mask", "not", "num_classes", "of", "one", "one_hot", "ones", "only", "or", "output", "output_attentions", "output_g", "output_h", "output_hidden_states", "outputs", "padding", "parameters", "perm_mask", "permute", "pos_emb", "pos_seq", "positional_embedding", "post_init", "pow", "prev_mem", "qlen", "r", "raise", "range", "relative_positional_encoding", "removed", "return", "return_dict", "reuse_len", "same", "same_length", "seg_mat", "self", "set_input_embeddings", "shape", "sin", "sinusoid_inp", "specify", "staticmethod", "super", "t", "target_mapping", "the", "time", "to", "token_type_ids", "torch", "training", "transpose", "tril", "triu_", "tuple", "type", "uni", "unsqueeze", "use", "use_cache", "use_mems", "use_mems_eval", "use_mems_train", "use_return_dict", "uses", "v", "version", "vocab_size", "warn", "warnings", "will", "with", "word_emb_k", "word_emb_q", "word_embedding", "zeros"], "xlnet/modeling_xlnet.py:XLNetLMHeadModel": ["CrossEntropyLoss", "GenerationMixin", "Linear", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_reorder_cache", "_tied_weights_keys", "attention_mask", "attentions", "attn_type", "auto_docstring", "beam_idx", "bias", "bool", "cat", "class", "config", "d_model", "def", "device", "dim", "dtype", "dummy_token", "effective_batch_size", "else", "float", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "in", "index_select", "input_ids", "input_mask", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "layer_past", "list", "lm_loss", "logits", "long", "loss", "loss_fct", "mems", "model_inputs", "new_embeddings", "nn", "not", "offset", "output", "output_attentions", "output_hidden_states", "past_key_values", "perm_mask", "pop", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "same_length", "self", "sequence_length", "set_output_embeddings", "shape", "size", "staticmethod", "super", "target_mapping", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_mems", "use_return_dict", "value", "view", "vocab_size", "weight", "zeros"], "xlnet/modeling_xlnet.py:XLNetForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Linear", "MSELoss", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelSequenceSummary", "None", "Optional", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "d_model", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "input_mask", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "logits_proj", "long", "loss", "loss_fct", "mems", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "perm_mask", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_summary", "single_label_classification", "squeeze", "super", "target_mapping", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_mems", "use_return_dict", "view"], "xlnet/modeling_xlnet.py:XLNetForTokenClassification": ["CrossEntropyLoss", "Linear", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "input_mask", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "mems", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "perm_mask", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "target_mapping", "token_type_ids", "torch", "transformer", "tuple", "use_mems", "use_return_dict", "view"], "xlnet/modeling_xlnet.py:XLNetForMultipleChoice": ["CrossEntropyLoss", "Linear", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelSequenceSummary", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "d_model", "def", "else", "flat_attention_mask", "flat_input_ids", "flat_input_mask", "flat_inputs_embeds", "flat_token_type_ids", "forward", "head_mask", "hidden_states", "if", "input_ids", "input_mask", "inputs_embeds", "is", "kwargs", "labels", "logits", "logits_proj", "loss", "loss_fct", "mems", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "perm_mask", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_summary", "shape", "size", "super", "target_mapping", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_mems", "use_return_dict", "view"], "xlnet/modeling_xlnet.py:XLNetForQuestionAnsweringSimple": ["CrossEntropyLoss", "Linear", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "input_mask", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "mems", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "perm_mask", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "target_mapping", "token_type_ids", "torch", "total_loss", "transformer", "tuple", "use_mems", "use_return_dict"], "xlnet/modeling_xlnet.py:XLNetForQuestionAnswering": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Model", "ModelModel", "ModelOutput", "ModelPoolerAnswerClass", "ModelPoolerEndLogits", "ModelPoolerStartLogits", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "and", "answer_class", "attention_mask", "attentions", "auto_docstring", "bh", "bl", "blh", "bool", "bsz", "class", "cls_index", "cls_logits", "cls_loss", "config", "def", "dim", "einsum", "else", "end_log_probs", "end_logits", "end_loss", "end_n_top", "end_positions", "end_top_index", "end_top_log_probs", "expand", "expand_as", "for", "forward", "functional", "gather", "head_mask", "hidden_states", "hidden_states_expanded", "hsz", "if", "in", "input_ids", "input_mask", "inputs_embeds", "is", "is_impossible", "kwargs", "loss", "loss_fct", "loss_fct_cls", "mems", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "p_mask", "perm_mask", "post_init", "r", "return", "return_dict", "self", "size", "slen", "softmax", "squeeze_", "start_log_probs", "start_logits", "start_loss", "start_n_top", "start_positions", "start_states", "start_top_index", "start_top_index_exp", "start_top_log_probs", "super", "target_mapping", "token_type_ids", "topk", "torch", "total_loss", "transformer", "transformer_outputs", "tuple", "unsqueeze", "use_mems", "use_return_dict", "view", "x"], "upernet/modeling_upernet.py:UperNetConvModule": ["BatchNorm2d", "Conv2d", "False", "ModelConvModule", "Module", "None", "ReLU", "Tensor", "Union", "__init__", "activation", "batch_norm", "bias", "bool", "class", "conv", "def", "dilation", "forward", "in_channels", "input", "int", "kernel_size", "nn", "out_channels", "output", "padding", "return", "sModel", "self", "str", "torch", "tuple"], "upernet/modeling_upernet.py:UperNetPyramidPoolingBlock": ["AdaptiveAvgPool2d", "ModelConvModule", "ModelPyramidPoolingBlock", "Module", "None", "Tensor", "__init__", "add_module", "channels", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_channels", "input", "int", "kernel_size", "layer", "layers", "nn", "pool_scale", "return", "sModel", "self", "str", "torch"], "upernet/modeling_upernet.py:UperNetPyramidPoolingModule": ["ModelPyramidPoolingBlock", "ModelPyramidPoolingModule", "Module", "None", "Tensor", "__init__", "add_module", "align_corners", "append", "bilinear", "block", "blocks", "bool", "channels", "class", "def", "enumerate", "for", "forward", "functional", "i", "in", "in_channels", "int", "interpolate", "list", "mode", "nn", "pool_scale", "pool_scales", "ppm", "ppm_out", "ppm_outs", "return", "sModel", "self", "size", "str", "torch", "tuple", "upsampled_ppm_out", "x"], "upernet/modeling_upernet.py:UperNetHead": ["Conv2d", "False", "ModelConvModule", "ModelHead", "ModelPyramidPoolingModule", "Module", "ModuleList", "Tensor", "__init__", "align_corners", "append", "bilinear", "bottleneck", "cat", "channels", "class", "classifier", "config", "def", "dim", "encoder_hidden_states", "enumerate", "extend", "for", "forward", "fpn_bottleneck", "fpn_conv", "fpn_convs", "fpn_outs", "functional", "hidden_size", "i", "in", "in_channels", "inputs", "interpolate", "kernel_size", "l_conv", "lateral_conv", "lateral_convs", "laterals", "len", "mode", "nn", "num_labels", "output", "padding", "pool_scales", "prev_shape", "psp_forward", "psp_modules", "psp_outs", "range", "return", "sModel", "self", "shape", "size", "torch", "used_backbone_levels", "x"], "upernet/modeling_upernet.py:UperNetFCNHead": ["Conv2d", "Identity", "ModelConvModule", "ModelFCNHead", "Module", "None", "Sequential", "Tensor", "Union", "__init__", "append", "auxiliary_channels", "auxiliary_concat_input", "auxiliary_in_channels", "auxiliary_num_convs", "cat", "channels", "class", "classifier", "concat_input", "config", "conv_cat", "conv_padding", "convs", "def", "dilation", "dim", "else", "encoder_hidden_states", "for", "forward", "hidden_states", "i", "if", "in", "in_channels", "in_index", "int", "is", "kernel_size", "nn", "num_convs", "num_labels", "output", "padding", "range", "return", "sModel", "self", "torch", "tuple"], "upernet/modeling_upernet.py:UperNetPreTrainedModel": ["BatchNorm2d", "Conv2d", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "upernet/modeling_upernet.py:UperNetForSemanticSegmentation": ["CrossEntropyLoss", "False", "ModelFCNHead", "ModelForSemanticSegmentation", "ModelHead", "ModelPreTrainedModel", "None", "Optional", "SemanticSegmenterOutput", "Tensor", "The", "Union", "ValueError", "__init__", "align_corners", "and", "attentions", "auto_docstring", "auxiliary_head", "auxiliary_logits", "auxiliary_loss", "auxiliary_loss_weight", "backbone", "be", "bilinear", "bool", "channels", "class", "config", "decode_head", "def", "else", "feature_maps", "features", "forward", "forward_with_filtered_kwargs", "functional", "greater", "hidden_states", "if", "ignore_index", "in_channels", "interpolate", "is", "labels", "load_backbone", "logits", "loss", "loss_fct", "loss_ignore_index", "mode", "nn", "not", "num_labels", "number", "of", "one", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "sModel", "self", "shape", "should", "size", "than", "torch", "tuple", "use_auxiliary_head", "use_return_dict"], "minimax/modeling_minimax.py:MiniMaxRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "minimax/modeling_minimax.py:MiniMaxCache": ["DynamicCache", "ModelCache", "None", "RuntimeError", "Tensor", "_", "__getitem__", "__init__", "__iter__", "__len__", "and", "append", "batch_repeat_interleave", "batch_select_indices", "class", "crop", "def", "dim", "doesnot", "else", "for", "get_linear_cache", "if", "in", "indices", "int", "layer_idx", "layers", "len", "linear_cache", "list", "max", "max_length", "method", "raise", "range", "repeat_interleave", "repeats", "return", "self", "set_linear_cache", "super", "support", "torch", "yield"], "minimax/modeling_minimax.py:MiniMaxLightningAttention": ["ACT2FN", "Cache", "F", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelConfig", "ModelLightningAttention", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "act_fn", "append", "arange", "attention_mask", "attn_output", "attn_output_inter", "attn_output_intra", "attn_weights_inter", "attn_weights_intra", "base", "batch_size", "bias", "block_decay", "block_size", "block_size_range", "bool", "cache_position", "cat", "class", "config", "current_attn_output", "current_attn_weights_inter", "current_block_size", "current_diagonal_decay", "current_key_decay", "current_key_states", "current_query_decay", "current_query_states", "current_value_states", "decay_factors", "def", "deprecate_kwarg", "diagonal_decay", "dim", "dtype", "else", "end_idx", "exp", "exponent", "factor", "float", "for", "forward", "get_linear_cache", "get_slope_rate", "getattr", "head_dim", "hidden_act", "hidden_size", "hidden_states", "i", "if", "in", "inf", "int", "is", "key_decay", "key_states", "kwargs", "layer_idx", "masked_fill", "matmul", "min", "new_name", "next_attn_weights_inter", "nn", "norm", "not", "num_attention_heads", "num_blocks", "num_hidden_layers", "or", "out_proj", "output_gate", "past_key_value", "past_key_values", "position_embeddings", "qkv_proj", "qkv_states", "query_decay", "query_states", "range", "rate", "ratio", "register_buffer", "reshape", "return", "self", "seq_len", "set_linear_cache", "shape", "sigmoid", "slope_rate", "split", "start_idx", "super", "to", "torch", "transpose", "tuple", "unsqueeze", "value_states", "version", "where", "zeros"], "minimax/modeling_minimax.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "minimax/modeling_minimax.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "minimax/modeling_minimax.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "minimax/modeling_minimax.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "minimax/modeling_minimax.py:MiniMaxAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "minimax/modeling_minimax.py:MiniMaxBlockSparseTop2MLP": ["ACT2FN", "False", "Linear", "ModelBlockSparseTop2MLP", "ModelConfig", "Module", "__init__", "act_fn", "bias", "class", "config", "current_hidden_states", "def", "ffn_dim", "forward", "hidden_act", "hidden_dim", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "w1", "w2", "w3"], "minimax/modeling_minimax.py:MiniMaxSparseMoeBlock": ["F", "False", "Linear", "ModelBlockSparseTop2MLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "and", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "empty_like", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "ffn_dim", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "jitter_noise", "keepdim", "nn", "nonzero", "num_classes", "num_experts", "num_experts_per_tok", "num_local_experts", "one_hot", "permute", "range", "reshape", "return", "router_jitter_noise", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "training", "uniform_", "view", "where", "zeros"], "minimax/modeling_minimax.py:MiniMaxDecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLightningAttention", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "attention_mask", "attn_alpha_factor", "attn_beta_factor", "block_sparse_moe", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "full_attn_alpha_factor", "full_attn_beta_factor", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_type", "layer_types", "linear_attention", "linear_attn_alpha_factor", "linear_attn_beta_factor", "mlp_alpha_factor", "mlp_beta_factor", "new_name", "output_attentions", "output_router_logits", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "minimax/modeling_minimax.py:MiniMaxPreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLightningAttention", "ModelPreTrainedModel", "ModelSparseMoeBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "index", "model", "past_key_values", "router_logits", "supports_gradient_checkpointing"], "minimax/modeling_minimax.py:MiniMaxRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "minimax/modeling_minimax.py:MiniMaxModel": ["Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelCache", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache", "cache_position", "causal_mask", "check_model_inputs", "class", "compatible", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "elif", "else", "embed_tokens", "eps", "exactly", "f", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_attention_mask", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "its", "kwargs", "last_hidden_state", "layer_idx", "layer_type", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "own", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "type", "unsqueeze", "use_cache", "uses", "vocab_size", "with"], "minimax/modeling_minimax.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "minimax/modeling_minimax.py:MiniMaxForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "minimax/modeling_minimax.py:MiniMaxForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "minimax/modeling_minimax.py:MiniMaxForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "minimax/modeling_minimax.py:MiniMaxForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "class", "pass"], "xlstm/modeling_xlstm.py:small_init_method": ["Model_init_method", "def", "dim", "init", "init_", "mean", "nn", "normal_", "return", "std", "tensor", "torch"], "xlstm/modeling_xlstm.py:wang_init_method": ["Model_init_method", "def", "dim", "init", "init_", "mean", "n_layers", "nn", "normal_", "return", "std", "tensor", "torch"], "xlstm/modeling_xlstm.py:xLSTMPreTrainedModel": ["Embedding", "Linear", "Model", "ModelBlock", "ModelConfig", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_layer_normalize", "_module_name_map", "_no_split_modules", "and", "backbone", "base_model_prefix", "bias", "class", "config", "config_class", "copy_", "def", "device", "dim", "dtype", "elif", "embeddings", "fgate", "for", "fused", "gate", "hasattr", "hidden_size", "if", "igate", "in", "init", "is", "isinstance", "linspace", "mod", "module", "n_layers", "name", "named_modules", "nn", "no_grad", "not", "num_heads", "num_hidden_layers", "ones_", "ones_like", "or", "out_proj", "proj_down", "return", "self", "shape", "single", "small_init_method", "supports_gradient_checkpointing", "to", "torch", "wang_init_method", "weight", "weight_mode", "with", "zeros_"], "xlstm/modeling_xlstm.py:xLSTMCache": ["Model", "ModelConfig", "None", "Optional", "__init__", "bfloat16", "class", "config", "def", "device", "dtype", "for", "in", "int", "kwargs", "layer", "max_batch_size", "num_heads", "num_hidden_layers", "qk_head_dim", "range", "reset", "rnn_state", "self", "seqlen_offset", "str", "torch", "v_head_dim", "zeros", "zeros_like"], "xlstm/modeling_xlstm.py:xLSTMOutput": ["FloatTensor", "Model", "ModelCache", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "xlstm/modeling_xlstm.py:xLSTMModel": ["Embedding", "False", "LongTensor", "Model", "ModelBlock", "ModelCache", "ModelOutput", "ModelPreTrainedModel", "ModelRMSNorm", "Model_block", "ModuleList", "None", "Optional", "Union", "ValueError", "You", "_", "__call__", "__init__", "_gradient_checkpointing_func", "all_hidden_states", "and", "auto_docstring", "blocks", "bool", "cache_params", "can_return_tuple", "class", "config", "copy_", "def", "device", "dtype", "else", "embedding_dim", "embeddings", "enumerate", "eps", "exactly", "final_state", "for", "forward", "get_input_embeddings", "gradient_checkpointing", "hidden_size", "hidden_states", "hidden_states_chunk", "if", "in", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "len", "local_rnn_state", "max_batch_size", "max_inference_chunksize", "min", "must", "new_embedding", "nn", "no_grad", "norm_eps", "not", "num_blocks", "of", "offset", "one", "or", "out_norm", "output_hidden_states", "post_init", "r", "raise", "range", "return", "rnn_state", "rnn_state_initial", "self", "seqlen_offset", "set_input_embeddings", "shape", "size", "specify", "state", "state_idx", "super", "torch", "training", "tuple", "use_cache", "vocab_size", "while", "with", "zeros_like"], "xlstm/modeling_xlstm.py:xLSTMCausalLMOutput": ["FloatTensor", "Model", "ModelCache", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "xlstm/modeling_xlstm.py:xLSTMForCausalLM": ["CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelCache", "ModelCausalLMOutput", "ModelModel", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Union", "__init__", "and", "attention_mask", "auto_docstring", "backbone", "bias", "bool", "cache_params", "can_return_tuple", "class", "config", "contiguous", "def", "device", "dtype", "else", "float", "for", "forward", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "lm_head", "logits", "loss", "loss_fct", "max_inference_chunksize", "min", "model_inputs", "new_embeddings", "nn", "no_grad", "not", "offset", "output_hidden_states", "output_logit_soft_cap", "post_init", "prepare_inputs_for_generation", "r", "return", "self", "set_input_embeddings", "set_output_embeddings", "shape", "shift_labels", "shift_logits", "size", "soft_cap", "super", "to", "torch", "training", "tuple", "update", "use_cache", "value", "view", "vocab_size", "weight", "while", "with"], "seed_oss/modeling_seed_oss.py:SeedOssRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "seed_oss/modeling_seed_oss.py:SeedOssMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "dropout", "forward", "functional", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "p", "residual_dropout", "return", "self", "super", "training", "up_proj", "x"], "seed_oss/modeling_seed_oss.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "seed_oss/modeling_seed_oss.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "seed_oss/modeling_seed_oss.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "seed_oss/modeling_seed_oss.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "seed_oss/modeling_seed_oss.py:SeedOssAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_out_bias", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "functional", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "p", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "residual_dropout", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "seed_oss/modeling_seed_oss.py:SeedOssDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "seed_oss/modeling_seed_oss.py:SeedOssPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "seed_oss/modeling_seed_oss.py:SeedOssRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "seed_oss/modeling_seed_oss.py:SeedOssModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "seed_oss/modeling_seed_oss.py:SeedOssForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "seed_oss/modeling_seed_oss.py:SeedOssForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "seed_oss/modeling_seed_oss.py:SeedOssForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "seed_oss/modeling_seed_oss.py:SeedOssForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerModelOutput": ["FloatTensor", "LongTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "decoder_attentions", "decoder_hidden_states", "duration_outputs", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "energy_outputs", "loss", "pitch_outputs", "r", "spectrogram", "torch", "tuple"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerWithHifiGanOutput": ["FloatTensor", "ModelModelOutput", "ModelWithHifiGanOutput", "None", "Optional", "class", "r", "torch", "waveform"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:length_regulator": ["Model_regulator", "ValueError", "be", "def", "device", "dim", "dtype", "duration_labels", "elif", "encoded_embedding", "encoded_embeddings", "enumerate", "eq", "float", "for", "greater", "hidden_states", "i", "if", "in", "long", "max", "max_len", "must", "raise", "repeat_interleave", "repeated", "return", "round", "size", "speaking_speed", "sum", "target_duration", "than", "torch", "zeros", "zip"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerDurationPredictor": ["Linear", "ModelConfig", "ModelDurationPredictor", "ModelPredictorLayer", "Module", "ModuleList", "__init__", "append", "clamp", "class", "config", "conv_layers", "def", "duration_predictor_channels", "duration_predictor_dropout_rate", "duration_predictor_kernel_size", "duration_predictor_layers", "else", "encoder_hidden_states", "exp", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_channels", "layer", "layer_idx", "linear", "log_domain_offset", "long", "min", "nn", "not", "num_chans", "range", "return", "round", "self", "squeeze", "super", "torch", "training", "transpose"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerBatchNormConvLayer": ["BatchNorm1d", "Conv1d", "Dropout", "False", "ModelBatchNormConvLayer", "Module", "None", "Tanh", "__init__", "activation", "batch_norm", "bias", "class", "config", "conv", "def", "dropout", "else", "forward", "hidden_states", "if", "in_conv_dim", "is", "kernel_size", "layer_id", "nn", "not", "num_mel_bins", "out_conv_dim", "padding", "return", "self", "speech_decoder_postnet_dropout", "speech_decoder_postnet_kernel", "speech_decoder_postnet_layers", "speech_decoder_postnet_units", "stride", "super"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerSpeechDecoderPostnet": ["Linear", "ModelBatchNormConvLayer", "ModelSpeechDecoderPostnet", "Module", "ModuleList", "Tensor", "__init__", "class", "config", "def", "feat_out", "for", "forward", "hidden_size", "hidden_states", "i", "in", "layer", "layer_output", "layers", "nn", "num_mel_bins", "outputs_after_postnet", "outputs_before_postnet", "range", "reduction_factor", "return", "self", "size", "speech_decoder_postnet_layers", "super", "torch", "transpose", "view"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerPredictorLayer": ["Conv1d", "Dropout", "LayerNorm", "ModelPredictorLayer", "Module", "ReLU", "__init__", "activation", "class", "conv", "def", "dropout", "dropout_rate", "forward", "hidden_states", "input_channels", "kernel_size", "layer_norm", "nn", "num_chans", "padding", "return", "self", "stride", "super", "transpose"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerVariancePredictor": ["Linear", "ModelConfig", "ModelPredictorLayer", "ModelVariancePredictor", "Module", "ModuleList", "None", "__init__", "append", "class", "config", "conv_layers", "def", "dropout_rate", "else", "encoder_hidden_states", "for", "forward", "hidden_size", "hidden_states", "idx", "if", "in", "input_channels", "is", "kernel_size", "layer", "linear", "masked_fill", "nn", "not", "num_chans", "num_layers", "padding_masks", "range", "return", "self", "super", "transpose"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerVarianceEmbedding": ["Conv1d", "Dropout", "ModelVarianceEmbedding", "Module", "__init__", "class", "conv", "def", "dropout", "dropout_rate", "forward", "hidden_states", "in_channels", "kernel_size", "nn", "out_channels", "padding", "return", "self", "super", "transpose"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerAttention": ["Attention", "Dropout", "False", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "_", "__init__", "attention_dropout_rate", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bsz", "bsz_pos", "but", "cat", "class", "config", "contiguous", "def", "device", "dim", "dim_key", "dropout", "dtype", "else", "eq", "expected_size", "f", "finfo", "float", "forward", "head_dim", "hidden_size", "hidden_states", "if", "is", "key_states", "linear_k", "linear_out", "linear_pos", "linear_q", "linear_v", "mask", "masked_fill", "math", "matmul", "matrix_ac", "matrix_bd", "min", "min_value", "module_config", "nn", "not", "num_attention_heads", "num_heads", "of", "output_attentions", "p", "permute", "pos_bias_u", "pos_bias_v", "pos_emb", "pos_encoding", "pos_tensor", "pos_tensor_padded", "q_len", "query_states", "query_with_bias_u", "query_with_bias_v", "raise", "return", "scores", "self", "shift_relative_position_tensor", "should", "size", "softmax", "sqrt", "super", "torch", "transpose", "tuple", "unsqueeze", "value_states", "view", "view_as", "zero_pad", "zeros"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerConvolutionModule": ["ACT2FN", "BatchNorm1d", "Conv1d", "ModelConfig", "ModelConvolutionModule", "Module", "None", "True", "__init__", "activation", "all", "all_masked_rows", "attention_mask", "bias", "channels", "class", "config", "conv_kernel_size", "def", "depthwise_conv", "dim", "else", "forward", "functional", "get", "getattr", "glu", "groups", "hidden_act", "hidden_size", "hidden_states", "if", "is", "kernel_size", "masked_fill", "module_config", "nn", "norm", "not", "padding", "pointwise_conv1", "pointwise_conv2", "return", "self", "silu", "stride", "super", "torch", "transpose"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerEncoderLayer": ["Dropout", "False", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelConvolutionModule", "ModelEncoderLayer", "ModelMultiLayeredConv1d", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "attention_output", "attention_scores", "cat", "class", "concat_after", "concat_linear", "config", "conv_layer_norm", "conv_module", "def", "dim", "dropout", "dropout_rate", "else", "feed_forward", "feed_forward_macaron", "ff_layer_norm", "ff_macaron_layer_norm", "ff_scale", "final_layer_norm", "forward", "hidden_size", "hidden_states", "if", "is", "macaron_style", "module_config", "nn", "normalize_before", "not", "output_attentions", "outputs", "pos_emb", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "size", "super", "torch", "use_cnn_in_conformer", "use_cnn_module", "use_macaron_style_in_conformer", "x_concat"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerMultiLayeredConv1d": ["Conv1d", "Dropout", "ModelConfig", "ModelMultiLayeredConv1d", "Module", "__init__", "class", "config", "conv1", "conv2", "def", "dropout", "dropout_rate", "forward", "hidden_channels", "hidden_size", "hidden_states", "input_channels", "kernel_size", "linear_units", "module_config", "nn", "padding", "positionwise_conv_kernel_size", "relu", "return", "self", "stride", "super", "torch", "transpose"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerRelPositionalEncoding": ["Dropout", "ModelConfig", "ModelRelPositionalEncoding", "Module", "None", "__init__", "arange", "cat", "center_idx", "class", "config", "cos", "def", "device", "dim", "div_term", "dropout", "dtype", "embed_dim", "exp", "expand", "extend_pos_enc", "feature_representation", "flip", "float", "forward", "hidden_size", "hidden_states", "if", "input_scale", "int64", "is", "log", "math", "max_len", "module_config", "nn", "not", "or", "p", "pos_emb", "pos_enc", "pos_enc_negative", "pos_enc_positive", "position", "positional_dropout_rate", "return", "self", "sin", "size", "sqrt", "super", "tensor", "to", "torch", "unsqueeze", "x", "zeros"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerEncoder": ["BaseModelOutput", "Embedding", "False", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelRelPositionalEncoding", "Module", "ModuleList", "None", "Optional", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "class", "config", "conformer_layer", "conformer_layers", "def", "else", "embed", "embedding_dim", "feature_representation", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_tensor", "is", "last_hidden_state", "layer_outputs", "layers", "module_config", "nn", "not", "num_embeddings", "output_attentions", "output_hidden_states", "padding_idx", "pos_emb", "pos_enc", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_encoder_input_layer", "v", "vocab_size"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerLoss": ["Either", "False", "L1Loss", "MSELoss", "ModelConfig", "ModelLoss", "Module", "None", "True", "ValueError", "__init__", "and", "be", "both", "but", "can", "class", "config", "def", "dim", "duration_criterion", "duration_labels", "duration_loss", "duration_mask", "duration_outputs", "duration_weights", "else", "energy_labels", "energy_loss", "energy_outputs", "float", "forward", "functional", "if", "is", "keepdim", "l1_criterion", "l1_loss", "log", "log_domain_offset", "masked_select", "mean", "mse_criterion", "mul", "nn", "none", "not", "or", "out_weights", "outputs_after_postnet", "outputs_before_postnet", "pad", "pitch_and_energy_masks", "pitch_labels", "pitch_loss", "pitch_outputs", "pitch_weights", "raise", "reduction", "return", "self", "size", "spectrogram_labels", "spectrogram_mask", "sum", "super", "torch", "transpose", "unsqueeze", "use_masking", "use_weighted_masking", "value"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerPreTrainedModel": ["BatchNorm1d", "Conv1d", "Embedding", "False", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEncoder", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_set_gradient_checkpointing", "a", "b", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "gradient_checkpointing", "groups", "if", "in_channels", "init", "input_ids", "is", "isinstance", "kaiming_normal_", "kernel_size", "key", "main_input_name", "math", "module", "nn", "normal_", "not", "padding_idx", "pos_bias_u", "pos_bias_v", "self", "size", "sqrt", "std", "uniform_", "value", "weight", "xavier_uniform_", "zero_"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerModel": ["All", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelDurationPredictor", "ModelEncoder", "ModelLoss", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModelSpeechDecoderPostnet", "ModelVarianceEmbedding", "ModelVariancePredictor", "None", "Optional", "True", "Union", "ValueError", "__init__", "and", "any", "attention_mask", "attentions", "audio_feature_predictions", "auto_docstring", "be", "bool", "cat", "class", "config", "criterion", "decoder", "decoder_attentions", "decoder_config", "decoder_hidden_states", "decoder_outputs", "def", "detach", "device", "dim", "dropout_rate", "duration_labels", "duration_mask", "duration_outputs", "duration_predictions", "duration_predictor", "else", "embedded_energy_curve", "embedded_pitch_curve", "embeddings_expanded", "encoder", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "energy_embed", "energy_embed_dropout", "energy_embed_kernel_size", "energy_labels", "energy_outputs", "energy_predictions", "energy_predictor", "energy_predictor_channels", "energy_predictor_dropout", "energy_predictor_kernel_size", "energy_predictor_layers", "expand", "forward", "functional", "has_missing_labels", "hidden_mask", "hidden_size", "hidden_states", "if", "in", "input_ids", "int", "is", "kernel_size", "labels", "lang_ids", "language_id_embbedings", "language_id_embedding", "last_hidden_state", "length_dim", "length_regulator", "loss", "loss_duration_mask", "loss_spectrogram_mask", "masked_fill", "mode", "multilingual_model", "multispeaker_model", "must", "nn", "normalize", "not", "num_chans", "num_languages", "num_layers", "num_mel_bins", "num_speakers", "ones", "or", "out_channels", "output_attentions", "output_hidden_states", "outputs", "outputs_after_postnet", "outputs_before_postnet", "padding", "pitch_embed", "pitch_embed_dropout", "pitch_embed_kernel_size", "pitch_labels", "pitch_outputs", "pitch_predictions", "pitch_predictor", "pitch_predictor_channels", "pitch_predictor_dropout", "pitch_predictor_kernel_size", "pitch_predictor_layers", "post_init", "postnet_outputs", "projection", "provided", "r", "raise", "reduction_factor", "return", "return_dict", "run", "self", "shape", "size", "speaker_embed_dim", "speaker_embedding", "speaker_id_embedding", "speaker_id_embeddings", "speaker_ids", "speaking_speed", "spectrogram", "spectrogram_labels", "spectrogram_mask", "speech_decoder_postnet", "stop_gradient_from_energy_predictor", "stop_gradient_from_pitch_predictor", "super", "text_masks", "to", "torch", "training", "tuple", "unsqueeze", "use_encoder_input_layer", "use_return_dict", "view", "vocab_size"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:HifiGanResidualBlock": ["Conv1d", "ModelGanResidualBlock", "Module", "ModuleList", "_", "__init__", "apply_weight_norm", "channels", "class", "conv1", "conv2", "convs1", "convs2", "def", "dilation", "for", "forward", "functional", "get_padding", "hasattr", "hidden_states", "i", "if", "in", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "nn", "padding", "parametrizations", "range", "remove_weight_norm", "residual", "return", "self", "stride", "super", "utils", "weight_norm", "zip"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerHifiGan": ["Conv1d", "ConvTranspose1d", "FloatTensor", "HifiGanResidualBlock", "ModelHifiGan", "ModelHifiGanConfig", "Module", "ModuleList", "None", "PreTrainedModel", "__init__", "_init_weights", "append", "apply_weight_norm", "auto_docstring", "bias", "channels", "class", "config", "conv_post", "conv_pre", "custom_intro", "data", "def", "dilation", "dim", "else", "enumerate", "for", "forward", "functional", "hasattr", "hidden_states", "i", "if", "in", "initializer_range", "is", "is_batched", "isinstance", "j", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "main_input_name", "mean", "model_in_dim", "module", "nn", "normal_", "normalize_before", "not", "num_kernels", "num_upsamples", "ones", "padding", "parametrizations", "post_init", "r", "range", "register_buffer", "remove_weight_norm", "res_state", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "scale", "self", "spectrogram", "squeeze", "std", "stride", "super", "tanh", "torch", "transpose", "unsqueeze", "upsample_initial_channel", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "utils", "view", "waveform", "weight", "weight_norm", "zero_", "zeros", "zip"], "fastspeech2_conformer/modeling_fastspeech2_conformer.py:FastSpeech2ConformerWithHifiGan": ["FloatTensor", "LongTensor", "ModelHifiGan", "ModelModel", "ModelModelOutput", "ModelWithHifiGan", "ModelWithHifiGanConfig", "ModelWithHifiGanOutput", "None", "Optional", "PreTrainedModel", "Union", "__init__", "attention_mask", "auto_docstring", "bool", "class", "config", "def", "duration_labels", "else", "energy_labels", "forward", "has_missing_labels", "if", "input_ids", "is", "lang_ids", "model", "model_config", "model_outputs", "not", "or", "output_attentions", "output_hidden_states", "pitch_labels", "r", "return", "return_dict", "self", "speaker_embedding", "speaker_ids", "spectrogram", "spectrogram_labels", "super", "torch", "tuple", "use_return_dict", "vocoder", "vocoder_config", "waveform"], "bert/modeling_bert.py:BertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "shape", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "bert/modeling_bert.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "bert/modeling_bert.py:BertSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "bert/modeling_bert.py:BertCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "bert/modeling_bert.py:BertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bert/modeling_bert.py:BertAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "bert/modeling_bert.py:BertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "bert/modeling_bert.py:BertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bert/modeling_bert.py:BertLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "bert/modeling_bert.py:BertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "bert/modeling_bert.py:BertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "bert/modeling_bert.py:BertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "bert/modeling_bert.py:BertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "bert/modeling_bert.py:BertOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "bert/modeling_bert.py:BertOnlyNSPHead": ["Linear", "ModelOnlyNSPHead", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "bert/modeling_bert.py:BertPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "bert/modeling_bert.py:BertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLMPredictionHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "bert/modeling_bert.py:BertForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "bert/modeling_bert.py:BertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "bert/modeling_bert.py:BertForPreTraining": ["CrossEntropyLoss", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "can_return_tuple", "class", "cls", "config", "decoder", "def", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "next_sentence_label", "next_sentence_loss", "not", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "total_loss", "tuple", "view", "vocab_size", "weight"], "bert/modeling_bert.py:BertLMHeadModel": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "Model", "ModelLMHeadModel", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "cls", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "list", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "bert/modeling_bert.py:BertForMaskedLM": ["CrossEntropyLoss", "False", "If", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bi", "bias", "bool", "can_generate", "can_return_tuple", "cat", "class", "classmethod", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "bert/modeling_bert.py:BertForNextSentencePrediction": ["CrossEntropyLoss", "FutureWarning", "Model", "ModelForNextSentencePrediction", "ModelModel", "ModelOnlyNSPHead", "ModelPreTrainedModel", "NextSentencePredictorOutput", "None", "Optional", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "a", "and", "argument", "attention_mask", "attentions", "auto_docstring", "be", "can_return_tuple", "class", "cls", "config", "def", "deprecated", "forward", "future", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "outputs", "pooled_output", "pop", "position_ids", "post_init", "r", "removed", "return", "return_dict", "self", "seq_relationship_scores", "super", "token_type_ids", "torch", "tuple", "use", "version", "view", "warn", "warnings", "will"], "bert/modeling_bert.py:BertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "bert/modeling_bert.py:BertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "bert/modeling_bert.py:BertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "view"], "bert/modeling_bert.py:BertForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "stablelm/modeling_stablelm.py:StableLmRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "stablelm/modeling_stablelm.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "stablelm/modeling_stablelm.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "stablelm/modeling_stablelm.py:StableLmMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "stablelm/modeling_stablelm.py:StableLmLayerNormPerHead": ["False", "LayerNorm", "ModelLayerNormPerHead", "Module", "ModuleList", "Tensor", "_", "__init__", "bias", "cat", "class", "def", "dim", "eps", "for", "forward", "hidden_states", "in", "nn", "norm", "norms", "num_heads", "range", "return", "self", "split", "states_per_heads", "super", "torch", "zip"], "stablelm/modeling_stablelm.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "stablelm/modeling_stablelm.py:StableLmAttention": ["Cache", "Dropout", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelLayerNormPerHead", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "cat", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dtype", "during", "eps", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_layernorm", "k_proj", "key_pass", "key_rot", "key_states", "layer_idx", "layer_norm_eps", "lead", "logger", "make", "math", "matmul", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "partial_rotary_factor", "partial_rotation_size", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_layernorm", "q_len", "q_proj", "qk_layernorm", "query_pass", "query_rot", "query_states", "raise", "recommended", "repeat_kv", "reshape", "return", "rope_theta", "rotary_emb", "rotary_ndims", "self", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "transpose", "tuple", "update", "use_cache", "use_qkv_bias", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "stablelm/modeling_stablelm.py:StableLmSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "cat", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "dim", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_size", "hidden_states", "if", "implementation", "is", "is_causal", "k_layernorm", "k_proj", "key_pass", "key_rot", "key_states", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "p", "partial_rotation_size", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_layernorm", "q_len", "q_proj", "qk_layernorm", "query_pass", "query_rot", "query_states", "removed", "repeat_kv", "required", "return", "rotary_ndims", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "stablelm/modeling_stablelm.py:StableLmFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "apply_rotary_pos_emb", "args", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "bool", "bsz", "cache_kwargs", "cache_position", "cat", "class", "contiguous", "cos", "def", "deprecate_kwarg", "dim", "dropout", "dropout_rate", "else", "flash_attn_supports_top_left_mask", "forward", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_layernorm", "k_proj", "key_pass", "key_rot", "key_states", "kwargs", "layer_idx", "new_name", "not", "num_heads", "num_key_value_heads", "o_proj", "output_attentions", "p", "partial_rotation_size", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_layernorm", "q_len", "q_proj", "qk_layernorm", "query_pass", "query_rot", "query_states", "reshape", "return", "rotary_ndims", "self", "sin", "size", "super", "torch", "training", "transpose", "tuple", "update", "use_cache", "use_top_left_mask", "v_proj", "value_states", "version", "view"], "stablelm/modeling_stablelm.py:StableLmDecoderLayer": ["ATTENTION_CLASSES", "Cache", "Dropout", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "dropout", "else", "eps", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "layer_idx", "layer_norm_eps", "mlp", "mlp_output", "nn", "not", "output_attentions", "outputs", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "self_attn_output", "self_attn_weights", "super", "torch", "tuple", "use_cache", "use_parallel_residual"], "stablelm/modeling_stablelm.py:StableLmPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "stablelm/modeling_stablelm.py:StableLmModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_eps", "layer_outputs", "layers", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "rotary_emb", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "type", "unsqueeze", "use_cache", "using_compilable_cache", "vocab_size", "warning_once", "with", "xpu"], "stablelm/modeling_stablelm.py:StableLmForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "stablelm/modeling_stablelm.py:StableLmForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "stablelm/modeling_stablelm.py:StableLmForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class"], "llava/modeling_llava.py:LlavaModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "llava/modeling_llava.py:LlavaCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "llava/modeling_llava.py:LlavaMultiModalProjector": ["ACT2FN", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "bias", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "image_features", "int", "isinstance", "len", "linear_1", "linear_2", "multimodal_projector_bias", "nn", "num_feature_layers", "projector_hidden_act", "return", "self", "super", "text_config", "vision_config", "vision_feature_layer"], "llava/modeling_llava.py:LlavaPreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "past_key_values", "supports_gradient_checkpointing"], "llava/modeling_llava.py:LlavaModel": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unexpected", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "default", "device", "dim", "do", "dtype", "else", "exactly", "expand_as", "f", "feature", "features", "for", "forward", "from_config", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "height", "hidden_states", "hs", "hs_pool", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_sizes", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "k", "kwargs", "language_model", "last_hidden_state", "layer_idx", "list", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_hidden_states", "outputs", "past_key_values", "patch_size", "pixel_values", "position_ids", "post_init", "raise", "return", "select", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "split", "split_sizes", "squeeze", "str", "strategy", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "v", "value", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "width"], "llava/modeling_llava.py:LlavaForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "str", "super", "text_config", "torch", "tuple", "value", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "roformer/modeling_roformer.py:RoFormerSinusoidalPositionalEmbedding": ["Embedding", "False", "FloatTensor", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_init_weight", "arange", "array", "bsz", "class", "cos", "def", "device", "dim", "dtype", "else", "embedding_dim", "empty", "for", "forward", "if", "in", "input_ids_shape", "int", "is", "j", "long", "n_pos", "nn", "no_grad", "np", "num_positions", "out", "padding_idx", "past_key_values_length", "pos", "position_enc", "position_ids", "power", "range", "requires_grad", "return", "self", "sentinel", "seq_len", "shape", "sin", "super", "torch", "weight"], "roformer/modeling_roformer.py:RoFormerEmbeddings": ["Dropout", "Embedding", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "class", "config", "def", "device", "dropout", "dtype", "else", "embedding_size", "embeddings", "eps", "forward", "hidden_dropout_prob", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "nn", "not", "pad_token_id", "padding_idx", "return", "self", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "roformer/modeling_roformer.py:RoFormerSelfAttention": ["Dropout", "EncoderDecoderCache", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "True", "ValueError", "_", "__init__", "a", "all_head_size", "and", "apply_rotary_position_embeddings", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "cache_position", "chunk", "class", "config", "context_layer", "contiguous", "cos", "cos_pos", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "dropout", "else", "embedding_size", "encoder_hidden_states", "f", "forward", "functional", "get", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "key", "key_layer", "keys", "layer_idx", "layers", "math", "matmul", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "past_key_value", "past_key_values", "permute", "query", "query_layer", "raise", "reshape_as", "return", "rotary_value", "rotate_half_key_layer", "rotate_half_query_layer", "rotate_half_value_layer", "self", "self_attention_cache", "seq_length", "shape", "sin", "sin_pos", "sinusoidal_pos", "size", "softmax", "sqrt", "stack", "staticmethod", "super", "the", "torch", "transpose", "update", "value", "value_layer", "values", "version", "view"], "roformer/modeling_roformer.py:RoFormerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "roformer/modeling_roformer.py:RoFormerAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "cache_position", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "layer_idx", "len", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "sinusoidal_pos", "super", "union", "value", "version"], "roformer/modeling_roformer.py:RoFormerIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "roformer/modeling_roformer.py:RoFormerOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "roformer/modeling_roformer.py:RoFormerLayer": ["False", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "True", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "deprecate_kwarg", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_idx", "layer_output", "layers", "model", "new_name", "not", "output", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "raise", "return", "self", "self_attention_outputs", "seq_len_dim", "setting", "should", "sinusoidal_pos", "super", "to", "used", "version", "with"], "roformer/modeling_roformer.py:RoFormerEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "EncoderDecoderCache", "False", "ModelEncoder", "ModelLayer", "ModelSinusoidalPositionalEmbedding", "Module", "ModuleList", "None", "Passing", "Setting", "Transformers", "True", "You", "__init__", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "attention_mask", "attentions", "be", "cache_position", "checkpointing", "class", "config", "cross_attentions", "def", "deprecated", "e", "else", "embed_positions", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "from_legacy_cache", "g", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_outputs", "logger", "max_position_embeddings", "nn", "not", "num_attention_heads", "num_hidden_layers", "of", "output_attentions", "output_hidden_states", "pass", "past_key_values", "past_key_values_length", "range", "removed", "return", "return_dict", "self", "shape", "should", "sinusoidal_pos", "super", "training", "tuple", "use_cache", "v", "v4", "warning_once", "will", "with"], "roformer/modeling_roformer.py:RoFormerSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "ModelConfig", "ModelSequenceSummary", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "roformer/modeling_roformer.py:RoFormerPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "embedding_size", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "roformer/modeling_roformer.py:RoFormerLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "embedding_size", "forward", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "roformer/modeling_roformer.py:RoFormerOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "roformer/modeling_roformer.py:RoFormerPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "None", "PreTrainedModel", "True", "_init_weight", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "roformer/modeling_roformer.py:RoFormerModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "False", "FloatTensor", "Linear", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cache_position", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embedding_size", "embeddings", "embeddings_project", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "long", "nn", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_key_values_length", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "roformer/modeling_roformer.py:RoFormerForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "Tensor", "The", "Union", "__init__", "_tied_weights_keys", "assert", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bi", "bias", "bool", "cat", "class", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "roformer/modeling_roformer.py:RoFormerForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "FloatTensor", "GenerationMixin", "If", "LongTensor", "Model", "ModelForCausalLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "a", "add", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "cls", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "use_return_dict", "vocab_size", "want", "warning", "weight", "you"], "roformer/modeling_roformer.py:RoFormerClassificationHead": ["ACT2FN", "Dropout", "Linear", "ModelClassificationHead", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "x"], "roformer/modeling_roformer.py:RoFormerForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "roformer/modeling_roformer.py:RoFormerForMultipleChoice": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_output", "sequence_summary", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "roformer/modeling_roformer.py:RoFormerForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "roformer/modeling_roformer.py:RoFormerForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "gpt_neo/modeling_gpt_neo.py:GPTNeoSelfAttention": ["Dropout", "False", "Linear", "Model", "Module", "None", "True", "ValueError", "__init__", "_attn", "_merge_heads", "_split_heads", "and", "attention_dropout", "attention_mask", "attention_type", "attn_dropout", "attn_head_size", "attn_output", "attn_weights", "be", "bias", "bitwise_xor", "bool", "by", "cache_kwargs", "cache_position", "causal_mask", "class", "config", "contiguous", "def", "device", "dim", "divisible", "dtype", "embed_dim", "f", "finfo", "float", "float32", "forward", "functional", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "key", "key_length", "layer_id", "layer_past", "local", "mask_value", "masked_bias", "matmul", "max_position_embeddings", "max_positions", "min", "must", "new_shape", "nn", "not", "num_heads", "ones", "out_proj", "output_attentions", "permute", "persistent", "q_proj", "query", "query_length", "raise", "register_buffer", "resid_dropout", "return", "self", "shape", "size", "softmax", "super", "tensor", "to", "torch", "transpose", "tril", "update", "use_cache", "v_proj", "value", "view", "where", "window_size"], "gpt_neo/modeling_gpt_neo.py:GPTNeoFlashAttention2": ["False", "Model", "ModelSelfAttention", "None", "The", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "_split_heads", "args", "attention_dropout", "attention_mask", "attn_dropout", "attn_output", "attn_weights_reshaped", "back", "be", "bsz", "cache_kwargs", "cache_position", "cast", "casted", "class", "config", "cpu", "def", "device", "device_type", "dropout", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "have", "head_dim", "head_mask", "hidden", "hidden_states", "if", "in", "input", "is", "is_autocast_enabled", "is_causal", "k_proj", "key", "kwargs", "layer", "layer_id", "layer_past", "layers", "logger", "might", "mps", "norm", "not", "num_heads", "or", "out_proj", "output_attentions", "q_proj", "query", "query_length", "related", "reshape", "resid_dropout", "return", "seems", "self", "shape", "silently", "size", "softmax_scale", "states", "super", "target_dtype", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "type", "upcasted", "update", "use_cache", "use_top_left_mask", "v_proj", "value", "view", "warning_once", "weight", "will", "you"], "gpt_neo/modeling_gpt_neo.py:GPTNeoAttention": ["False", "Model", "Model_ATTENTION_CLASSES", "Module", "None", "NotImplementedError", "Only", "Select", "__init__", "_attn_implementation", "and", "attention", "attention_layers", "attention_mask", "attention_type", "attn", "but", "cache_position", "class", "config", "def", "else", "exist", "f", "forward", "from", "global", "got", "head_mask", "hidden_states", "if", "in", "layer", "layer_id", "layer_past", "local", "nn", "only", "output_attentions", "raise", "return", "self", "super", "types", "use_cache"], "gpt_neo/modeling_gpt_neo.py:GPTNeoMLP": ["ACT2FN", "Dropout", "Linear", "Model", "Module", "__init__", "act", "activation_function", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "float", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "resid_dropout", "return", "self", "super"], "gpt_neo/modeling_gpt_neo.py:GPTNeoBlock": ["False", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelMLP", "None", "__init__", "attention_mask", "attn", "attn_output", "attn_weights", "cache_position", "class", "config", "def", "else", "eps", "feed_forward_hidden_states", "forward", "head_mask", "hidden_size", "hidden_states", "if", "inner_dim", "intermediate_size", "is", "layer_id", "layer_norm_epsilon", "layer_past", "ln_1", "ln_2", "mlp", "nn", "not", "output_attentions", "residual", "return", "self", "super", "use_cache"], "gpt_neo/modeling_gpt_neo.py:GPTNeoPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "Model", "ModelBlock", "ModelConfig", "None", "PreTrainedModel", "True", "__init__", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "inputs", "is", "isinstance", "kwargs", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "gpt_neo/modeling_gpt_neo.py:GPTNeoModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelBlock", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attentions", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "block", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "drop", "dtype", "else", "embed_dim", "embed_dropout", "enumerate", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "float", "for", "forward", "full", "get_head_mask", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "h", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_id", "layer_norm_epsilon", "layer_past", "ln_f", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "max_position_embeddings", "min", "min_dtype", "must", "new_embeddings", "nn", "not", "npu", "num_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeds", "position_ids", "post_init", "r", "raise", "range", "reshape", "return", "return_dict", "sdpa", "self", "seq_length", "sequence_length", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "target_length", "to", "token_type_embeds", "token_type_ids", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "view", "vocab_size", "warning_once", "with", "wpe", "wte", "xpu"], "gpt_neo/modeling_gpt_neo.py:GPTNeoForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "def", "device", "dtype", "else", "float32", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_function", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "gpt_neo/modeling_gpt_neo.py:GPTNeoForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "sequence_length", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "gpt_neo/modeling_gpt_neo.py:GPTNeoForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view"], "gpt_neo/modeling_gpt_neo.py:GPTNeoForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "phi/modeling_phi.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "phi/modeling_phi.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "phi/modeling_phi.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "phi/modeling_phi.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "phi/modeling_phi.py:PhiAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "LayerNorm", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "cat", "class", "config", "contiguous", "cos", "def", "dense", "deprecate_kwarg", "dim", "dropout", "eager", "eager_attention_forward", "elementwise_affine", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_layernorm", "k_proj", "key_pass", "key_rot", "key_states", "kwargs", "layer_idx", "layer_norm_eps", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "partial_rotary_factor", "past_key_value", "past_key_values", "position_embeddings", "q_layernorm", "q_proj", "qk_layernorm", "query_pass", "query_rot", "query_states", "reshape", "return", "rotary_ndims", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "phi/modeling_phi.py:PhiMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "phi/modeling_phi.py:PhiDecoderLayer": ["Cache", "Dropout", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_outputs", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "feed_forward_hidden_states", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_norm_eps", "mlp", "new_name", "nn", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "resid_dropout", "resid_pdrop", "residual", "return", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "phi/modeling_phi.py:PhiRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "phi/modeling_phi.py:PhiPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "phi/modeling_phi.py:PhiModel": ["BaseModelOutputWithPast", "Cache", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "checkpointing", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embd_pdrop", "embed_dropout", "embed_tokens", "eps", "exactly", "final_layernorm", "for", "forward", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_eps", "layer_outputs", "layers", "logger", "must", "nn", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "shape", "specify", "super", "torch", "training", "unsqueeze", "use_cache", "vocab_size", "warning_once", "with"], "phi/modeling_phi.py:PhiForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "phi/modeling_phi.py:PhiForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "phi/modeling_phi.py:PhiForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "vit_msn/modeling_vit_msn.py:ViTMSNEmbeddings": ["BoolTensor", "Dropout", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "else", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_length", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_mask_token", "view", "width", "zeros"], "vit_msn/modeling_vit_msn.py:ViTMSNPatchEmbeddings": ["Conv2d", "Expected", "False", "Input", "Iterable", "Make", "ModelConfig", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "bool", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "flatten", "forward", "got", "height", "hidden_size", "if", "image", "image_size", "in", "interpolate_pos_encoding", "isinstance", "kernel_size", "match", "model", "nn", "not", "num_channels", "num_patches", "of", "one", "or", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "torch", "transpose", "values", "width", "with"], "vit_msn/modeling_vit_msn.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "vit_msn/modeling_vit_msn.py:ViTMSNSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "vit_msn/modeling_vit_msn.py:ViTMSNSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "vit_msn/modeling_vit_msn.py:ViTMSNAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "vit_msn/modeling_vit_msn.py:ViTMSNIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "vit_msn/modeling_vit_msn.py:ViTMSNOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "vit_msn/modeling_vit_msn.py:ViTMSNLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "vit_msn/modeling_vit_msn.py:ViTMSNEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "vit_msn/modeling_vit_msn.py:ViTMSNPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSdpaAttention", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "vit", "weight", "zero_"], "vit_msn/modeling_vit_msn.py:ViTMSNModel": ["BaseModelOutput", "BoolTensor", "False", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "bool", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dict", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "int", "interpolate_pos_encoding", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "num_hidden_layers", "patch_embeddings", "pixel_values", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch", "use_mask_token"], "vit_msn/modeling_vit_msn.py:ViTMSNForImageClassification": ["BaseModelOutput", "Identity", "ImageClassifierOutput", "Linear", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch", "vit"], "xglm/modeling_xglm.py:XGLMScaledWordEmbedding": ["Embedding", "Model", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "xglm/modeling_xglm.py:XGLMSinusoidalPositionalEmbedding": ["False", "Model", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "cos", "def", "detach", "device", "dim", "dtype", "emb", "emb_weights", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "index_select", "int", "int64", "is", "log", "make_weights", "math", "max_pos", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "unsqueeze", "view", "weights", "zeros"], "xglm/modeling_xglm.py:XGLMAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "class", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "device", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "finfo", "float", "float16", "float32", "for", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "mask", "max", "min", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "should", "single", "size", "softmax", "src_len", "super", "tensor", "tgt_len", "to", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view", "weights"], "xglm/modeling_xglm.py:XGLMDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "add_cross_attention", "attention_dropout", "attention_heads", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "ffn_dim", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "xglm/modeling_xglm.py:XGLMPreTrainedModel": ["Embedding", "Linear", "Model", "ModelConfig", "ModelDecoderLayer", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "if", "init_std", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "xglm/modeling_xglm.py:XGLMModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "Model", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "auto_docstring", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_layer", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "float", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "long", "mask_name", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "num_layers", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "warning_once", "will", "with", "zip"], "xglm/modeling_xglm.py:XGLMForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "logits", "loss", "loss_function", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "swin2sr/modeling_swin2sr.py:Swin2SREncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "torch", "tuple"], "swin2sr/modeling_swin2sr.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "height", "input_feature", "num_channels", "permute", "return", "shape", "view", "width"], "swin2sr/modeling_swin2sr.py:window_reverse": ["Model_reverse", "Model_size", "Models", "contiguous", "def", "height", "num_channels", "permute", "return", "shape", "view", "width"], "swin2sr/modeling_swin2sr.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "swin2sr/modeling_swin2sr.py:Swin2SRDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "swin2sr/modeling_swin2sr.py:Swin2SREmbeddings": ["Dropout", "FloatTensor", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "class", "config", "def", "dropout", "else", "embed_dim", "embeddings", "forward", "hidden_dropout_prob", "if", "is", "nn", "not", "num_patches", "output_dimensions", "patch_embeddings", "pixel_values", "position_embeddings", "return", "self", "super", "torch", "tuple", "use_absolute_embeddings", "window_size", "zeros"], "swin2sr/modeling_swin2sr.py:Swin2SRPatchEmbeddings": ["Conv2d", "FloatTensor", "Iterable", "LayerNorm", "ModelPatchEmbeddings", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "abc", "class", "collections", "config", "def", "else", "embed_dim", "embeddings", "flatten", "forward", "height", "if", "image_size", "int", "is", "isinstance", "kernel_size", "layernorm", "nn", "normalize_patches", "not", "num_channels", "num_patches", "output_dimensions", "patch_size", "patches_resolution", "projection", "return", "self", "shape", "stride", "super", "torch", "transpose", "tuple", "width"], "swin2sr/modeling_swin2sr.py:Swin2SRPatchUnEmbeddings": ["ModelPatchUnEmbeddings", "Module", "__init__", "batch_size", "class", "config", "def", "embed_dim", "embeddings", "forward", "height_width", "nn", "num_channels", "r", "return", "self", "shape", "super", "transpose", "view", "x_size"], "swin2sr/modeling_swin2sr.py:Swin2SRPatchMerging": ["False", "LayerNorm", "Linear", "ModelPatchMerging", "Module", "None", "Tensor", "__init__", "batch_size", "bias", "cat", "class", "def", "dim", "forward", "functional", "height", "if", "input_dimensions", "input_feature", "input_feature_0", "input_feature_1", "input_feature_2", "input_feature_3", "input_resolution", "int", "maybe_pad", "nn", "norm", "norm_layer", "num_channels", "or", "pad", "pad_values", "reduction", "return", "self", "shape", "should_pad", "super", "torch", "tuple", "view", "width"], "swin2sr/modeling_swin2sr.py:Swin2SRSelfAttention": ["Dropout", "False", "FloatTensor", "Iterable", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Parameter", "ReLU", "Sequential", "Tensor", "The", "True", "ValueError", "__init__", "a", "abc", "abs", "all_head_size", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "clamp", "class", "collections", "config", "context_layer", "contiguous", "continuous_position_bias_mlp", "coords", "coords_flatten", "coords_h", "coords_w", "def", "dim", "dropout", "dtype", "elif", "else", "exp", "f", "flatten", "float", "forward", "functional", "head_mask", "heads", "hidden", "hidden_states", "if", "ij", "indexing", "inplace", "int", "int64", "is", "isinstance", "key", "key_layer", "log", "log2", "logit_scale", "mask_shape", "math", "matmul", "max", "meshgrid", "multiple", "new_context_layer_shape", "next", "nn", "normalize", "not", "num_attention_heads", "num_channels", "num_heads", "number", "of", "ones", "output_attentions", "outputs", "parameters", "permute", "persistent", "pretrained_window_size", "qkv_bias", "query", "query_layer", "raise", "register_buffer", "relative_coords", "relative_coords_h", "relative_coords_table", "relative_coords_w", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "return", "self", "shape", "sigmoid", "sign", "size", "softmax", "stack", "sum", "super", "the", "to", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view", "window_size"], "swin2sr/modeling_swin2sr.py:Swin2SRSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "swin2sr/modeling_swin2sr.py:Swin2SRAttention": ["False", "FloatTensor", "Iterable", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "abc", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "collections", "config", "def", "dense", "dim", "else", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "isinstance", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "pretrained_window_size", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "swin2sr/modeling_swin2sr.py:Swin2SRIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "swin2sr/modeling_swin2sr.py:Swin2SROutput": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "swin2sr/modeling_swin2sr.py:Swin2SRLayer": ["False", "FloatTensor", "Identity", "Iterable", "LayerNorm", "ModelAttention", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "_compute_window_shift", "abc", "attention", "attention_output", "attention_outputs", "attention_windows", "attn_mask", "batch_size", "bool", "channels", "class", "collections", "config", "contiguous", "count", "def", "device", "dim", "dims", "drop_path", "drop_path_rate", "dtype", "else", "eps", "for", "forward", "functional", "get_attn_mask", "head_mask", "height", "height_pad", "height_slice", "height_slices", "hidden_states", "hidden_states_windows", "if", "img_mask", "in", "input_dimensions", "input_resolution", "int", "intermediate", "is", "isinstance", "layer_norm_eps", "layer_output", "layer_outputs", "layernorm_after", "layernorm_before", "mask_windows", "masked_fill", "maybe_pad", "nn", "not", "num_heads", "or", "output", "output_attentions", "pad", "pad_bottom", "pad_right", "pad_values", "pretrained_window_size", "r", "return", "roll", "s", "self", "shape", "shift_size", "shifted_hidden_states", "shifted_windows", "shifts", "shortcut", "size", "slice", "super", "target_shift_size", "target_window_size", "to", "torch", "tuple", "unsqueeze", "view", "w", "was_padded", "width", "width_pad", "width_slice", "width_slices", "window_partition", "window_reverse", "window_size", "zeros", "zip"], "swin2sr/modeling_swin2sr.py:Swin2SRStage": ["Conv2d", "False", "FloatTensor", "GradientCheckpointingLayer", "LeakyReLU", "ModelLayer", "ModelPatchEmbeddings", "ModelPatchUnEmbeddings", "ModelStage", "ModuleList", "None", "Optional", "Sequential", "Tensor", "True", "_", "__init__", "bool", "class", "config", "conv", "def", "depth", "dim", "drop_path", "elif", "else", "enumerate", "for", "forward", "head_mask", "height", "hidden_states", "i", "if", "in", "inplace", "input_dimensions", "input_resolution", "int", "is", "layer_head_mask", "layer_module", "layer_outputs", "layers", "negative_slope", "nn", "normalize_patches", "not", "num_heads", "output_attentions", "output_dimensions", "patch_embed", "patch_unembed", "pretrained_window_size", "range", "resi_connection", "residual", "return", "self", "shift_size", "stage_outputs", "super", "torch", "tuple", "width", "window_size"], "swin2sr/modeling_swin2sr.py:Swin2SREncoder": ["False", "FloatTensor", "ModelEncoder", "ModelEncoderOutput", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_input_dimensions", "all_self_attentions", "attentions", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "dim", "dpr", "drop_path", "drop_path_rate", "else", "embed_dim", "enumerate", "for", "forward", "gradient_checkpointing", "grid_size", "head_mask", "hidden_states", "i", "if", "in", "input_dimensions", "input_resolution", "int", "is", "item", "last_hidden_state", "layer_head_mask", "layer_outputs", "len", "linspace", "nn", "not", "num_heads", "num_stages", "output_attentions", "output_dimensions", "output_hidden_states", "pretrained_window_size", "range", "return", "return_dict", "self", "stage_idx", "stage_module", "stages", "sum", "super", "torch", "tuple", "v", "x"], "swin2sr/modeling_swin2sr.py:Swin2SRPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "module", "nn", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "torch", "trunc_normal_", "weight", "zero_"], "swin2sr/modeling_swin2sr.py:Swin2SRModel": ["BaseModelOutput", "Conv2d", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchUnEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Union", "_", "__init__", "_prune_heads", "and", "attention", "attentions", "auto_docstring", "bool", "class", "config", "conv_after_body", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "first_convolution", "for", "forward", "functional", "get_head_mask", "get_input_embeddings", "grid_size", "head_mask", "heads", "heads_to_prune", "height", "hidden_states", "if", "img_range", "in", "input_dimensions", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "len", "mean", "modulo_pad_height", "modulo_pad_width", "nn", "not", "num_channels", "num_channels_out", "output", "output_attentions", "output_hidden_states", "pad", "pad_and_normalize", "patch_embeddings", "patch_unembed", "patches_resolution", "persistent", "pixel_values", "post_init", "prune_heads", "reflect", "register_buffer", "return", "return_dict", "self", "sequence_output", "shape", "size", "super", "tensor", "torch", "tuple", "type_as", "use_return_dict", "view", "width", "window_size", "zeros"], "swin2sr/modeling_swin2sr.py:Upsample": ["Conv2d", "Model", "Module", "PixelShuffle", "Scale", "Supported", "ValueError", "__getattr__", "__init__", "add_module", "and", "class", "convolution", "convolution_", "def", "elif", "else", "f", "for", "forward", "hidden_state", "i", "if", "in", "int", "is", "log", "math", "n", "nn", "not", "num_features", "pixelshuffle", "pixelshuffle_", "raise", "range", "return", "scale", "scales", "self", "super", "supported"], "swin2sr/modeling_swin2sr.py:UpsampleOneStep": ["Conv2d", "ModelOneStep", "Module", "PixelShuffle", "__init__", "class", "conv", "def", "forward", "in_channels", "nn", "out_channels", "pixel_shuffle", "return", "scale", "self", "super", "x"], "swin2sr/modeling_swin2sr.py:PixelShuffleUpsampler": ["Conv2d", "LeakyReLU", "ModelShuffleUpsampler", "Module", "True", "Upsample", "__init__", "activation", "class", "config", "conv_before_upsample", "def", "embed_dim", "final_convolution", "forward", "inplace", "nn", "num_channels_out", "num_features", "return", "self", "sequence_output", "super", "upsample", "upscale", "x"], "swin2sr/modeling_swin2sr.py:NearestConvUpsampler": ["Conv2d", "LeakyReLU", "Model", "ModelConvUpsampler", "Module", "The", "True", "ValueError", "__init__", "activation", "an", "at", "class", "config", "conv", "conv_before_upsample", "conv_hr", "conv_up1", "conv_up2", "def", "embed_dim", "factor", "final_convolution", "forward", "functional", "if", "inplace", "interpolate", "lrelu", "mode", "moment", "negative_slope", "nn", "num_channels_out", "num_features", "of", "only", "raise", "reconstruction", "return", "scale_factor", "self", "sequence_output", "super", "supports", "the", "torch", "upsampler", "upscale"], "swin2sr/modeling_swin2sr.py:PixelShuffleAuxUpsampler": ["Conv2d", "LeakyReLU", "ModelShuffleAuxUpsampler", "Module", "Sequential", "True", "Upsample", "__init__", "activation", "aux", "bicubic", "class", "config", "conv_after_aux", "conv_aux", "conv_before_upsample", "conv_bicubic", "def", "embed_dim", "final_convolution", "forward", "height", "inplace", "nn", "num_channels", "num_channels_out", "num_features", "reconstruction", "return", "self", "sequence_output", "super", "upsample", "upscale", "width"], "swin2sr/modeling_swin2sr.py:Swin2SRForImageSuperResolution": ["Conv2d", "False", "FloatTensor", "ImageSuperResolutionOutput", "LongTensor", "Model", "ModelForImageSuperResolution", "ModelModel", "ModelPreTrainedModel", "NearestConvUpsampler", "None", "NotImplementedError", "Optional", "PixelShuffleAuxUpsampler", "PixelShuffleUpsampler", "Training", "Union", "UpsampleOneStep", "__init__", "align_corners", "at", "attentions", "auto_docstring", "aux", "bicubic", "bool", "class", "config", "conv", "def", "elif", "else", "embed_dim", "final_convolution", "forward", "functional", "head_mask", "height", "hidden_states", "if", "img_range", "in", "interpolate", "is", "labels", "loss", "mean", "mode", "moment", "nearest", "nn", "not", "num_channels_out", "num_features", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pixelshuffle", "pixelshuffle_aux", "pixelshuffledirect", "post_init", "r", "raise", "reconstruction", "return", "return_dict", "self", "sequence_output", "shape", "size", "super", "supported", "the", "torch", "tuple", "upsample", "upsampler", "upscale", "use_return_dict", "width"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "bool", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "nn", "return", "self", "super", "up_proj"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VisionPatchEmbed": ["Conv3d", "False", "Model_5_VisionPatchEmbed", "Module", "None", "Tensor", "__init__", "bias", "class", "def", "dtype", "embed_dim", "forward", "hidden_states", "in_channels", "int", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VisionRotaryEmbedding": ["False", "Model_5_VisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLPatchMerger": ["GELU", "Linear", "ModelPatchMerger", "ModelRMSNorm", "Module", "None", "Sequential", "Tensor", "__init__", "class", "context_dim", "def", "dim", "eps", "forward", "hidden_size", "int", "ln_q", "mlp", "nn", "return", "self", "spatial_merge_size", "super", "torch", "view", "x"], "qwen2_5_vl/modeling_qwen2_5_vl.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen2_5_vl/modeling_qwen2_5_vl.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "qwen2_5_vl/modeling_qwen2_5_vl.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen2_5_vl/modeling_qwen2_5_vl.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLVisionBlock": ["GradientCheckpointingLayer", "ModelMLP", "ModelRMSNorm", "ModelVisionAttention", "ModelVisionBlock", "None", "Optional", "Tensor", "True", "__init__", "attn", "attn_implementation", "bias", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "norm1", "norm2", "position_embeddings", "return", "rotary_pos_emb", "sdpa", "self", "str", "super", "torch", "tuple"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLPreTrainedModel": ["ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VisionTransformerPretrainedModel": ["F", "False", "ModelPatchMerger", "ModelPreTrainedModel", "ModelVisionBlock", "ModelVisionConfig", "Model_5_VisionPatchEmbed", "Model_5_VisionRotaryEmbedding", "Model_5_VisionTransformerPretrainedModel", "ModuleList", "None", "Tensor", "_", "__init__", "_no_split_modules", "append", "arange", "argsort", "blk", "blocks", "cat", "class", "config", "constant", "context_dim", "cos", "cu_seqlens", "cu_seqlens_now", "cu_seqlens_tmp", "cu_window_seqlens", "cumsum", "def", "depth", "device", "dim", "dtype", "else", "emb", "embed_dim", "enumerate", "expand", "extend", "flatten", "for", "forward", "fullatt_block_indexes", "get_window_index", "gradient_checkpointing", "grid_h", "grid_t", "grid_thw", "grid_w", "h", "head_dim", "hidden_size", "hidden_states", "hpos_ids", "if", "in", "in_channels", "index", "index_new", "index_padded", "inputs", "int32", "is_tracing", "item", "jit", "kwargs", "layer_num", "list", "llm_grid_h", "llm_grid_w", "max", "max_grid_size", "merger", "nn", "num_heads", "num_windows_h", "num_windows_w", "out_hidden_size", "pad", "pad_h", "pad_w", "patch_embed", "patch_size", "permute", "pos_ids", "position_embeddings", "range", "repeat", "repeat_interleave", "reshape", "return", "reverse_indices", "rot_pos_emb", "rotary_pos_emb", "rotary_pos_emb_full", "self", "seq_len", "seqlens", "sin", "size", "spatial_merge_size", "spatial_merge_unit", "stack", "sum", "super", "t", "temporal_patch_size", "tensor", "tolist", "torch", "unique_consecutive", "unsqueeze", "value", "vit_merger_window_size", "w", "window_index", "window_index_id", "window_size", "wpos_ids"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLModelOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLRotaryEmbedding": ["False", "ModelRotaryEmbedding", "ModelTextConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen2_5_vl/modeling_qwen2_5_vl.py:apply_multimodal_rotary_pos_emb": ["Model_multimodal_rotary_pos_emb", "cat", "cos", "def", "dim", "enumerate", "for", "i", "in", "k", "k_embed", "m", "mrope_section", "q", "q_embed", "return", "rotate_half", "sin", "split", "torch", "unsqueeze", "unsqueeze_dim"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelRotaryEmbedding", "ModelTextConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "apply_multimodal_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "errors", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "logger", "make", "mrope_section", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "rope_scaling", "rotary_emb", "scaling", "self", "sin", "size", "sliding_attention", "sliding_window", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLDecoderLayer": ["Attention", "Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelTextConfig", "None", "Optional", "Sliding", "Tensor", "Unpack", "Window", "__init__", "_attn_implementation", "and", "attention_mask", "attention_type", "be", "bool", "but", "cache_position", "class", "config", "def", "deprecate_kwarg", "enabled", "encountered", "eps", "f", "flash_attention_2", "for", "forward", "hidden_size", "hidden_states", "if", "implemented", "input_layernorm", "int", "is", "kwargs", "layer_idx", "layer_types", "logger", "may", "mlp", "new_name", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "results", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "unexpected", "use_cache", "use_sliding_window", "version", "warning_once"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTextConfig", "ModelTextModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "checkpointing", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "elif", "else", "embed_tokens", "eps", "exactly", "expand", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "isinstance", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layer_types", "layers", "logger", "mask_kwargs", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "return_dict", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "text_position_ids", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warning_once", "with"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLModel": ["Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionBlock", "Model_5_VisionTransformerPretrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "_", "__init__", "_checkpoint_conversion_mapping", "_from_config", "_no_split_modules", "accepts_loss_kwargs", "all", "and", "append", "arange", "argwhere", "as_tensor", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "bool", "cache_position", "cat", "class", "config", "cumsum", "decoder", "def", "delta", "device", "dim", "do", "dtype", "ed", "ed_image", "ed_video", "else", "enumerate", "expand", "expand_as", "expanded_range", "f", "features", "flatten", "for", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_seq_length", "get_video_features", "grid_thw", "h", "h_index", "hidden_states", "i", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_index", "image_mask", "image_nums", "image_token_id", "in", "index", "input_ids", "input_tokens", "inputs_embeds", "is", "is_torchdynamo_compiling", "item", "keepdim", "kwargs", "language_model", "last_hidden_state", "len", "list", "llm_grid_h", "llm_grid_t", "llm_grid_w", "llm_pos_ids_list", "llm_positions", "long", "masked_fill_", "masked_scatter", "match", "max", "max_position_ids", "model", "mrope_position_deltas", "n_image_tokens", "n_video_tokens", "not", "numel", "ones", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prefill_compiled_stage", "prefill_noncompiled_stage", "prod", "r", "raise", "range", "range_tensor", "remain_images", "remain_videos", "repeat_interleave", "reshape", "return", "return_dict", "rope_deltas", "second_per_grid_t", "second_per_grid_ts", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "special_video_mask", "split", "split_sizes", "squeeze", "st", "st_idx", "stack", "sum", "super", "t", "t_index", "tensor", "text_config", "text_len", "time_tensor", "time_tensor_long", "to", "to_tuple", "tokens", "tokens_per_second", "tolist", "torch", "total_input_ids", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "value", "video", "video_embeds", "video_features", "video_grid_thw", "video_index", "video_mask", "video_nums", "video_token_id", "view", "vision_config", "vision_start_indices", "vision_start_token_id", "vision_tokens", "visual", "w", "w_index", "zeros"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen2_5_vl/modeling_qwen2_5_vl.py:Qwen2_5_VLForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "If", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_expand_dict_for_generation", "_expand_dict_for_generation_visual", "_expand_inputs_for_generation", "_get_image_nums_and_video_nums", "_repeat_interleave_samples", "_tied_weights_keys", "accepts_loss_kwargs", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "defined", "delta", "device", "dict", "dict_to_expand", "dim", "dims", "dtype", "elif", "else", "encoder_outputs", "expand", "expand_as", "expand_size", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "get_rope_index", "get_video_features", "hidden_size", "hidden_states", "if", "image_grid_thw", "image_mask", "image_nums", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "isinstance", "key", "kwargs", "labels", "language_model", "lengths", "list", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "make", "model", "model_inputs", "model_kwargs", "nn", "not", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "prod", "property", "r", "raise", "repeat", "repeat_args", "repeat_interleave", "repeat_times", "result", "return", "return_dict", "roll", "rope_deltas", "sample", "samples", "second_per_grid_ts", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "shifts", "slice", "slice_indices", "split", "str", "sum", "super", "sure", "tensor", "text_config", "text_positions", "that", "torch", "tuple", "use_cache", "value", "video_grid_thw", "video_mask", "video_nums", "video_token_id", "view", "vision_first_mask", "vision_positions", "vision_start_mask", "vision_start_token_id", "visual", "visual_keys", "vocab_size", "weight", "x"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "not", "return", "self", "super", "up_proj", "use_bias", "x"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "ernie4_5_moe/modeling_ernie4_5_moe.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "ernie4_5_moe/modeling_ernie4_5_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dim", "dtype", "float", "k", "k_embed", "original_dtype", "position_ids", "q", "q_embed", "repeat_interleave", "return", "rotate_half", "shape", "sin", "to", "unsqueeze", "unsqueeze_dim"], "ernie4_5_moe/modeling_ernie4_5_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "ernie4_5_moe/modeling_ernie4_5_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "use_bias", "v_proj", "value_states", "version", "view"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeStatics": ["False", "ModelStatics", "Module", "Parameter", "__init__", "class", "config", "def", "dtype", "e_score_correction_bias", "float32", "forward", "hidden_states", "moe_num_experts", "nn", "num_experts", "num_experts_groups", "requires_grad", "return", "self", "squeeze", "super", "torch", "zeros"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeSparseMoeBlock": ["F", "False", "Linear", "ModelMLP", "ModelSparseMoeBlock", "ModelStatics", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "and", "autocast", "batch_size", "bias", "clamp", "class", "config", "cpu", "current_hidden_states", "current_state", "def", "device", "device_type", "dim", "dtype", "else", "enabled", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "float32", "for", "forward", "functional", "gate", "gather", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index", "index_add_", "is", "isinstance", "keepdim", "min", "moe_intermediate_size", "moe_k", "moe_norm_min", "moe_num_experts", "moe_num_shared_experts", "moe_statics", "mps", "nn", "nonzero", "norm_min", "not", "num_classes", "num_experts", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "shared_experts", "shared_output", "softmax", "squeeze", "str", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "tuple", "type", "view", "where", "with", "zeros"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeDecoderLayer": ["Cache", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "and", "attention_mask", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "isinstance", "kwargs", "layer_idx", "mlp", "moe_layer_end_index", "moe_layer_interval", "moe_layer_start_index", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoePreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSparseMoeBlock", "ModelStatics", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_keep_in_fp32_modules_strict", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "e_score_correction_bias", "gate", "hidden_states", "if", "index", "isinstance", "model", "module", "moe_statics", "mtp", "past_key_values", "router_logits", "self", "super", "supports_gradient_checkpointing", "zero_"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeModel": ["Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "ernie4_5_moe/modeling_ernie4_5_moe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "ernie4_5_moe/modeling_ernie4_5_moe.py:Ernie4_5_MoeForCausalLM": ["Cache", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "moe_k", "moe_num_experts", "nn", "not", "num_experts", "num_experts_per_tok", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_bias", "use_cache", "vocab_size", "weight"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoContrastiveEmbedding": ["BoolTensor", "FloatTensor", "Model", "Module", "None", "Parameter", "__init__", "bias", "class", "config", "def", "device", "float", "forward", "full", "inf", "masked_fill_", "math", "max_text_len", "new_res", "nn", "res", "return", "self", "shape", "sqrt", "super", "tensor", "text_hidden_state", "text_token_mask", "torch", "transpose", "vision_hidden_state"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MultiScaleDeformableAttention": ["False", "ModelScaleDeformableAttention", "Module", "Tensor", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "class", "contiguous", "def", "dim", "enumerate", "flatten", "for", "forward", "functional", "grid_sample", "height", "hidden_dim", "im2col_step", "in", "int", "level_id", "level_start_index", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "self", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "value_spatial_shapes_list", "view", "width", "zeros"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoLearnedPositionEmbedding": ["Embedding", "Model", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "config", "d_model", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoMultiscaleDeformableAttention": ["CUDA", "F", "False", "Last", "Linear", "Make", "Model", "ModelConfig", "Module", "MultiScaleDeformableAttention", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "attn", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "config", "d", "d_model", "def", "dim", "dim_per_head", "dimension", "disable_custom_kernels", "divisible", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "forward", "got", "head", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "make", "masked_fill", "more", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_coordinates", "num_feature_levels", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "value", "value_proj", "view", "warn", "warnings", "which", "with", "with_pos_embed"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoBiMultiHeadAttention": ["Attention", "BoolTensor", "F", "FloatTensor", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_reshape", "and", "attn_weights", "attn_weights_transposed", "batch_size", "be", "bmm", "but", "by", "clamp", "class", "config", "contiguous", "d_model", "def", "dim", "divisible", "dropout", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "f", "flatten", "float", "forward", "fusion_dropout", "got", "head_dim", "if", "inf", "int", "is", "keepdim", "masked_fill_", "max", "min", "must", "nn", "not", "num_heads", "of", "out_text_proj", "out_vision_proj", "p", "proj_shape", "raise", "repeat", "reshape", "return", "scale", "self", "seq_len", "should", "size", "softmax", "src_len", "super", "tensor", "text_attention_mask", "text_attn_output", "text_attn_probs", "text_attn_weights", "text_dim", "text_features", "text_key_states", "text_proj", "text_value_states", "tgt_len", "torch", "training", "transpose", "tuple", "values_text_proj", "values_vision_proj", "view", "vision_attention_mask", "vision_attn_output", "vision_attn_probs", "vision_attn_weights", "vision_dim", "vision_features", "vision_proj", "vision_query_states", "vision_value_states", "weights"], "mm_grounding_dino/modeling_mm_grounding_dino.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoDropPath": ["Model", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoFusionLayer": ["BoolTensor", "FloatTensor", "Identity", "LayerNorm", "Model", "ModelBiMultiHeadAttention", "ModelDropPath", "Module", "None", "Optional", "Parameter", "True", "__init__", "attention_mask_text", "attention_mask_vision", "attn", "class", "config", "d_model", "def", "delta_t", "delta_v", "drop_path", "else", "forward", "fusion_droppath", "if", "init_values", "layer_norm_eps", "layer_norm_text", "layer_norm_vision", "nn", "ones", "requires_grad", "return", "self", "super", "text_attention_mask", "text_attn", "text_features", "text_param", "torch", "tuple", "vision_attention_mask", "vision_attn", "vision_features", "vision_param"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "False", "GroupNorm", "LayerNorm", "Linear", "Model", "ModelBiMultiHeadAttention", "ModelConfig", "ModelContrastiveEmbedding", "ModelDecoder", "ModelFusionLayer", "ModelLearnedPositionEmbedding", "ModelMLPPredictionHead", "ModelMultiscaleDeformableAttention", "None", "Parameter", "PreTrainedModel", "True", "_init_weights", "_set_gradient_checkpointing", "abs", "and", "arange", "attention_weights", "base_model_prefix", "bias", "class", "column_embeddings", "config", "constant_", "cos", "data", "def", "default_dtype", "dtype", "elif", "fill_", "for", "gain", "get_default_dtype", "gradient_checkpointing", "grid_init", "hasattr", "i", "if", "in", "init", "init_std", "int64", "is", "isinstance", "keepdim", "layers", "level_embed", "log", "main_input_name", "math", "max", "mean", "model", "module", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "out_text_proj", "out_vision_proj", "output_proj", "padding_idx", "pi", "pixel_values", "range", "reference_points", "repeat", "row_embeddings", "sampling_offsets", "self", "sin", "stack", "std", "text_param", "text_proj", "thetas", "to", "torch", "two_stage", "uniform_", "value", "value_proj", "values_text_proj", "values_vision_proj", "view", "vision_param", "vision_proj", "weight", "with", "xavier_uniform_", "zero_"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoFrozenBatchNorm2d": ["Model", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "mm_grounding_dino/modeling_mm_grounding_dino.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoConvEncoder": ["Either", "False", "Model", "Module", "None", "Tensor", "True", "ValueError", "__init__", "and", "append", "backbone", "backbone_config", "backbone_kwargs", "backbone_model_type", "be", "bool", "channels", "class", "config", "create_model", "def", "elif", "else", "feature_info", "feature_map", "feature_maps", "features", "features_only", "float", "for", "forward", "functional", "if", "in", "intermediate_channel_sizes", "interpolate", "is", "layer2", "layer3", "layer4", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "or", "out", "parameter", "pixel_mask", "pixel_values", "pretrained", "provided", "raise", "replace_batch_norm", "requires_backends", "requires_grad_", "resnet", "return", "self", "shape", "should", "size", "stage", "super", "the", "timm", "to", "torch", "use_pretrained_backbone", "use_timm_backbone", "with"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoConvModel": ["Model", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoEncoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "last_hidden_state_text", "last_hidden_state_vision", "r", "text_hidden_states", "torch", "tuple", "vision_hidden_states"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoMultiheadAttention": ["Dropout", "False", "FloatTensor", "Linear", "Model", "Module", "None", "Optional", "Tensor", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_mask", "attention_probs", "attention_scores", "batch_size", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "functional", "hasattr", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "keys", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "out_proj", "output_attentions", "outputs", "permute", "queries", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "values", "view"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoTextEnhancerLayer": ["ACT2FN", "BoolTensor", "FloatTensor", "LayerNorm", "Linear", "Model", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "activation", "activation_function", "and", "attention_mask", "attention_masks", "attention_output", "attention_weights", "class", "config", "d_model", "def", "dim", "dropout", "dtype", "else", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "finfo", "forward", "functional", "hidden_state", "hidden_states", "if", "is", "keys", "layer_norm_after", "layer_norm_before", "layer_norm_eps", "min", "nn", "num_attention_heads", "num_heads", "output_attentions", "p", "position_embeddings", "queries", "repeat", "residual", "return", "self", "self_attn", "shape", "super", "text_enhancer_dropout", "to", "torch", "training", "tuple", "values", "with_pos_embed"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoDeformableLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_attention_mask", "encoder_ffn_dim", "encoder_hidden_states", "encoder_n_points", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "layer_norm_eps", "level_start_index", "max", "min", "n_points", "nn", "num_heads", "or", "output_attentions", "p", "position_embeddings", "reference_points", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training"], "mm_grounding_dino/modeling_mm_grounding_dino.py:get_sine_pos_embed": ["Model_sine_pos_embed", "Tensor", "True", "arange", "bool", "cat", "cos", "def", "device", "dim", "dim_t", "div", "dtype", "exchange_xy", "flatten", "float32", "floor", "for", "if", "in", "int", "math", "num_pos_feats", "pi", "pos_tensor", "position_embeddings", "return", "rounding_mode", "scale", "shape", "sin", "sin_x", "sine_func", "split", "stack", "temperature", "torch", "x"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoEncoderLayer": ["False", "Model", "ModelDeformableLayer", "ModelFusionLayer", "ModelTextEnhancerLayer", "Module", "None", "Optional", "Tensor", "_", "__init__", "and", "arange", "attention_mask", "attention_mask_text", "attention_mask_vision", "attention_masks", "batch_size", "class", "config", "d_model", "def", "deformable_layer", "device", "else", "exchange_xy", "float", "forward", "fusion_layer", "get_sine_pos_embed", "get_text_position_embeddings", "hidden_states", "if", "int", "is", "key_padding_mask", "level_start_index", "list", "nn", "not", "num_pos_feats", "position_embeddings", "reference_points", "repeat", "return", "self", "seq_length", "shape", "spatial_shapes", "spatial_shapes_list", "super", "text_attention_mask", "text_enhanced_attn", "text_enhancer_layer", "text_features", "text_fused_attn", "text_position_embedding", "text_position_ids", "text_self_attention_masks", "torch", "tuple", "unsqueeze", "vision_deformable_attn", "vision_features", "vision_fused_attn", "vision_position_embedding"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoEncoder": ["Model", "ModelConfig", "ModelLayer", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_attn_deformable", "all_attn_enhanced_text", "all_attn_fused_text", "all_attn_fused_vision", "all_attns", "append", "attentions", "cat", "class", "config", "def", "device", "dropout", "dtype", "else", "enc_outputs", "encoder_layer", "encoder_layers", "encoder_text_states", "encoder_vision_states", "enumerate", "float32", "for", "forward", "get_reference_points", "height", "i", "if", "ij", "in", "indexing", "int", "is", "key_padding_mask", "last_hidden_state_text", "last_hidden_state_vision", "layers", "level", "level_start_index", "linspace", "list", "meshgrid", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "r", "range", "ref", "ref_x", "ref_y", "reference_points", "reference_points_list", "reshape", "return", "return_dict", "self", "spatial_shapes", "spatial_shapes_list", "stack", "staticmethod", "super", "text_attention_mask", "text_features", "text_hidden_states", "text_position_embedding", "text_position_ids", "text_self_attention_masks", "torch", "tuple", "use_return_dict", "v", "valid_ratios", "vision_attention_mask", "vision_features", "vision_hidden_states", "vision_position_embedding", "width"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoDecoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoDecoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "decoder_n_points", "def", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_attn_text", "encoder_attn_text_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "keys", "layer_norm_eps", "level_start_index", "n_points", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "queries", "reference_points", "residual", "return", "second_residual", "self", "self_attn", "self_attn_layer_norm", "self_attn_mask", "self_attn_weights", "spatial_shapes", "spatial_shapes_list", "super", "tensor", "text_cross_attn_weights", "text_encoder_attention_mask", "text_encoder_hidden_states", "third_residual", "torch", "training", "values", "vision_encoder_attention_mask", "vision_encoder_hidden_states", "with_pos_embed"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoDecoder": ["False", "Last", "LayerNorm", "Model", "ModelConfig", "ModelLayer", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "ValueError", "_", "__init__", "all_attns", "all_cross_attns_text", "all_cross_attns_vision", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "be", "but", "cat", "checkpoint", "class", "class_embed", "config", "create_custom_forward", "custom_forward", "d_model", "decoder_attention_heads", "decoder_layer", "decoder_layers", "def", "detach", "dim", "dropout", "dtype", "elif", "else", "enumerate", "eps", "f", "finfo", "for", "forward", "get_sine_pos_embed", "got", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs", "inputs_embeds", "intermediate", "intermediate_hidden_states", "intermediate_reference_points", "is", "last_hidden_state", "layer_norm", "layer_norm_eps", "layer_outputs", "layers", "level_start_index", "logit", "min", "module", "must", "new_reference_points", "nn", "not", "num_coordinates", "num_pos_feats", "num_queries", "of", "or", "output_attentions", "output_hidden_states", "position_embeddings", "post_init", "query_dim", "query_pos", "query_scale", "r", "raise", "range", "reference_points", "reference_points_head", "reference_points_input", "repeat", "return", "return_dict", "self", "self_attn_mask", "shape", "sigmoid", "spatial_shapes", "spatial_shapes_list", "special", "stack", "super", "text_encoder_attention_mask", "text_encoder_hidden_states", "tmp", "to", "torch", "training", "tuple", "use_return_dict", "utils", "v", "valid_ratios", "vision_encoder_attention_mask", "vision_encoder_hidden_states"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "decoder_attentions", "decoder_hidden_states", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "init_reference_points", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoSinePositionEmbedding": ["Model", "Module", "None", "__init__", "arange", "cat", "class", "config", "cos", "cumsum", "d_model", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "eps", "flatten", "float32", "floor", "forward", "math", "nn", "permute", "pi", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "positional_embedding_temperature", "return", "rounding_mode", "scale", "self", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "mm_grounding_dino/modeling_mm_grounding_dino.py:build_position_encoding": ["ModelLearnedPositionEmbedding", "ModelSinePositionEmbedding", "Model_position_encoding", "Not", "ValueError", "config", "def", "elif", "else", "f", "if", "learned", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "mm_grounding_dino/modeling_mm_grounding_dino.py:generate_masks_with_special_tokens_and_transfer_map": ["LongTensor", "Model_masks_with_special_tokens_and_transfer_map", "SPECIAL_TOKENS", "Tensor", "True", "arange", "attention_mask", "batch_size", "bool", "col", "def", "device", "else", "eye", "for", "i", "idxs", "if", "in", "input_ids", "logical_or", "long", "nonzero", "num_token", "or", "position_ids", "previous_col", "range", "repeat", "return", "row", "shape", "special_token", "special_tokens_mask", "to", "torch", "tuple", "unsqueeze", "zeros"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoModel": ["AutoModel", "Conv2d", "Embedding", "False", "GroupNorm", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelContrastiveEmbedding", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelEncoderOutput", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "Sequential", "Tensor", "True", "_", "__init__", "_len_sources", "add_pooling_layer", "all", "and", "append", "as_tensor", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_size", "bool", "build_position_encoding", "cat", "class", "config", "conv_encoder", "cumsum", "current_position", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_outputs", "def", "delta_bbox", "detach", "device", "dim", "dtype", "elif", "else", "embedding_init_target", "enc_output", "enc_output_norm", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "encoder", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_output_bbox_embed", "encoder_output_class_embed", "encoder_outputs", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "enumerate", "expand", "feature_maps", "flatten", "float", "float32", "for", "forward", "freeze_backbone", "from_config", "functional", "gather", "generate_encoder_output_proposals", "generate_masks_with_special_tokens_and_transfer_map", "get_valid_ratio", "grid", "grid_x", "grid_y", "height", "hidden_dim", "hidden_size", "hidden_states", "i", "if", "ij", "in", "in_channels", "indexing", "inf", "init_reference_points", "input_dim", "input_ids", "input_proj_list", "input_proj_vision", "inputs_embeds", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_reference_points", "interpolate", "is", "isinstance", "keepdim", "kernel_size", "last_hidden_state", "last_hidden_state_text", "last_hidden_state_vision", "layer_norm_eps", "len", "level", "level_embed", "level_start_index", "linspace", "log", "long", "lvl_pos_embed", "lvl_pos_embed_flatten", "m", "mask", "mask_flatten", "mask_flatten_", "masked_fill", "masks", "max", "max_text_len", "meshgrid", "model", "name", "named_parameters", "new_zeros", "nn", "not", "num_backbone_outs", "num_channels", "num_feature_levels", "num_layers", "num_queries", "object_query", "object_query_embedding", "ones", "ones_like", "or", "output_attentions", "output_dim", "output_hidden_states", "output_proposals", "output_proposals_valid", "padding", "padding_mask", "param", "pixel_mask", "pixel_values", "pos_embed", "pos_l", "position_embedding", "position_embeddings", "position_embeddings_list", "position_ids", "post_init", "prod", "proposal", "proposals", "query_embeds", "query_position_embeddings", "r", "range", "reference_points", "repeat", "requires_grad_", "return", "return_dict", "scale", "self", "self_attn_mask", "shape", "sigmoid", "size", "source", "source_flatten", "spatial_shape", "spatial_shapes", "spatial_shapes_list", "stack", "stride", "sum", "super", "target", "text_attention_mask", "text_backbone", "text_config", "text_encoder_attention_mask", "text_encoder_hidden_states", "text_features", "text_hidden_states", "text_outputs", "text_position_embedding", "text_position_ids", "text_projection", "text_self_attention_masks", "text_token_mask", "to", "token_type_ids", "topk", "topk_coords_logits", "topk_logits", "topk_proposals", "torch", "transpose", "tuple", "tuple_outputs", "two_stage", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_height", "valid_ratio", "valid_ratio_height", "valid_ratio_width", "valid_ratios", "valid_width", "value", "view", "vision_attention_mask", "vision_encoder_attention_mask", "vision_encoder_hidden_states", "vision_features", "vision_hidden_states", "vision_position_embedding", "weight", "width", "width_height", "zeros_like", "zip"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoMLPPredictionHead": ["Linear", "Model", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoObjectDetectionOutput": ["FloatTensor", "LongTensor", "Model", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "decoder_attentions", "decoder_hidden_states", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "init_reference_points", "input_ids", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "mm_grounding_dino/modeling_mm_grounding_dino.py:build_label_maps": ["F", "FloatTensor", "LongTensor", "Model_label_maps", "None", "SPECIAL_TOKENS", "cumsum", "def", "delimiter_token_masks", "delimiter_tokens", "device", "dim", "for", "in", "input_ids", "int32", "isin", "label_group", "label_groups", "label_map", "label_maps", "logits", "max_seq_len", "num_labels", "pad", "repeat", "return", "shape", "tensor", "to", "torch", "tuple", "unique", "unique_labels", "unsqueeze", "value", "where"], "mm_grounding_dino/modeling_mm_grounding_dino.py:build_text_mask": ["Model_text_mask", "None", "attention_mask", "bool", "def", "device", "dtype", "logits", "return", "seq_len", "shape", "text_mask", "torch", "zeros_like"], "mm_grounding_dino/modeling_mm_grounding_dino.py:MMGroundingDinoForObjectDetection": ["BoolTensor", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelContrastiveEmbedding", "ModelEncoderOutput", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Union", "ValueError", "_", "__init__", "_tied_weights_keys", "append", "attention_mask", "auto_docstring", "auxiliary_outputs", "bbox_embed", "be", "bool", "build_label_maps", "build_text_mask", "but", "class", "class_embed", "config", "d", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_layers", "def", "delta_bbox", "device", "dict", "dict_outputs", "elif", "else", "enc_outputs_class", "enc_outputs_coord_logits", "enc_text_hidden_state", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_outputs", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "eps", "f", "for", "forward", "got", "hidden_dim", "hidden_states", "idx", "if", "in", "init_reference_points", "input_dim", "input_ids", "inter_references_points", "intermediate_hidden_states", "intermediate_reference_points", "is", "label_maps", "labels", "last_hidden_state", "level", "list", "logit", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_layers", "num_levels", "ones_like", "or", "out", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "outputs_class", "outputs_classes", "outputs_coord", "outputs_coord_logits", "outputs_coords", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "r", "raise", "range", "reference", "reference_coordinates", "return", "return_dict", "self", "shape", "should", "sigmoid", "special", "stack", "str", "super", "text_hidden_state", "text_mask", "text_token_mask", "token_type_ids", "torch", "tuple", "use_return_dict", "vision_hidden_state"], "umt5/modeling_umt5.py:UMT5LayerNorm": ["Model", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "umt5/modeling_umt5.py:UMT5DenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "umt5/modeling_umt5.py:UMT5DenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "umt5/modeling_umt5.py:UMT5LayerFF": ["DenseReluDense", "Dropout", "Model", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelLayerNorm", "Module", "__init__", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is_gated_act", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "umt5/modeling_umt5.py:UMT5Attention": ["Cache", "Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "Model", "Module", "None", "Optional", "Please", "Tensor", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "_shape", "a", "abs", "and", "arange", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "encoder_hidden_states", "errors", "f", "float", "forward", "full_like", "functional", "get", "get_seq_length", "has_relative_attention_bias", "hidden_states", "if", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "keys", "layer_head_mask", "layer_idx", "layers", "list", "log", "log_ratio", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "new_projection", "new_projection_shape", "nn", "not", "num_buckets", "num_heads", "o", "ones", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "projection", "provide", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "size", "softmax", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "unsqueeze", "update", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "umt5/modeling_umt5.py:UMT5LayerSelfAttention": ["Dropout", "Model", "ModelAttention", "ModelLayerNorm", "Module", "None", "Optional", "SelfAttention", "True", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "new_name", "nn", "normed_hidden_states", "outputs", "past_key_value", "past_key_values", "return", "self", "super", "version"], "umt5/modeling_umt5.py:UMT5LayerCrossAttention": ["Dropout", "EncDecAttention", "False", "Model", "ModelAttention", "ModelLayerNorm", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "encoder_hidden_states", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "new_name", "nn", "normed_hidden_states", "outputs", "past_key_value", "past_key_values", "return", "self", "super", "version"], "umt5/modeling_umt5.py:UMT5Block": ["False", "GradientCheckpointingLayer", "Model", "ModelLayerCrossAttention", "ModelLayerFF", "ModelLayerSelfAttention", "ModuleList", "None", "Optional", "__init__", "and", "any", "append", "attention_mask", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "encoder_attention_mask", "encoder_hidden_states", "finfo", "float16", "forward", "hidden_states", "if", "int", "is", "is_decoder", "isinf", "layer", "layer_head_mask", "layer_idx", "max", "max_dtype", "min", "new_name", "nn", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "return", "self", "self_attn_weights", "super", "torch", "use_cache", "version", "where"], "umt5/modeling_umt5.py:UMT5ClassificationHead": ["Dropout", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "classifier_dropout", "config", "d_model", "def", "dense", "dropout", "forward", "hidden_states", "nn", "num_labels", "out_proj", "p", "return", "self", "super", "tanh", "torch"], "umt5/modeling_umt5.py:UMT5PreTrainedModel": ["DUMMY_INPUTS", "DUMMY_MASK", "In", "Model", "ModelAttention", "ModelBlock", "ModelClassificationHead", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelEncoderModel", "ModelForConditionalGeneration", "ModelForQuestionAnswering", "ModelForTokenClassification", "ModelLayerNorm", "ModelModel", "None", "PreTrainedModel", "See", "True", "ValueError", "_can_compile_fullgraph", "_init_weights", "_keep_in_fp32_modules", "_no_split_modules", "_shift_right", "and", "base_model_prefix", "be", "bias", "cat", "class", "classifier", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "def", "defined", "dense", "dim", "docs", "dummy_inputs", "elif", "else", "factor", "fill_", "for", "full", "has", "has_relative_attention_bias", "hasattr", "if", "information", "initializer_factor", "input_ids", "input_mask", "is", "is_torch_fx_proxy", "isinstance", "it", "k", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "normal_", "not", "num_heads", "o", "out_proj", "pad_token_id", "property", "q", "qa_outputs", "raise", "relative_attention_bias", "return", "self", "set", "shape", "shared", "shifted_input_ids", "std", "supports_gradient_checkpointing", "tensor", "the", "tie_word_embeddings", "to", "torch", "transformer", "usually", "v", "weight", "wi", "wi_0", "wi_1", "wo", "zero_"], "umt5/modeling_umt5.py:UMT5Stack": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "Dropout", "DynamicCache", "EncoderDecoderCache", "False", "Model", "ModelBlock", "ModelLayerNorm", "ModelPreTrainedModel", "ModuleList", "None", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "a", "all_attentions", "all_cross_attentions", "all_hidden_states", "and", "any", "arange", "as", "at", "attention_mask", "attentions", "batch_size", "be", "block", "bool", "both", "cache_position", "can", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "d_model", "decoder", "decoder_", "def", "device", "diagonal", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_torchdynamo_compiling", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "logger", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "min", "min_dtype", "model", "new_embeddings", "nn", "not", "npu", "num_layers", "ones", "only", "or", "output_attentions", "output_hidden_states", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "post_init", "raise", "range", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "used", "using_compilable_cache", "v", "valid", "view", "warning_once", "with", "xpu"], "umt5/modeling_umt5.py:UMT5Model": ["BaseModelOutput", "BoolTensor", "Cache", "Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "True", "Union", "__init__", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "len", "model_type", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "prune_heads", "r", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "tie_encoder_decoder", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "umt5/modeling_umt5.py:UMT5ForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "True", "Union", "__init__", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "device", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "labels", "last_hidden_state", "len", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "model_dim", "model_type", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "r", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "shared", "size", "super", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "umt5/modeling_umt5.py:UMT5EncoderModel": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "SelfAttention", "Union", "__init__", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "attention_mask", "auto_docstring", "block", "bool", "class", "config", "copy", "d_model", "deepcopy", "def", "else", "embed_tokens", "encoder", "encoder_config", "encoder_outputs", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "is", "is_encoder_decoder", "items", "layer", "model_type", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "prune_heads", "r", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "umt5/modeling_umt5.py:UMT5ForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "EncDecAttention", "False", "FloatTensor", "If", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Please", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "_", "__class__", "__init__", "__name__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tied_weights_keys", "and", "are", "attention_mask", "auto_docstring", "batch_size", "be", "block", "bool", "cannot", "class", "classification_head", "config", "cross_attentions", "cross_attn_head_mask", "currently", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_size", "if", "input", "input_ids", "inputs_embeds", "int", "is", "labels", "layer", "len", "list", "logits", "long", "loss", "loss_fct", "model_parallel", "multi_label_classification", "must", "no", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pass", "passed", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "relative_attention_bias", "return", "return_dict", "same", "self", "sentence_representation", "sequence_output", "shape", "single_label_classification", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "transformer", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "umt5/modeling_umt5.py:UMT5ForTokenClassification": ["CrossEntropyLoss", "Dropout", "EncDecAttention", "Linear", "Model", "ModelConfig", "ModelEncoderModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "_keys_to_ignore_on_load_unexpected", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "block", "bool", "class", "classifier", "classifier_dropout", "config", "decoder", "def", "dropout", "else", "embed_tokens", "encoder", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "layer", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "r", "relative_attention_bias", "return", "return_dict", "self", "super", "torch", "transformer", "tuple", "use_return_dict", "view", "weight"], "umt5/modeling_umt5.py:UMT5ForQuestionAnswering": ["BaseModelOutput", "BoolTensor", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "If", "Linear", "LongTensor", "Model", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Please", "Seq2SeqQuestionAnsweringModelOutput", "Tensor", "True", "Union", "ValueError", "__init__", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "are", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cannot", "clamp", "class", "config", "contiguous", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "device", "dim", "either", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "last_hidden_state", "len", "logits", "loss", "loss_fct", "model_dim", "new_embeddings", "nn", "no", "not", "num_decoder_layers", "num_labels", "num_layers", "or", "output", "output_attentions", "output_hidden_states", "pass", "passed", "past_key_values", "post_init", "qa_outputs", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "shared", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "funnel/modeling_funnel.py:FunnelEmbeddings": ["Dropout", "Embedding", "LayerNorm", "ModelConfig", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "class", "config", "d_model", "def", "dropout", "embeddings", "eps", "forward", "hidden_dropout", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "layer_norm", "layer_norm_eps", "nn", "pad_token_id", "padding_idx", "return", "self", "super", "torch", "vocab_size", "word_embeddings"], "funnel/modeling_funnel.py:FunnelAttentionStructure": ["Dropout", "ModelAttentionStructure", "ModelConfig", "Module", "None", "NotImplementedError", "Optional", "Tensor", "The", "True", "Union", "__init__", "and", "append", "arange", "are", "attention_inputs", "attention_mask", "attention_type", "avg_pool2d", "ax", "axis", "axis_slice", "block_index", "cat", "ceil_mode", "class", "cls_ids", "cls_mask", "cls_mat", "cls_pos", "cls_slice", "cls_token_type_id", "config", "cos", "cos_dropout", "cos_embed", "cos_embed_d", "d_model", "def", "device", "dim", "dtype", "elif", "else", "enc_slice", "expand", "factorized", "for", "freq_seq", "functional", "gather", "get_position_embeds", "hidden_dropout", "if", "in", "init_attention_inputs", "inputs_embeds", "int", "int64", "inv_freq", "is", "isinstance", "len", "list", "long", "max", "max_dist", "max_pool2d", "mean", "min", "min_dist", "mode", "modes", "ndim", "new_ones", "new_tensor", "nn", "not", "num_blocks", "num_remove", "omega", "output", "pad", "phi", "pi", "pool_q_only", "pool_tensor", "pooled_pos", "pooled_pos_id", "pooling_mult", "pooling_type", "pos", "pos_embed", "pos_id", "pos_seq", "position_embeds", "position_embeds_list", "position_embeds_no_pooling", "position_embeds_pooling", "post_attention_pooling", "pre_attention_pooling", "psi", "raise", "range", "ref_point", "rel_pos", "rel_pos_id", "relative_pos", "return", "self", "separate_cls", "seq_len", "shift", "sin", "sin_dropout", "sin_embed", "sin_embed_d", "sinusoid", "size", "slice", "str", "stride", "stride_pool", "stride_pool_pos", "suffix", "super", "supported", "tensor", "to", "token_type_ids", "token_type_ids_to_mat", "token_type_mat", "torch", "truncate_seq", "tuple", "type", "x", "zero_offset"], "funnel/modeling_funnel.py:_relative_shift_gather": ["Tensor", "_relative_shift_gather", "batch_size", "context_len", "def", "int", "max_rel_len", "n_head", "positional_attn", "reshape", "return", "seq_len", "shape", "shift", "torch"], "funnel/modeling_funnel.py:FunnelRelMultiheadAttention": ["Dropout", "False", "INF", "LayerNorm", "Linear", "ModelConfig", "ModelRelMultiheadAttention", "Module", "None", "Parameter", "Tensor", "_", "__init__", "_relative_shift_gather", "attention_dropout", "attention_inputs", "attention_mask", "attention_type", "attn_out", "attn_prob", "attn_score", "attn_vec", "batch_size", "bias", "bind", "binh", "bjnd", "block_index", "bnij", "bnis", "bnit", "bool", "class", "cls_mask", "config", "content_score", "context_len", "d_head", "d_model", "def", "diff_token_type", "dim", "dnh", "dtype", "einsum", "else", "eps", "expand", "factorized", "float", "forward", "hidden_dropout", "if", "int", "is", "jd", "k_head", "key", "layer_norm", "layer_norm_eps", "n_head", "nn", "not", "omega", "output", "output_attentions", "phi", "pi", "position_embeds", "positional_attn", "post_proj", "psi", "q_head", "q_r_attention", "q_r_attention_1", "q_r_attention_2", "query", "r", "r_head", "r_kernel", "r_r_bias", "r_s_bias", "r_w_bias", "relative_positional_attention", "relative_token_type_attention", "reshape", "return", "same_token_type", "scale", "seg_embed", "self", "seq_len", "shape", "shift", "snd", "softmax", "split", "super", "td", "tnh", "token_type_attn", "token_type_bias", "token_type_mat", "torch", "tuple", "u", "v", "v_head", "value", "view", "w_r", "where", "zeros"], "funnel/modeling_funnel.py:FunnelPositionwiseFFN": ["ACT2FN", "Dropout", "LayerNorm", "Linear", "ModelConfig", "ModelPositionwiseFFN", "Module", "None", "Tensor", "__init__", "activation_dropout", "activation_function", "class", "config", "d_inner", "d_model", "def", "dropout", "forward", "h", "hidden", "hidden_act", "hidden_dropout", "layer_norm", "layer_norm_eps", "linear_1", "linear_2", "nn", "return", "self", "super", "torch"], "funnel/modeling_funnel.py:FunnelLayer": ["False", "ModelConfig", "ModelLayer", "ModelPositionwiseFFN", "ModelRelMultiheadAttention", "Module", "None", "Tensor", "__init__", "attention", "attention_inputs", "attn", "block_index", "bool", "class", "config", "def", "else", "ffn", "forward", "if", "int", "key", "nn", "output", "output_attentions", "query", "return", "self", "super", "torch", "tuple", "value"], "funnel/modeling_funnel.py:FunnelEncoder": ["BaseModelOutput", "False", "ModelAttentionStructure", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_attentions", "all_hidden_states", "and", "attention_inputs", "attention_mask", "attention_structure", "attentions", "block", "block_index", "block_repeats", "block_size", "block_sizes", "blocks", "bool", "class", "config", "def", "do_pooling", "else", "enumerate", "for", "forward", "hidden", "hidden_states", "if", "in", "init_attention_inputs", "inputs_embeds", "is", "key", "last_hidden_state", "layer", "layer_index", "layer_output", "nn", "not", "output_attentions", "output_hidden_states", "pool_q_only", "pooled_hidden", "pooling_flag", "post_attention_pooling", "pre_attention_pooling", "query", "range", "repeat_index", "return", "return_dict", "self", "separate_cls", "size", "super", "token_type_ids", "torch", "tuple", "type_as", "v", "value"], "funnel/modeling_funnel.py:upsample": ["False", "Model", "Tensor", "True", "bool", "cat", "cls", "def", "dim", "else", "functional", "if", "int", "nn", "output", "pad", "repeat_interleave", "repeats", "return", "separate_cls", "stride", "target_len", "torch", "truncate_seq", "x"], "funnel/modeling_funnel.py:FunnelDecoder": ["BaseModelOutput", "False", "ModelAttentionStructure", "ModelConfig", "ModelDecoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_attentions", "all_hidden_states", "attention_inputs", "attention_mask", "attention_structure", "attentions", "block_sizes", "bool", "class", "config", "def", "else", "final_hidden", "first_block_hidden", "for", "forward", "hidden", "hidden_states", "if", "in", "init_attention_inputs", "is", "last_hidden_state", "layer", "layer_output", "layers", "len", "nn", "not", "num_decoder_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "separate_cls", "shape", "stride", "super", "target_len", "token_type_ids", "torch", "truncate_seq", "tuple", "upsample", "upsampled_hidden", "v"], "funnel/modeling_funnel.py:FunnelDiscriminatorPredictions": ["ACT2FN", "Linear", "ModelConfig", "ModelDiscriminatorPredictions", "Module", "None", "Tensor", "__init__", "class", "config", "d_model", "def", "dense", "dense_prediction", "discriminator_hidden_states", "forward", "hidden_act", "hidden_states", "logits", "nn", "return", "self", "squeeze", "super", "torch"], "funnel/modeling_funnel.py:FunnelPreTrainedModel": ["Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "ModelRelMultiheadAttention", "None", "PreTrainedModel", "__class__", "__name__", "_init_weights", "b", "base_model_prefix", "bias", "class", "classname", "config", "constant_", "data", "def", "elif", "else", "fan_in", "fan_out", "find", "float", "getattr", "if", "init", "initializer_range", "initializer_std", "is", "module", "nn", "normal_", "not", "np", "padding_idx", "r_kernel", "r_r_bias", "r_s_bias", "r_w_bias", "seg_embed", "self", "shape", "sqrt", "std", "uniform_", "weight", "word_embeddings", "zero_"], "funnel/modeling_funnel.py:FunnelClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "ModelConfig", "Module", "None", "Tensor", "__init__", "class", "config", "d_model", "def", "dropout", "forward", "hidden", "hidden_dropout", "int", "linear_hidden", "linear_out", "n_labels", "nn", "return", "self", "super", "tanh", "torch"], "funnel/modeling_funnel.py:FunnelForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "funnel/modeling_funnel.py:FunnelBaseModel": ["BaseModelOutput", "Embedding", "ModelBaseModel", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "and", "at", "attention_mask", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embeddings", "encoder", "encoder_outputs", "forward", "get_input_embeddings", "have", "head_mask", "if", "input_ids", "input_shape", "inputs_embeds", "is", "long", "new_embeddings", "nn", "not", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "raise", "return", "return_dict", "same", "self", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "funnel/modeling_funnel.py:FunnelModel": ["BaseModelOutput", "Embedding", "ModelConfig", "ModelDecoder", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "and", "at", "attention_mask", "attentions", "auto_docstring", "block_sizes", "bool", "both", "cannot", "class", "config", "decoder", "decoder_outputs", "def", "device", "dtype", "either", "elif", "else", "embeddings", "encoder", "encoder_outputs", "final_hidden", "first_block_hidden", "forward", "get_input_embeddings", "have", "hidden_states", "idx", "if", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "long", "new_embeddings", "nn", "not", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "post_init", "raise", "return", "return_dict", "same", "self", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "funnel/modeling_funnel.py:FunnelForPreTraining": ["BCEWithLogitsLoss", "Model", "ModelConfig", "ModelDiscriminatorPredictions", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "active_labels", "active_logits", "active_loss", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "discriminator_hidden_states", "discriminator_predictions", "discriminator_sequence_output", "else", "float", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "output", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "shape", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "funnel/modeling_funnel.py:FunnelForMaskedLM": ["CrossEntropyLoss", "Embedding", "Linear", "MaskedLMOutput", "Model", "ModelConfig", "ModelForMaskedLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "d_model", "def", "else", "forward", "get_output_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "lm_head", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "prediction_logits", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "funnel/modeling_funnel.py:FunnelForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelBaseModel", "ModelClassificationHead", "ModelConfig", "ModelForSequenceClassification", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "last_hidden_state", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "funnel/modeling_funnel.py:FunnelForMultipleChoice": ["CrossEntropyLoss", "Model", "ModelBaseModel", "ModelClassificationHead", "ModelConfig", "ModelForMultipleChoice", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "logits", "loss", "loss_fct", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "funnel/modeling_funnel.py:FunnelForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelConfig", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "r", "return", "return_dict", "self", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "funnel/modeling_funnel.py:FunnelForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelConfig", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "last_hidden_state", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "qa_outputs", "return", "return_dict", "self", "size", "split", "squeeze", "squeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3PatchEmbeddings": ["Conv2d", "F", "Iterable", "ModelPatchEmbeddings", "Module", "None", "__init__", "abc", "bicubic", "class", "collections", "config", "def", "else", "embeddings", "flatten", "forward", "hidden_size", "if", "image_size", "input_size", "interpolate", "is", "isinstance", "kernel_size", "mode", "nn", "not", "num_channels", "patch_height", "patch_shape", "patch_size", "patch_width", "permute", "pixel_values", "position_embedding", "proj", "return", "self", "shape", "size", "stride", "super", "transpose", "view"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3TextEmbeddings": ["Dropout", "Embedding", "False", "IndexError", "LayerNorm", "ModelTextEmbeddings", "Module", "None", "The", "__init__", "arange", "as", "bbox", "be", "calculate_spatial_position_embeddings", "cat", "class", "clip", "config", "coordinate", "coordinate_size", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "e", "else", "embeddings", "eps", "except", "expand", "forward", "from", "h_position_embeddings", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "left_position_embeddings", "long", "lower_position_embeddings", "mask", "max_2d_position_embeddings", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "raise", "range", "register_buffer", "return", "right_position_embeddings", "self", "sequence_length", "shape_size", "should", "size", "spatial_position_embeddings", "super", "to", "token_type_embeddings", "token_type_ids", "torch", "try", "type_as", "type_vocab_size", "unsqueeze", "upper_position_embeddings", "values", "vocab_size", "w_position_embeddings", "within", "word_embeddings", "x_position_embeddings", "y_position_embeddings", "zeros"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3PreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "pos_embed", "self", "std", "visual_embed", "weight", "zero_"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3SelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "Softmax", "The", "ValueError", "_", "__init__", "a", "all_head_size", "alpha", "amax", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "class", "cogview_attention", "config", "context_layer", "contiguous", "def", "dim", "dropout", "elif", "else", "embedding_size", "f", "forward", "has_relative_attention_bias", "has_spatial_attention_bias", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "max_value", "multiple", "new_attention_scores", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "query", "query_layer", "raise", "rel_2d_pos", "rel_pos", "return", "scaled_attention_scores", "self", "seq_length", "shape", "size", "sqrt", "super", "the", "torch", "transpose", "unsqueeze", "value", "value_layer", "view"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3SelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3Attention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "attention_mask", "attention_output", "class", "config", "def", "forward", "head_mask", "hidden_states", "nn", "output", "output_attentions", "outputs", "rel_2d_pos", "rel_pos", "return", "self", "self_outputs", "super"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3Layer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "outputs", "rel_2d_pos", "rel_pos", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3Encoder": ["BaseModelOutput", "False", "Linear", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "_cal_1d_pos_emb", "_cal_2d_pos_emb", "abs", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bbox", "bias", "bidirectional", "class", "config", "contiguous", "def", "else", "enumerate", "float", "for", "forward", "full_like", "gradient_checkpointing", "has_relative_attention_bias", "has_spatial_attention_bias", "head_mask", "hidden_states", "i", "if", "in", "is", "is_small", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "log", "long", "math", "max", "max_distance", "max_exact", "max_rel_2d_pos", "max_rel_pos", "min", "n", "nn", "no_grad", "not", "num_attention_heads", "num_buckets", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_height", "patch_width", "permute", "position_coord_x", "position_coord_y", "position_ids", "range", "rel_2d_pos", "rel_2d_pos_bins", "rel_pos", "rel_pos_bias", "rel_pos_bins", "rel_pos_mat", "rel_pos_x", "rel_pos_x_2d_mat", "rel_pos_x_bias", "rel_pos_y", "rel_pos_y_2d_mat", "rel_pos_y_bias", "relative_position", "relative_position_bucket", "ret", "return", "return_dict", "self", "super", "t", "to", "torch", "tuple", "unsqueeze", "v", "val_if_large", "weight", "where", "with", "zeros_like"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3Intermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3Output": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3Model": ["BaseModelOutput", "Dropout", "FloatTensor", "LayerNorm", "LongTensor", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelTextEmbeddings", "None", "Optional", "Parameter", "Tensor", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "arange", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bbox", "bool", "calculate_visual_bbox", "cat", "class", "cls_token", "cls_token_box", "cls_tokens", "config", "def", "device", "dim", "div", "dropout", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "expand", "expand_as", "extended_attention_mask", "final_bbox", "final_position_ids", "for", "forward", "forward_image", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "has_relative_attention_bias", "has_spatial_attention_bias", "have", "head_mask", "heads", "heads_to_prune", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image_size", "in", "init_visual_bbox", "init_weights", "input_ids", "input_shape", "input_size", "inputs_embeds", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "len", "list", "long", "max_len", "nn", "norm", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "p", "patch_embed", "patch_height", "patch_size", "patch_width", "pixel_values", "pos_drop", "pos_embed", "position_ids", "prune_heads", "r", "raise", "repeat", "return", "return_dict", "rounding_mode", "self", "seq_len", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "stack", "super", "tensor", "text_embed", "to", "token_type_ids", "torch", "torch_int", "transpose", "trunc", "tuple", "type", "unsqueeze", "use_return_dict", "value", "view", "visual_attention_mask", "visual_bbox", "visual_bbox_x", "visual_bbox_y", "visual_embed", "visual_embeddings", "visual_position_ids", "word_embeddings", "zeros"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3ClassificationHead": ["Dropout", "False", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "nn", "not", "num_labels", "out_proj", "pool_feature", "return", "self", "super", "tanh", "torch", "x"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3ForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelClassificationHead", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_weights", "input_ids", "input_shape", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pool_feature", "position_ids", "r", "return", "return_dict", "self", "seq_length", "sequence_output", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3ForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "Model", "ModelClassificationHead", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_states", "if", "ignore_index", "ignored_index", "init_weights", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pool_feature", "position_ids", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "layoutlmv3/modeling_layoutlmv3.py:LayoutLMv3ForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pool_feature", "position_ids", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "paligemma/modeling_paligemma.py:PaligemmaModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "paligemma/modeling_paligemma.py:PaliGemmaCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "paligemma/modeling_paligemma.py:PaliGemmaMultiModalProjector": ["Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "True", "__init__", "bias", "class", "config", "def", "forward", "hidden_size", "hidden_states", "image_features", "linear", "nn", "projection_dim", "return", "self", "super", "vision_config"], "paligemma/modeling_paligemma.py:token_type_ids_mask_function": ["Callable", "Model_type_ids", "Model_type_ids_at_kv_idx", "Model_type_ids_mask_function", "None", "Optional", "Tensor", "batch_idx", "bool", "def", "head_idx", "if", "image_group_ids", "image_group_ids_at_kv_idx", "inner_mask", "int", "is", "is_image_block", "kv_idx", "q_idx", "return", "safe_idx", "same_image_block", "shape", "torch", "where"], "paligemma/modeling_paligemma.py:create_causal_mask_mapping": ["Cache", "False", "FloatTensor", "Model_causal_mask_mapping", "Model_masks_for_generate", "None", "Optional", "PretrainedConfig", "Tensor", "The", "ValueError", "We", "a", "and", "as", "attention", "attention_mask", "bad", "be", "bool", "but", "cache_position", "config", "cumsum", "def", "device", "dict", "dim", "else", "full_like", "functional", "get_text_config", "if", "image_group_ids", "input", "input_embeds", "int", "is", "is_image", "is_initialized", "is_previous_image", "is_training", "kwargs", "logger", "mask_kwargs", "masking", "may", "maybe_is_prompt", "model", "new_image_start", "nn", "not", "ones_like", "or", "or_mask_function", "pad", "passing", "past_key_values", "pixel_values", "position_ids", "prevent", "prompt", "provided", "raise", "recommend", "required", "return", "shape", "the", "to", "token_type_ids", "token_type_ids_mask_function", "torch", "training", "value", "warning_once", "when", "where"], "paligemma/modeling_paligemma.py:PaliGemmaPreTrainedModel": ["False", "Linear", "ModelConfig", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "get_text_config", "getattr", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "paligemma/modeling_paligemma.py:PaliGemmaModel": ["AutoModel", "Cache", "False", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "accepts_loss_kwargs", "all", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "causal_mask_mapping", "class", "clone", "config", "create_causal_mask_mapping", "decoder", "def", "device", "dict", "do", "dtype", "else", "exactly", "expand_as", "f", "features", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_seq_length", "get_text_config", "hidden_size", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "input_ids", "inputs_embeds", "is", "is_training", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "llm_input_ids", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "past_seen_tokens", "pixel_values", "position_ids", "post_init", "r", "raise", "return", "return_dict", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "text_config_dtype", "to", "token_type_ids", "tokens", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_config", "vision_tower", "vocab_size"], "paligemma/modeling_paligemma.py:PaliGemmaForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PretrainedConfig", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "create_causal_mask_mapping", "create_masks_for_generate", "decoder", "def", "dict", "else", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "in", "input_embeds", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "k", "kwargs", "labels", "language_model", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "staticmethod", "super", "text_config", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "v", "value", "vision_tower", "vocab_size", "weight"], "nystromformer/modeling_nystromformer.py:NystromformerEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "nystromformer/modeling_nystromformer.py:NystromformerSelfAttention": ["Conv2d", "Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "ValueError", "_", "__init__", "a", "absolute", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "class", "config", "context_layer", "contiguous", "conv", "conv_kernel_size", "def", "device", "dim", "dropout", "else", "embedding_size", "eye", "f", "for", "forward", "functional", "getattr", "groups", "hasattr", "heads", "hidden", "hidden_size", "hidden_states", "identity", "if", "in", "in_channels", "init_option", "int", "inv_coeff_init_option", "inv_init_coeff_option", "is", "iterative_inv", "k_landmarks", "kernel_1", "kernel_2", "kernel_3", "kernel_size", "key", "key_layer", "key_value", "mat", "math", "matmul", "max", "mean", "multiple", "n_iter", "new_context_layer_shape", "new_value_layer", "nn", "not", "num_attention_heads", "num_landmarks", "number", "of", "or", "original", "out_channels", "output_attentions", "outputs", "padding", "permute", "position_embedding_type", "q_landmarks", "query", "query_layer", "raise", "range", "reshape", "return", "segment_means_seq_len", "self", "seq_len", "seq_length", "shape", "size", "softmax", "sqrt", "sum", "super", "the", "torch", "transpose", "value", "value_layer", "values", "view"], "nystromformer/modeling_nystromformer.py:NystromformerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "nystromformer/modeling_nystromformer.py:NystromformerAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "nystromformer/modeling_nystromformer.py:NystromformerIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "nystromformer/modeling_nystromformer.py:NystromformerOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "nystromformer/modeling_nystromformer.py:NystromformerLayer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "add_cross_attention", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "nystromformer/modeling_nystromformer.py:NystromformerEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "nystromformer/modeling_nystromformer.py:NystromformerPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "nystromformer/modeling_nystromformer.py:NystromformerLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "nystromformer/modeling_nystromformer.py:NystromformerOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "nystromformer/modeling_nystromformer.py:NystromformerPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "nystromformer/modeling_nystromformer.py:NystromformerModel": ["BaseModelOutputWithPastAndCrossAttentions", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "nystromformer/modeling_nystromformer.py:NystromformerForMaskedLM": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size"], "nystromformer/modeling_nystromformer.py:NystromformerClassificationHead": ["ACT2FN", "Dropout", "Linear", "ModelClassificationHead", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "x"], "nystromformer/modeling_nystromformer.py:NystromformerForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "nystromformer/modeling_nystromformer.py:NystromformerForMultipleChoice": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "ReLU", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_state", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "pre_classifier", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "nystromformer/modeling_nystromformer.py:NystromformerForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "nystromformer/modeling_nystromformer.py:NystromformerForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "dinov2/modeling_dinov2.py:Dinov2Embeddings": ["Dropout", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "dtype", "embeddings", "expand", "float32", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask_token", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "projection", "randn", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "super", "target_dtype", "to", "torch", "torch_int", "unsqueeze", "use_mask_token", "view", "weight", "where", "width", "zeros"], "dinov2/modeling_dinov2.py:Dinov2PatchEmbeddings": ["Conv2d", "Expected", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "f", "flatten", "forward", "got", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "with"], "dinov2/modeling_dinov2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "dinov2/modeling_dinov2.py:Dinov2SelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "dinov2/modeling_dinov2.py:Dinov2SelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "dinov2/modeling_dinov2.py:Dinov2Attention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "dinov2/modeling_dinov2.py:Dinov2LayerScale": ["ModelLayerScale", "Module", "None", "Parameter", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_size", "hidden_state", "lambda1", "layerscale_value", "nn", "ones", "return", "self", "super", "torch"], "dinov2/modeling_dinov2.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "dinov2/modeling_dinov2.py:Dinov2DropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "dinov2/modeling_dinov2.py:Dinov2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "None", "Tensor", "True", "__init__", "activation", "bias", "class", "config", "def", "else", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "hidden_size", "hidden_state", "if", "in_features", "int", "isinstance", "mlp_ratio", "nn", "out_features", "return", "self", "str", "super", "torch"], "dinov2/modeling_dinov2.py:Dinov2SwiGLUFFN": ["Linear", "ModelSwiGLUFFN", "Module", "None", "Tensor", "True", "__init__", "bias", "chunk", "class", "config", "def", "dim", "forward", "functional", "hidden", "hidden_features", "hidden_size", "hidden_state", "in_features", "int", "mlp_ratio", "nn", "out_features", "return", "self", "silu", "super", "torch", "weights_in", "weights_out", "x1", "x2"], "dinov2/modeling_dinov2.py:Dinov2Layer": ["GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelLayerScale", "ModelMLP", "ModelSwiGLUFFN", "None", "Optional", "Tensor", "__init__", "attention", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "if", "layer_norm_eps", "layer_output", "layer_scale1", "layer_scale2", "mlp", "nn", "norm1", "norm2", "return", "self", "self_attention_output", "super", "torch", "use_swiglu_ffn"], "dinov2/modeling_dinov2.py:Dinov2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_hidden_states", "append", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "output_hidden_states", "range", "return", "self", "super", "torch", "tuple"], "dinov2/modeling_dinov2.py:Dinov2PreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelLayerScale", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "dtype", "elif", "fill_", "float32", "if", "init", "initializer_range", "is", "isinstance", "lambda1", "layerscale_value", "main_input_name", "mask_token", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "use_mask_token", "weight", "zero_"], "dinov2/modeling_dinov2.py:Dinov2Model": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Tensor", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "bool", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dict", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "int", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "num_hidden_layers", "output_hidden_states", "patch_embeddings", "pixel_values", "pooled_output", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch"], "dinov2/modeling_dinov2.py:Dinov2ForImageClassification": ["BaseModelOutputWithPooling", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attentions", "auto_docstring", "can_return_tuple", "cat", "class", "classifier", "cls_token", "config", "def", "dim", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "is", "kwargs", "labels", "last_hidden_state", "linear_input", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "outputs", "patch_tokens", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch"], "dinov2/modeling_dinov2.py:Dinov2Backbone": ["BackboneMixin", "BackboneOutput", "BaseModelOutput", "LayerNorm", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "_", "__init__", "_init_backbone", "append", "apply_layernorm", "auto_docstring", "batch_size", "bool", "check_model_inputs", "class", "config", "contiguous", "def", "else", "embedding_output", "embeddings", "encoder", "eps", "feature_maps", "for", "forward", "get_input_embeddings", "height", "hidden_size", "hidden_state", "hidden_states", "if", "in", "is", "kwargs", "layer_norm_eps", "layernorm", "nn", "num_features", "num_hidden_layers", "out_features", "output", "output_hidden_states", "patch_embeddings", "patch_size", "permute", "pixel_values", "post_init", "r", "range", "reshape", "reshape_hidden_states", "return", "self", "shape", "stage", "stage_names", "super", "torch", "tuple", "width", "zip"], "lxmert/modeling_lxmert.py:GeLU": ["GeLU", "Module", "__init__", "class", "def", "forward", "gelu", "nn", "return", "self", "super", "x"], "lxmert/modeling_lxmert.py:LxmertModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "cross_encoder_attentions", "language_attentions", "language_hidden_states", "language_output", "pooled_output", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states", "vision_output"], "lxmert/modeling_lxmert.py:LxmertForQuestionAnsweringOutput": ["FloatTensor", "ModelForQuestionAnsweringOutput", "ModelOutput", "None", "Optional", "class", "cross_encoder_attentions", "language_attentions", "language_hidden_states", "loss", "question_answering_score", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states"], "lxmert/modeling_lxmert.py:LxmertForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "class", "cross_encoder_attentions", "cross_relationship_score", "language_attentions", "language_hidden_states", "loss", "prediction_logits", "question_answering_score", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states"], "lxmert/modeling_lxmert.py:LxmertEmbeddings": ["Dropout", "Embedding", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "arange", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "long", "max_position_embeddings", "nn", "not", "padding_idx", "position_embeddings", "position_ids", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "lxmert/modeling_lxmert.py:LxmertAttention": ["Dropout", "False", "Linear", "ModelAttention", "Module", "None", "The", "ValueError", "_", "__init__", "a", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "class", "config", "context", "context_layer", "contiguous", "ctx_dim", "def", "dim", "dropout", "else", "f", "forward", "functional", "head_size", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "value", "value_layer", "view"], "lxmert/modeling_lxmert.py:LxmertAttentionOutput": ["Dropout", "LayerNorm", "Linear", "ModelAttentionOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertCrossAttentionLayer": ["False", "ModelAttention", "ModelAttentionOutput", "ModelCrossAttentionLayer", "Module", "None", "__init__", "att", "attention_output", "attention_probs", "class", "config", "ctx_att_mask", "ctx_tensor", "def", "else", "forward", "if", "input_tensor", "nn", "output", "output_attentions", "outputs", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertSelfAttentionLayer": ["False", "ModelAttention", "ModelAttentionOutput", "ModelSelfAttentionLayer", "Module", "__init__", "attention_mask", "attention_output", "attention_probs", "class", "config", "def", "else", "forward", "if", "input_tensor", "nn", "output", "output_attentions", "outputs", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "__init__", "class", "config", "def", "dense", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_act_fn", "intermediate_size", "nn", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertLayer": ["False", "ModelIntermediate", "ModelLayer", "ModelOutput", "ModelSelfAttentionLayer", "Module", "None", "__init__", "attention", "attention_mask", "attention_output", "class", "config", "def", "forward", "hidden_states", "intermediate", "intermediate_output", "layer_output", "nn", "output", "output_attentions", "outputs", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertXLayer": ["False", "ModelCrossAttentionLayer", "ModelIntermediate", "ModelOutput", "ModelSelfAttentionLayer", "ModelXLayer", "Module", "__init__", "attention_probs", "class", "config", "cross_att", "ctx_att_mask", "def", "else", "forward", "if", "lang_att_output", "lang_attention_mask", "lang_feats", "lang_input", "lang_inter", "lang_inter_output", "lang_output", "lang_self_att", "nn", "output_attentions", "output_fc", "output_x_attentions", "return", "self", "self_att", "super", "visn_inter", "visn_output", "visn_self_att", "visual_att_output", "visual_attention", "visual_attention_mask", "visual_feats", "visual_input", "visual_inter_output", "visual_output"], "lxmert/modeling_lxmert.py:LxmertVisualFeatureEncoder": ["Dropout", "LayerNorm", "Linear", "ModelVisualFeatureEncoder", "Module", "__init__", "box_fc", "box_layer_norm", "class", "config", "def", "dropout", "eps", "feat_dim", "forward", "hidden_dropout_prob", "hidden_size", "nn", "output", "pos_dim", "return", "self", "super", "visn_fc", "visn_layer_norm", "visual_feat_dim", "visual_feats", "visual_pos", "visual_pos_dim", "x", "y"], "lxmert/modeling_lxmert.py:LxmertEncoder": ["ModelEncoder", "ModelLayer", "ModelVisualFeatureEncoder", "ModelXLayer", "Module", "ModuleList", "None", "_", "__init__", "class", "config", "cross_encoder_attentions", "def", "else", "for", "forward", "if", "in", "is", "l_layers", "l_outputs", "lang_attention_mask", "lang_encoder_outputs", "lang_feats", "language_attentions", "language_hidden_states", "layer", "layer_module", "nn", "not", "num_l_layers", "num_r_layers", "num_x_layers", "or", "output_attentions", "r_layers", "range", "return", "self", "super", "v_outputs", "vision_attentions", "vision_hidden_states", "visn_fc", "visual_attention_mask", "visual_encoder_outputs", "visual_feats", "visual_pos", "x_layers", "x_outputs"], "lxmert/modeling_lxmert.py:LxmertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "nn", "return", "self", "super", "transform_act_fn"], "lxmert/modeling_lxmert.py:LxmertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Model_model_embedding_weights", "Module", "Parameter", "__init__", "bias", "class", "config", "decoder", "def", "forward", "hidden_states", "nn", "return", "self", "size", "super", "torch", "transform", "weight", "zeros"], "lxmert/modeling_lxmert.py:LxmertVisualAnswerHead": ["GeLU", "LayerNorm", "Linear", "ModelVisualAnswerHead", "Module", "Sequential", "__init__", "class", "config", "def", "eps", "forward", "hid_dim", "hidden_size", "hidden_states", "logit_fc", "nn", "num_labels", "return", "self", "super"], "lxmert/modeling_lxmert.py:LxmertVisualObjHead": ["Linear", "ModelPredictionHeadTransform", "ModelVisualObjHead", "Module", "ModuleDict", "__init__", "attr", "class", "config", "decoder_dict", "def", "feat", "for", "forward", "hidden_size", "hidden_states", "if", "in", "key", "nn", "num", "num_attr_labels", "num_object_labels", "obj", "output", "return", "self", "shape", "super", "transform", "visual_attr_loss", "visual_feat_dim", "visual_feat_loss", "visual_losses", "visual_obj_loss"], "lxmert/modeling_lxmert.py:LxmertPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Model_model_embedding_weights", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "lxmert/modeling_lxmert.py:LxmertPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_supports_param_buffer_assignment", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "lxmert/modeling_lxmert.py:LxmertModel": ["FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Union", "ValueError", "You", "__init__", "all_attentions", "and", "at", "attention_mask", "auto_docstring", "be", "bool", "both", "cannot", "class", "config", "cross_encoder_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "extended_visual_attention_mask", "finfo", "forward", "get_input_embeddings", "have", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "lang_encoder_outputs", "lang_output", "language_attentions", "language_hidden_states", "language_output", "long", "min", "new_embeddings", "not", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "post_init", "r", "raise", "return", "return_dict", "same", "self", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "unsqueeze", "use_return_dict", "vision_attentions", "vision_hidden_states", "vision_output", "visual_attention_mask", "visual_encoder_outputs", "visual_feats", "visual_output", "visual_pos", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "lxmert/modeling_lxmert.py:LxmertForPreTraining": ["CrossEntropyLoss", "Embedding", "False", "FloatTensor", "FutureWarning", "Linear", "LongTensor", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "ModelVisualAnswerHead", "ModelVisualObjHead", "Model_output", "Module", "None", "Optional", "Parameter", "SmoothL1Loss", "Tensor", "The", "True", "Union", "__init__", "_get_resized_qa_labels", "_init_weights", "_resize_bias", "_resize_qa_labels", "_set_qa_logit_layer", "_tie_weights", "_tied_weights_keys", "a", "and", "ans", "answer_head", "answer_loss", "answer_score", "argument", "attention_mask", "attr", "auto_docstring", "be", "bias", "bool", "cat", "ce", "class", "cls", "config", "cross_encoder_attentions", "cross_relationship_score", "cur_qa_labels", "cur_qa_logit_layer", "data", "decoder", "def", "deprecated", "device", "dict", "dim", "else", "embeddings", "extra_bias", "feat", "for", "forward", "future", "get_qa_logit_layer", "getattr", "hasattr", "hidden_dim", "if", "in", "input_ids", "inputs_embeds", "instead", "int", "is", "items", "key", "key_info", "kwargs", "l2", "label", "label_shape", "labels", "lang_output", "lang_prediction_scores", "language_attentions", "language_hidden_states", "logit_fc", "loss", "loss_fct_name", "loss_fcts", "mask_conf", "masked_lm_labels", "masked_lm_loss", "matched_label", "matched_loss", "mean", "mean_resizing", "min", "new_bias", "new_embeddings", "new_num_tokens", "new_qa_logit_layer", "nn", "none", "not", "num", "num_attr_labels", "num_labels", "num_labels_to_copy", "num_object_labels", "num_qa_labels", "obj", "obj_labels", "obj_predict_head", "old_num_tokens", "or", "output", "output_attentions", "output_dim", "output_hidden_states", "pad_to_multiple_of", "pooled_output", "pop", "post_init", "prediction_logits", "predictions", "qa_logit_layer", "question_answering_score", "r", "reduction", "removed", "resize_num_qa_labels", "resize_token_embeddings", "return", "return_dict", "self", "shape", "size", "str", "super", "task_mask_lm", "task_matched", "task_obj_predict", "task_qa", "tensor", "to", "token_type_ids", "torch", "total_loss", "total_visual_loss", "tuple", "use", "use_return_dict", "version", "view", "vision_attentions", "vision_hidden_states", "visual_attention_mask", "visual_attr_loss", "visual_ce", "visual_feat_dim", "visual_feat_loss", "visual_feats", "visual_loss", "visual_loss_fct", "visual_loss_normalizer", "visual_losses", "visual_obj_loss", "visual_output", "visual_pos", "visual_prediction_scores", "visual_prediction_scores_dict", "vocab_size", "warn", "warnings", "weight", "will", "word_embeddings", "zeros"], "lxmert/modeling_lxmert.py:LxmertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelForQuestionAnsweringOutput", "ModelModel", "ModelPreTrainedModel", "ModelVisualAnswerHead", "Model_output", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_get_resized_qa_labels", "_init_weights", "_resize_qa_labels", "_set_qa_logit_layer", "answer_head", "answer_score", "attention_mask", "auto_docstring", "bias", "bool", "class", "config", "cross_encoder_attentions", "cur_qa_labels", "cur_qa_logit_layer", "data", "def", "device", "else", "forward", "get_qa_logit_layer", "getattr", "hasattr", "hidden_dim", "if", "input_ids", "inputs_embeds", "is", "labels", "language_attentions", "language_hidden_states", "logit_fc", "loss", "min", "new_qa_logit_layer", "nn", "not", "num_labels", "num_labels_to_copy", "num_qa_labels", "or", "output", "output_attentions", "output_hidden_states", "pooled_output", "post_init", "qa_logit_layer", "question_answering_score", "r", "resize_num_qa_labels", "return", "return_dict", "self", "size", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vision_attentions", "vision_hidden_states", "visual_attention_mask", "visual_feats", "visual_loss_normalizer", "visual_pos", "weight"], "mistral/modeling_mistral.py:MistralMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "mistral/modeling_mistral.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "mistral/modeling_mistral.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "mistral/modeling_mistral.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "mistral/modeling_mistral.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "mistral/modeling_mistral.py:MistralAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "mistral/modeling_mistral.py:MistralRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "mistral/modeling_mistral.py:MistralDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "mistral/modeling_mistral.py:MistralPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "mistral/modeling_mistral.py:MistralRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "mistral/modeling_mistral.py:MistralModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "mistral/modeling_mistral.py:MistralForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "mistral/modeling_mistral.py:MistralForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "mistral/modeling_mistral.py:MistralForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "mistral/modeling_mistral.py:MistralForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "class"], "t5/modeling_t5.py:T5LayerNorm": ["Module", "Parameter", "T5LayerNorm", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "t5/modeling_t5.py:T5DenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "Module", "T5Config", "T5DenseActDense", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "t5/modeling_t5.py:T5DenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "Module", "T5Config", "T5DenseGatedActDense", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "t5/modeling_t5.py:T5LayerFF": ["DenseReluDense", "Dropout", "Module", "T5Config", "T5DenseActDense", "T5DenseGatedActDense", "T5LayerFF", "T5LayerNorm", "__init__", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is_gated_act", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "t5/modeling_t5.py:T5Attention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "Module", "None", "Optional", "Please", "T5Attention", "T5Config", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "if", "index", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "len", "list", "log", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "union", "unsqueeze", "update", "use_cache", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "t5/modeling_t5.py:T5LayerSelfAttention": ["Dropout", "False", "Module", "None", "Optional", "SelfAttention", "T5Attention", "T5LayerNorm", "T5LayerSelfAttention", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "t5/modeling_t5.py:T5LayerCrossAttention": ["Dropout", "EncDecAttention", "False", "Module", "None", "Optional", "T5Attention", "T5LayerCrossAttention", "T5LayerNorm", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "t5/modeling_t5.py:T5Block": ["False", "GradientCheckpointingLayer", "ModuleList", "None", "Optional", "T5Block", "T5LayerCrossAttention", "T5LayerFF", "T5LayerSelfAttention", "True", "__init__", "and", "any", "append", "attention_mask", "attention_outputs", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "encoder_attention_mask", "encoder_decoder_position_bias", "encoder_hidden_states", "finfo", "float16", "forward", "has_relative_attention_bias", "hidden_states", "if", "int", "is", "is_decoder", "isinf", "key_value_states", "layer", "layer_head_mask", "layer_idx", "max", "min", "new_name", "nn", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "return_dict", "self", "self_attention_outputs", "super", "torch", "use_cache", "version", "where"], "t5/modeling_t5.py:T5ClassificationHead": ["Dropout", "Linear", "Module", "T5ClassificationHead", "T5Config", "Tensor", "__init__", "class", "classifier_dropout", "config", "d_model", "def", "dense", "dropout", "forward", "hidden_states", "nn", "num_labels", "out_proj", "p", "return", "self", "super", "tanh", "torch"], "t5/modeling_t5.py:T5PreTrainedModel": ["DUMMY_INPUTS", "DUMMY_MASK", "In", "None", "PreTrainedModel", "See", "T5", "T5Attention", "T5Block", "T5ClassificationHead", "T5Config", "T5DenseActDense", "T5DenseGatedActDense", "T5EncoderModel", "T5ForConditionalGeneration", "T5ForQuestionAnswering", "T5ForTokenClassification", "T5LayerNorm", "T5Model", "T5PreTrainedModel", "True", "ValueError", "_can_compile_fullgraph", "_init_weights", "_keep_in_fp32_modules", "_no_split_modules", "_shift_right", "and", "base_model_prefix", "be", "bias", "cat", "class", "classifier", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "def", "defined", "dense", "dim", "docs", "dummy_inputs", "elif", "else", "factor", "fill_", "for", "full", "has", "has_relative_attention_bias", "hasattr", "if", "information", "initializer_factor", "input_ids", "input_mask", "is", "is_parallelizable", "is_torch_fx_proxy", "isinstance", "it", "k", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "normal_", "not", "num_heads", "o", "out_proj", "pad_token_id", "property", "q", "qa_outputs", "raise", "relative_attention_bias", "return", "self", "set", "shape", "shared", "shifted_input_ids", "std", "supports_gradient_checkpointing", "tensor", "the", "tie_word_embeddings", "to", "torch", "transformer", "usually", "v", "weight", "wi", "wi_0", "wi_1", "wo", "zero_"], "t5/modeling_t5.py:T5Stack": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "DEPARALLELIZE_DOCSTRING", "Dropout", "DynamicCache", "EncoderDecoderCache", "False", "FutureWarning", "Like", "ModuleList", "None", "PARALLELIZE_DOCSTRING", "Setting", "T5Block", "T5LayerNorm", "T5PreTrainedModel", "T5Stack", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "a", "add_start_docstrings", "all_attentions", "all_cross_attentions", "all_hidden_states", "also", "and", "any", "arange", "as", "assert_device_map", "at", "attention_mask", "attentions", "balanced", "batch_size", "be", "block", "bool", "both", "but", "cache_position", "call", "can", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cpu", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "cuda_device", "d_model", "decoder", "decoder_", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "diagonal", "dictionary", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "empty_cache", "encoder_attention_mask", "encoder_batch_size", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "first_device", "flash_attention_2", "flex_attention", "for", "forward", "from_pretrained", "full", "get_device_map", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_relative_attention_bias", "have", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "instance", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_torchdynamo_compiling", "is_training", "isinstance", "it", "items", "k", "keys", "kwargs", "last_device", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "len", "load", "logger", "long", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "max", "min", "min_dtype", "model", "model_parallel", "module_name", "needs", "new_embeddings", "nn", "not", "npu", "num_layers", "of", "ones", "only", "or", "output_attentions", "output_hidden_states", "own", "padding_mask", "parallelize", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_bias", "post_init", "provide", "raise", "range", "removed", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set", "set_device", "set_input_embeddings", "shape", "should", "size", "so", "specify", "staticmethod", "str", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "used", "using_compilable_cache", "v", "v5", "valid", "view", "warn", "warning_once", "warnings", "will", "with", "xpu", "you", "your"], "t5/modeling_t5.py:T5Model": ["BaseModelOutput", "BoolTensor", "Cache", "DEPARALLELIZE_DOCSTRING", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "Like", "LongTensor", "None", "Optional", "PARALLELIZE_DOCSTRING", "Seq2SeqModelOutput", "T5Config", "T5Model", "T5PreTrainedModel", "T5Stack", "Tensor", "Transformers", "True", "Union", "You", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "block", "bool", "but", "cache_position", "call", "can", "class", "config", "copy", "cpu", "cross_attentions", "cross_attn_head_mask", "cuda", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "elif", "else", "embed_tokens", "empty_cache", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "first_device", "for", "forward", "from_pretrained", "get_device_map", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instance", "is", "is_decoder", "isinstance", "it", "items", "last_hidden_state", "layer", "len", "load", "model", "model_parallel", "module_name", "needs", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "of", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "post_init", "provide", "prune_heads", "r", "range", "relative_attention_bias", "removed", "return", "return_dict", "self", "set_device", "set_input_embeddings", "shared", "should", "so", "super", "the", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "v5", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "t5/modeling_t5.py:T5ForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "Cache", "CrossEntropyLoss", "DEPARALLELIZE_DOCSTRING", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Like", "Linear", "LongTensor", "None", "Optional", "PARALLELIZE_DOCSTRING", "Seq2SeqLMOutput", "T5Config", "T5ForConditionalGeneration", "T5PreTrainedModel", "T5Stack", "Tensor", "Transformers", "True", "Union", "You", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "bias", "block", "bool", "but", "cache_position", "call", "can", "class", "config", "copy", "cpu", "cross_attentions", "cross_attn_head_mask", "cuda", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "elif", "else", "embed_tokens", "empty_cache", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "first_device", "for", "forward", "from_pretrained", "get_device_map", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "instance", "is", "is_decoder", "isinstance", "it", "labels", "last_hidden_state", "layer", "len", "lm_head", "lm_logits", "load", "logits", "loss", "loss_fct", "model", "model_dim", "model_parallel", "module_name", "needs", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "of", "output", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provide", "r", "range", "relative_attention_bias", "removed", "return", "return_dict", "self", "sequence_output", "set_device", "set_input_embeddings", "shared", "should", "size", "so", "super", "the", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "v5", "view", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "t5/modeling_t5.py:T5EncoderModel": ["BaseModelOutput", "DEPARALLELIZE_DOCSTRING", "Embedding", "False", "FloatTensor", "FutureWarning", "Like", "LongTensor", "None", "Optional", "PARALLELIZE_DOCSTRING", "SelfAttention", "T5Config", "T5EncoderModel", "T5PreTrainedModel", "T5Stack", "Transformers", "True", "Union", "You", "__init__", "_keys_to_ignore_on_load_unexpected", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "auto_docstring", "balanced", "be", "block", "bool", "but", "call", "can", "class", "config", "cpu", "cuda", "d_model", "decoder", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "else", "embed_tokens", "empty_cache", "encoder", "encoder_config", "encoder_outputs", "for", "forward", "from_pretrained", "get_device_map", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "instance", "is", "is_encoder_decoder", "it", "items", "layer", "len", "load", "model", "model_parallel", "module_name", "needs", "new_embeddings", "nn", "not", "of", "output_attentions", "output_hidden_states", "own", "parallelize", "post_init", "provide", "prune_heads", "r", "range", "removed", "return", "return_dict", "self", "set_input_embeddings", "shared", "should", "so", "super", "the", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "v5", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "t5/modeling_t5.py:T5ForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "EncDecAttention", "False", "FloatTensor", "If", "LongTensor", "MSELoss", "None", "NotImplementedError", "Optional", "Passing", "Please", "Seq2SeqSequenceClassifierOutput", "T5ClassificationHead", "T5Config", "T5ForSequenceClassification", "T5Model", "T5PreTrainedModel", "Tensor", "Union", "ValueError", "_", "__class__", "__init__", "__name__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tied_weights_keys", "and", "are", "attention_mask", "auto_docstring", "batch_size", "be", "block", "bool", "cannot", "class", "classification_head", "config", "cross_attentions", "cross_attn_head_mask", "currently", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_size", "if", "input", "input_ids", "inputs_embeds", "int", "is", "labels", "layer", "len", "list", "logits", "long", "loss", "loss_fct", "model_parallel", "multi_label_classification", "must", "no", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pass", "passed", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "relative_attention_bias", "return", "return_dict", "same", "self", "sentence_representation", "sequence_output", "shape", "single_label_classification", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "transformer", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "t5/modeling_t5.py:T5ForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "None", "Optional", "T5Config", "T5EncoderModel", "T5ForTokenClassification", "T5PreTrainedModel", "Tensor", "TokenClassifierOutput", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "embed_tokens", "encoder", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "r", "return", "return_dict", "self", "super", "torch", "transformer", "tuple", "use_return_dict", "view", "weight"], "t5/modeling_t5.py:T5ForQuestionAnswering": ["BaseModelOutput", "BoolTensor", "CrossEntropyLoss", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "If", "Linear", "LongTensor", "None", "Optional", "Please", "Seq2SeqQuestionAnsweringModelOutput", "T5Config", "T5ForQuestionAnswering", "T5PreTrainedModel", "T5Stack", "Tensor", "True", "Union", "ValueError", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "are", "attention_mask", "attentions", "auto_docstring", "be", "block", "bool", "cannot", "clamp", "class", "config", "contiguous", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "device", "dim", "either", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "last_hidden_state", "layer", "len", "logits", "loss", "loss_fct", "model_dim", "model_parallel", "new_embeddings", "nn", "no", "not", "num_decoder_layers", "num_labels", "num_layers", "or", "output", "output_attentions", "output_hidden_states", "pass", "passed", "past_key_values", "post_init", "qa_outputs", "r", "raise", "relative_attention_bias", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "shared", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "vocab_size", "warn", "warnings", "weight"], "rag/modeling_rag.py:RetrievAugLMMarginOutput": ["Cache", "FloatTensor", "LongTensor", "ModelAugLMMarginOutput", "ModelOutput", "Modeled_doc_embeds", "Modeled_doc_ids", "None", "Optional", "class", "context_attention_mask", "context_input_ids", "doc_scores", "generator_cross_attentions", "generator_dec_attentions", "generator_dec_hidden_states", "generator_enc_attentions", "generator_enc_hidden_states", "generator_enc_last_hidden_state", "logits", "loss", "past_key_values", "question_enc_attentions", "question_enc_hidden_states", "question_encoder_last_hidden_state", "r", "torch", "tuple"], "rag/modeling_rag.py:RetrievAugLMOutput": ["Cache", "FloatTensor", "LongTensor", "ModelAugLMOutput", "ModelOutput", "Modeled_doc_embeds", "Modeled_doc_ids", "None", "Optional", "class", "context_attention_mask", "context_input_ids", "doc_scores", "generator_cross_attentions", "generator_dec_attentions", "generator_dec_hidden_states", "generator_enc_attentions", "generator_enc_hidden_states", "generator_enc_last_hidden_state", "logits", "past_key_values", "question_enc_attentions", "question_enc_hidden_states", "question_encoder_last_hidden_state", "r", "torch", "tuple"], "rag/modeling_rag.py:RagPreTrainedModel": ["AutoConfig", "AutoModel", "AutoModelForSeq2SeqLM", "If", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelRetriever", "None", "Optional", "PreTrainedModel", "True", "_supports_flash_attn", "_supports_sdpa", "a", "an", "argument", "as", "assert", "base_model_prefix", "be", "class", "classmethod", "cls", "config", "def", "defined", "del", "for", "from_pretrained", "from_pretrained_question_encoder_generator", "from_question_encoder_generator_configs", "generator", "generator_", "generator_config", "generator_model", "generator_pretrained_model_name_or_path", "get", "has", "if", "in", "is", "items", "key", "kwargs", "kwargs_generator", "kwargs_question_encoder", "len", "model", "not", "pop", "question_encoder", "question_encoder_", "question_encoder_config", "question_encoder_pretrained_model_name_or_path", "r", "retriever", "return", "return_unused_kwargs", "startswith", "str", "to", "value"], "rag/modeling_rag.py:RagModel": ["Alternatively", "AutoModel", "AutoModelForSeq2SeqLM", "BoolTensor", "Cache", "Either", "False", "FloatTensor", "LongTensor", "Make", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelRetriever", "None", "Optional", "PreTrainedModel", "PretrainedConfig", "RetrievAugLMOutput", "Tensor", "The", "True", "Union", "__init__", "a", "an", "and", "are", "assert", "attention_mask", "attentions", "auto_docstring", "be", "bmm", "bool", "but", "can", "class", "config", "config_class", "configuration", "context_attention_mask", "context_encoder_training", "context_input_ids", "cpu", "cross_attentions", "ctx_encoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "def", "detach", "device", "dim", "dimension", "doc_ids", "doc_scores", "dtype", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "first", "float32", "forward", "from_config", "from_question_encoder_generator_configs", "function", "gen_outputs", "generator", "generator_cross_attentions", "generator_dec_attentions", "generator_dec_hidden_states", "generator_enc_attentions", "generator_enc_hidden_states", "generator_enc_last_hidden_state", "has", "has_to_retrieve", "hidden_states", "if", "input_ids", "int", "is", "isinstance", "kwargs", "logits", "multiple", "n_docs", "no", "not", "numpy", "of", "or", "output_attentions", "output_hidden_states", "output_retrieved", "passed", "passing", "past_key_values", "pooler_output", "prefix", "provided", "pt", "question_enc_attentions", "question_enc_hidden_states", "question_enc_outputs", "question_encoder", "question_encoder_last_hidden_state", "r", "repeat_interleave", "retrieved_doc_attention_mask", "retrieved_doc_embeds", "retrieved_doc_ids", "retrieved_doc_input_ids", "retriever", "retriever_outputs", "return", "return_dict", "return_tensors", "self", "set", "set_retriever", "shape", "should", "squeeze", "super", "sure", "that", "the", "to", "tokenized_doc_attention_mask", "tokenized_doc_ids", "torch", "transpose", "tuple", "type", "unsqueeze", "use_cache", "using", "view", "when", "you"], "rag/modeling_rag.py:RagSequenceForGeneration": ["Alternatively", "At", "BoolTensor", "Cache", "Either", "False", "FloatTensor", "LongTensor", "Make", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelRetriever", "ModelSequenceForGeneration", "Model_logprobs", "None", "Optional", "PreTrainedModel", "PretrainedConfig", "RetrievAugLMMarginOutput", "Tensor", "True", "__init__", "_cat_and_pad", "_mask_pads", "a", "all", "an", "and", "any", "append", "are", "assert", "attention_mask", "auto_docstring", "batch_size", "be", "bool", "bos_token_id", "can", "cat", "class", "config", "configuration", "context_attention_mask", "context_encoder_training", "context_input_ids", "cpu", "ctx_encoder", "decoder_attention_mask", "decoder_input_ids", "def", "detach", "device", "dim", "do_deduplication", "doc_logprobs", "doc_scores", "dtype", "else", "encoder", "encoder_outputs", "eps_i", "epsilon", "eq", "exclude_bos_score", "fill_", "first_token_scores", "float32", "for", "forward", "from_question_encoder_generator_configs", "function", "functional", "gather", "generate", "generator", "generator_cross_attentions", "generator_dec_attentions", "generator_dec_hidden_states", "generator_enc_attentions", "generator_enc_hidden_states", "generator_enc_last_hidden_state", "generator_input_ids", "get_nll", "given", "has", "hypos", "if", "in", "ind", "index", "individual_attention_mask", "individual_doc_scores", "individual_input_ids", "input_ids", "int", "is", "k", "keepdim", "kwargs", "label_smoothing", "labels", "least", "list", "ll", "log_softmax", "logits", "logsumexp", "loss", "masked_fill_", "max", "model_kwargs", "must", "n_docs", "new", "new_input_ids", "nll_loss", "nn", "no", "no_grad", "not", "num_beams", "num_candidates", "num_doc_return_sequences", "num_return_sequences", "numpy", "of", "one", "or", "output", "output_attentions", "output_hidden_states", "output_retrieved", "output_sequences", "outputs", "pad_mask", "pad_token_id", "passed", "past_key_values", "prefix", "property", "provided", "pt", "question_enc_attentions", "question_enc_hidden_states", "question_encoder", "question_encoder_last_hidden_state", "question_hidden_states", "r", "range", "reduce_loss", "remainder", "repeat", "retrieved_doc_embeds", "retrieved_doc_ids", "retriever", "return", "return_tensors", "second_token_scores", "self", "seq_logits", "seq_logprobs", "set", "set_context_encoder_for_training", "set_retriever", "shape", "size", "smooth_loss", "smooth_obj", "squeeze", "stack", "staticmethod", "str", "sum", "super", "sure", "t", "target", "tensors", "that", "the", "to", "tolist", "top_cand_inds", "topk", "torch", "tuple", "unsqueeze", "use_bos", "use_cache", "using", "values", "view", "you"], "rag/modeling_rag.py:RagTokenForGeneration": ["BoolTensor", "Cache", "Callable", "Either", "EncoderDecoderCache", "False", "FloatTensor", "GenerationConfig", "GenerationMixin", "LogitsProcessorList", "LongTensor", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelRetriever", "ModelTokenForGeneration", "Model_logprobs", "None", "Optional", "PreTrainedModel", "PretrainedConfig", "RetrievAugLMMarginOutput", "StoppingCriteriaList", "Tensor", "The", "True", "ValueError", "__init__", "_beam_search", "_get_logits_processor", "_get_stopping_criteria", "_mask_pads", "_prepare_cache_for_generation", "_prepare_special_tokens", "_reorder_cache", "_reorder_stacked", "_sample", "_temporary_reorder_cache", "a", "an", "and", "any", "assert", "attention_mask", "auto_docstring", "batch_size", "be", "beam_idx", "bmm", "bool", "but", "cat", "class", "clone", "config", "configuration", "context_attention_mask", "context_encoder_training", "context_input_ids", "copy", "cpu", "ctx_encoder", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "deepcopy", "def", "detach", "device", "dim", "dimension", "do_marginalize", "doc_logprobs", "doc_scores", "doing", "dtype", "elif", "else", "encoder", "encoder_input_ids", "encoder_outputs", "eps_i", "epsilon", "eq", "equal", "expand", "extend_enc_output", "f", "fill_", "first", "float32", "for", "forward", "from_legacy_cache", "from_question_encoder_generator_configs", "full", "functional", "gather", "generate", "generation_config", "generation_mode", "generator", "generator_cross_attentions", "generator_dec_attentions", "generator_dec_hidden_states", "generator_enc_attentions", "generator_enc_hidden_states", "generator_enc_last_hidden_state", "get", "get_encoder", "get_input_embeddings", "get_nll", "get_output_embeddings", "greedy", "has", "hidden_states", "if", "in", "index", "index_select", "input_ids", "input_ids_seq_length", "int", "integer", "is", "isinstance", "keepdim", "kwargs", "kwargs_has_attention_mask", "label_smoothing", "labels", "last_hidden_state", "layer_past", "list", "ll", "log_prob_sum", "log_softmax", "logits", "logits_processor", "logsumexp", "long", "loss", "marginalize", "masked_fill_", "max_cache_length", "max_length", "model_kwargs", "multiple", "n_docs", "new", "new_embeddings", "new_order", "new_zeros", "next", "nll_loss", "nn", "no_grad", "not", "num_beams", "num_return_sequences", "numpy", "of", "or", "out", "output_attentions", "output_hidden_states", "output_retrieved", "outputs", "pad_mask", "pad_token_id", "parameters", "past_key_values", "past_state", "pre_processor", "prefix", "prefix_allowed_tokens_fn", "prepare_inputs_for_generation", "prepared_stopping_criteria", "property", "provided", "pt", "question_enc_attentions", "question_enc_hidden_states", "question_encoder", "question_encoder_last_hidden_state", "question_hidden_states", "r", "raise", "reduce_loss", "reordered_past", "repeat_interleave", "reshape", "result", "retrieved_doc_embeds", "retrieved_doc_ids", "retriever", "return", "return_dict", "return_tensors", "search", "self", "seq_logits", "seq_logprobs", "set_context_encoder_for_training", "set_output_embeddings", "set_retriever", "shape", "shift_tokens_right", "shifted_input_ids", "should", "size", "smaller", "smooth_loss", "smooth_obj", "squeeze", "start_token_id", "staticmethod", "stopping_criteria", "streamer", "strictly", "sum", "super", "superior", "synced_gpus", "target", "tensor", "to", "torch", "transpose", "tuple", "unsqueeze", "update", "use_cache", "view", "when"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXMLP": ["ACT2FN", "Linear", "Model", "Module", "__init__", "act", "class", "config", "def", "dense_4h_to_h", "dense_h_to_4h", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super"], "gpt_neox/modeling_gpt_neox.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "gpt_neox/modeling_gpt_neox.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "gpt_neox/modeling_gpt_neox.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "FloatTensor", "Linear", "LongTensor", "Model", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "cache_kwargs", "cache_position", "chunk", "class", "config", "contiguous", "cos", "def", "dense", "dim", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_mask", "head_size", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "key_states", "kwargs", "layer_idx", "layer_past", "nn", "not", "num_attention_heads", "output_attentions", "partial_rotation_size", "position_embeddings", "qkv", "query_key_value", "query_states", "reshape", "return", "rotary_ndims", "rotary_pct", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "value_states", "view"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXLayer": ["Cache", "Dropout", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "Model", "ModelAttention", "ModelMLP", "None", "Optional", "Tensor", "Unpack", "__init__", "attention", "attention_mask", "attn_output", "attn_weights", "bool", "cache_position", "class", "config", "def", "else", "eps", "forward", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "kwargs", "layer_idx", "layer_norm_eps", "layer_past", "mlp", "mlp_output", "nn", "output_attentions", "outputs", "position_embeddings", "position_ids", "post_attention_dropout", "post_attention_layernorm", "post_mlp_dropout", "return", "self", "super", "torch", "tuple", "use_cache", "use_parallel_residual"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXRotaryEmbedding": ["False", "Model", "ModelConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXRMSNorm": ["Model", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXPreTrainedModel": ["Model", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLayer", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention", "attentions", "base_model_prefix", "bias", "class", "config", "hidden_states", "masked_bias", "past_key_values", "r", "supports_gradient_checkpointing"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXModel": ["BaseModelOutputWithPast", "Cache", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelLayer", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "all_attentions", "all_hidden_states", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "checkpointing", "class", "config", "converted_head_mask", "create_causal_mask", "def", "device", "dtype", "else", "emb_dropout", "embed_in", "enumerate", "eps", "exactly", "final_layer_norm", "finfo", "for", "forward", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layer_past", "layers", "logger", "min", "must", "nn", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "set_input_embeddings", "shape", "specify", "super", "to", "torch", "training", "unsqueeze", "use_cache", "value", "vocab_size", "warning_once", "with"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "embed_out", "forward", "get_output_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "logits", "logits_to_keep", "loss", "loss_function", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "set_output_embeddings", "slice", "slice_indices", "super", "torch", "tuple", "use_cache", "vocab_size", "weight"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXForSequenceClassification": ["BaseModelOutputWithPast", "Cache", "Cannot", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "can_return_tuple", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int32", "is", "labels", "last_hidden_state", "last_non_pad_token", "logger", "logits", "loss", "loss_function", "may", "nn", "no", "non_pad_mask", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "r", "raise", "return", "score", "self", "shape", "sizes", "super", "to", "token", "token_indices", "tokens", "torch", "tuple", "unexpected", "use_cache", "using", "warning_once", "will", "with"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXForTokenClassification": ["BaseModelOutputWithPast", "Cache", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "super", "token_type_ids", "torch", "tuple", "use_cache"], "gpt_neox/modeling_gpt_neox.py:GPTNeoXForQuestionAnswering": ["BaseModelOutputWithPast", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "contiguous", "def", "dim", "end_logits", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "self", "sequence_output", "split", "squeeze", "start_logits", "start_positions", "super", "token_type_ids", "torch"], "bigbird_pegasus/modeling_bigbird_pegasus.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusLearnedPositionalEmbedding": ["Embedding", "ModelLearnedPositionalEmbedding", "None", "Optional", "Size", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "embedding_dim", "forward", "if", "input_ids_shape", "int", "is", "long", "nn", "num_embeddings", "past_key_values_length", "position_ids", "return", "self", "seq_len", "super", "torch", "weight"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "cache_position", "class", "config", "context_layer", "contiguous", "current_states", "def", "deprecate_kwarg", "dim", "dropout", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "functional", "get_seq_length", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "keys", "layer_idx", "layers", "math", "matmul", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "past_key_value", "past_key_values", "permute", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "update", "use_bias", "value", "value_layer", "values", "version", "view"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusBlockSparseAttention": ["Error", "Key", "Linear", "Make", "ModelBlockSparseAttention", "Modelbird_block_sparse_attention", "Module", "None", "Query", "The", "Value", "ValueError", "_", "_Modelbird_block_rand_mask", "_Modelbird_block_rand_mask_with_head", "__init__", "_create_rand_mask_from_inputs", "_get_rand_attn_plan", "_get_single_block_row_attention", "a", "all_head_size", "and", "append", "arange", "are", "array", "attention", "attention_head_size", "attention_probs", "attn_mask_penalty", "attn_probs_view", "attn_weights", "axis", "band_mask", "band_product", "batch_size", "be", "bhkd", "bhlk", "bhlqd", "bhlqk", "bias", "blk_rw_idx", "block", "block_id", "block_size", "blocked_key_matrix", "blocked_query_matrix", "blocked_value_matrix", "blocks", "blq", "bmm", "break", "bsz", "but", "cat", "class", "concatenate", "config", "context_layer", "contiguous", "continue", "curr_r_cnt", "def", "device", "dim", "dimensions", "div", "dtype", "einsum", "elif", "else", "end", "exp_blocked_key_matrix", "exp_blocked_value_matrix", "extend", "f", "first", "first_attn_weights", "first_band_product", "first_context_layer", "first_product", "flatten", "flattened_indices", "flattened_params", "float", "floor", "for", "forward", "from", "from_block_size", "from_blocked_mask", "from_mask", "from_seq_len", "from_seq_length", "from_start_block_id", "functional", "gathered_key", "gathered_value", "global_block_bottom", "global_block_left", "global_block_right", "global_block_top", "h", "heads", "hidden", "hidden_size", "hidden_states", "i", "i1", "i2", "identical", "if", "illegal_blocks", "in", "index", "index_select", "indices", "indices_shift", "inner_band_product", "inp_1", "inp_2", "int", "int32", "is", "key", "key_layer", "last", "last_attn_weights", "last_band_product", "last_context_layer", "last_idx", "last_product", "len", "length", "list", "long", "math", "max_plan_idx", "max_position_embeddings", "max_seqlen", "middle_query_matrix", "middle_seq", "minimum", "multiple", "must", "n_heads", "n_rand_blocks", "ndim", "needs", "new_ones", "nh", "nn", "not", "np", "num_attention_heads", "num_blocks", "num_heads", "num_indices_to_gather", "num_indices_to_pick_from", "num_rand_blocks", "num_random_blocks", "num_windows", "number", "of", "out", "out_flattened", "output_attentions", "p1", "p2", "params", "perm_block", "permutation", "pl_id", "plan", "plan_block_length", "plan_from_length", "plan_idx", "plan_num_rand_blocks", "q_idx", "query", "query_layer", "r", "raise", "rand_attn", "rand_band_product", "rand_mask", "random", "range", "reshape", "return", "right_slice", "rnd_r_cnt", "rounding_mode", "rsqrt_d", "same", "second_attn_weights", "second_context_layer", "second_key_mat", "second_last_attn_weights", "second_last_context_layer", "second_last_key_mat", "second_last_product", "second_last_rand_pad", "second_last_seq_pad", "second_last_value_mat", "second_product", "second_rand_pad", "second_seq_pad", "second_value_mat", "seed", "selected_random_blocks", "self", "seqlen", "sequence", "shape", "shift", "sided", "size", "softmax", "sqrt", "stack", "start", "staticmethod", "sum", "super", "sure", "tensor", "that", "the", "they", "to", "to_block_list", "to_block_size", "to_blocked_mask", "to_end_block_id", "to_mask", "to_seq_len", "to_seq_length", "to_start_block_id", "torch", "torch_bmm_nd", "torch_bmm_nd_transpose", "torch_gather_b2", "training", "transpose", "two", "unsqueeze", "unsqueeze_", "use_bias", "value", "value_layer", "view", "vs", "w1", "w2", "window_block_left", "window_block_right", "zeros", "zip"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusEncoderAttention": ["False", "Linear", "ModelBlockSparseAttention", "ModelEncoderAttention", "ModelSelfAttention", "Module", "None", "ValueError", "__init__", "attention_mask", "attention_output", "attention_type", "attn_weights", "band_mask", "be", "bias", "block_sparse", "but", "can", "class", "config", "def", "either", "elif", "else", "eval", "f", "forward", "from_blocked_mask", "from_mask", "head_mask", "hidden_size", "hidden_states", "if", "in", "is", "key", "nn", "not", "only", "or", "original_full", "output", "output_attentions", "outputs", "query", "raise", "reshape", "return", "seed", "self", "self_outputs", "set", "set_attention_type", "str", "super", "to", "to_blocked_mask", "to_mask", "training", "use_bias", "value"], "bigbird_pegasus/modeling_bigbird_pegasus.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusDecoderAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelConfig", "ModelDecoderAttention", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderAttention", "ModelEncoderLayer", "None", "Tensor", "ValueError", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_mask", "attention_type", "band_mask", "be", "block_sparse", "bool", "but", "can", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "either", "embed_dim", "encoder_ffn_dim", "f", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "from_blocked_mask", "from_mask", "functional", "head_mask", "hidden_states", "if", "in", "is", "isinf", "isnan", "layer_head_mask", "max", "min", "nn", "not", "only", "or", "original_full", "output_attentions", "outputs", "p", "raise", "residual", "return", "seed", "self", "self_attention_outputs", "self_attn", "self_attn_layer_norm", "set", "set_attention_type", "str", "super", "to", "to_blocked_mask", "to_mask", "torch", "training", "value"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderAttention", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bias", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_bias", "use_cache"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "Tensor", "__init__", "class", "def", "dense", "dropout", "float", "forward", "hidden_states", "inner_dim", "input_dim", "int", "nn", "num_classes", "out_proj", "p", "pooler_dropout", "return", "self", "super", "tanh", "torch"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_skip_keys_device_placement", "_supports_param_buffer_assignment", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "def", "device", "diagonal", "dim", "dtype", "dummy_inputs", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "ne", "nn", "normal_", "not", "npu", "ones", "pad_token", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "property", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tensor", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusEncoder": ["Attention", "BaseModelOutput", "Changing", "Embedding", "False", "Input", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Sequence", "Tensor", "The", "True", "ValueError", "You", "__init__", "_pad_to_block_size", "_prepare_4d_attention_mask", "a", "additional", "all_attentions", "and", "are", "at", "attention", "attention_mask", "attention_type", "attentions", "automatically", "band_mask", "batch_size", "be", "blk", "block", "block_size", "block_sparse", "blocked_encoder_mask", "blq", "blqk", "bool", "both", "buffer", "but", "can", "cannot", "cat", "class", "config", "create_band_mask_from_inputs", "create_masks_for_block_sparse_attn", "d_model", "def", "device", "dim", "dropout", "dropout_probability", "dtype", "einsum", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "exp_blocked_to_pad", "f", "for", "forward", "from", "from_blocked_mask", "from_mask", "functional", "global", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "ids", "idx", "if", "in", "input_ids", "input_ids_padding", "input_shape", "inputs_embeds", "inputs_embeds_padding", "int", "is", "it", "last_hidden_state", "layer", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "length", "logger", "long", "math", "max_position_embeddings", "max_source_positions", "max_tokens_to_attend", "min", "multiple", "must", "nn", "not", "num", "num_random_blocks", "of", "ones", "only", "or", "original_full", "output_attentions", "output_hidden_states", "p", "pad", "pad_id", "pad_token_id", "padded", "padding_idx", "padding_len", "possible", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "seed", "self", "seq_len", "seq_length", "sequence", "sequence_length", "set", "set_attention_type", "shape", "should", "size", "sliding", "specified", "specify", "sqrt", "staticmethod", "str", "super", "the", "time", "to", "to_blocked_mask", "to_drop", "to_mask", "tokens", "torch", "training", "tuple", "type", "unsqueeze_", "use_return_dict", "v", "value", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "warning", "warning_once", "weight", "while", "with"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with", "zip"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusModel": ["BaseModelOutput", "Cache", "FloatTensor", "If", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Please", "Seq2SeqModelOutput", "Tensor", "Union", "ValueError", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "are", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache_position", "cannot", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "either", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "list", "math", "no", "not", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "pass", "passed", "past_key_values", "post_init", "r", "raise", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "shift_tokens_right", "sqrt", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "list", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "super", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "ModelClassificationHead", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "classification_head", "classifier_dropout", "config", "cross_attentions", "cross_attn_head_mask", "currently", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_states", "if", "input", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "len", "list", "logits", "long", "loss", "loss_fct", "model", "multi_label_classification", "must", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "same", "self", "sentence_representation", "single_label_classification", "size", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqQuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "clamp", "class", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "list", "logits", "loss", "loss_fct", "model", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "weight"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "bigbird_pegasus/modeling_bigbird_pegasus.py:BigBirdPegasusForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "phi3/modeling_phi3.py:Phi3MLP": ["ACT2FN", "False", "FloatTensor", "Linear", "ModelMLP", "Module", "__init__", "activation_fn", "bias", "chunk", "class", "config", "def", "dim", "down_proj", "forward", "gate", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch", "up_states"], "phi3/modeling_phi3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "phi3/modeling_phi3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "phi3/modeling_phi3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "phi3/modeling_phi3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "phi3/modeling_phi3.py:Phi3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "op_size", "past_key_value", "past_key_values", "position_embeddings", "qkv", "qkv_proj", "query_pos", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "value_states", "version", "view"], "phi3/modeling_phi3.py:Phi3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "phi3/modeling_phi3.py:Phi3DecoderLayer": ["Cache", "Dropout", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "nn", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "resid_attn_dropout", "resid_mlp_dropout", "resid_pdrop", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "phi3/modeling_phi3.py:Phi3PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_version", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "phi3/modeling_phi3.py:Phi3RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "phi3/modeling_phi3.py:Phi3Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "phi3/modeling_phi3.py:Phi3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "nn", "not", "original_max_position_embeddings", "outputs", "past_key_values", "past_length", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "rope_scaling", "self", "shape", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "phi3/modeling_phi3.py:Phi3ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "phi3/modeling_phi3.py:Phi3ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "unispeech/modeling_unispeech.py:UniSpeechForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "codevector_perplexity", "hidden_states", "loss", "projected_quantized_states", "projected_states", "r", "torch", "tuple"], "unispeech/modeling_unispeech.py:UniSpeechSamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "unispeech/modeling_unispeech.py:UniSpeechPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "unispeech/modeling_unispeech.py:UniSpeechNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "unispeech/modeling_unispeech.py:UniSpeechLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "unispeech/modeling_unispeech.py:UniSpeechGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "unispeech/modeling_unispeech.py:UniSpeechFeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "unispeech/modeling_unispeech.py:UniSpeechFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "unispeech/modeling_unispeech.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "unispeech/modeling_unispeech.py:UniSpeechAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "unispeech/modeling_unispeech.py:UniSpeechFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "unispeech/modeling_unispeech.py:UniSpeechEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelFeedForward", "None", "_", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super"], "unispeech/modeling_unispeech.py:UniSpeechEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalConvEmbedding", "Modelon", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "tensor", "torch", "training", "tuple", "unsqueeze", "v"], "unispeech/modeling_unispeech.py:UniSpeechAttnAdapterLayer": ["FloatTensor", "LayerNorm", "Linear", "ModelAttnAdapterLayer", "Module", "ReLU", "__init__", "act_fn", "adapter_attn_dim", "class", "config", "def", "forward", "hidden_dim", "hidden_size", "hidden_states", "input_dim", "linear_1", "linear_2", "nn", "norm", "return", "self", "super", "torch"], "unispeech/modeling_unispeech.py:UniSpeechEncoderLayerStableLayerNorm": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelAttnAdapterLayer", "ModelEncoderLayerStableLayerNorm", "ModelFeedForward", "None", "Optional", "Tensor", "_", "__init__", "adapter_attn_dim", "adapter_layer", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "bool", "class", "config", "def", "dropout", "else", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "getattr", "hidden_dropout", "hidden_size", "hidden_states", "if", "is", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super", "torch"], "unispeech/modeling_unispeech.py:UniSpeechEncoderStableLayerNorm": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoderLayerStableLayerNorm", "ModelEncoderStableLayerNorm", "ModelPositionalConvEmbedding", "Modelon", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "unsqueeze", "v"], "unispeech/modeling_unispeech.py:UniSpeechGumbelVectorQuantizer": ["FloatTensor", "Linear", "ModelGumbelVectorQuantizer", "Module", "Parameter", "True", "ValueError", "__init__", "_compute_perplexity", "argmax", "batch_size", "be", "by", "class", "codevector_dim", "codevector_idx", "codevector_probs", "codevector_soft_dist", "codevectors", "codevectors_per_group", "concatenation", "config", "conv_dim", "def", "dim", "divisible", "else", "exp", "f", "float", "for", "forward", "functional", "gumbel_softmax", "hard", "hidden_size", "hidden_states", "if", "log", "marginal_probs", "mean", "must", "new_zeros", "nn", "num_codevector_groups", "num_codevectors_per_group", "num_groups", "num_vars", "perplexity", "probs", "raise", "return", "scatter_", "self", "sequence_length", "shape", "softmax", "staticmethod", "sum", "super", "tau", "temperature", "torch", "training", "type_as", "unsqueeze", "view", "weight_proj"], "unispeech/modeling_unispeech.py:UniSpeechPreTrainedModel": ["Conv1d", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelFeatureProjection", "ModelGumbelVectorQuantizer", "ModelPositionalConvEmbedding", "ModelPreTrainedModel", "Modelform_", "Modelon", "None", "PreTrainedModel", "True", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "arange", "attention_mask", "b", "base_model_prefix", "batch_size", "bias", "bool", "class", "codevectors", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "data", "def", "device", "dim", "div", "dtype", "elif", "feature_vector_length", "fill_", "flip", "floor", "for", "groups", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "module", "nn", "non_padded_lengths", "normal_", "not", "output_lengths", "projection", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "supports_gradient_checkpointing", "to", "torch", "weight", "weight_proj", "zero_", "zeros", "zip"], "unispeech/modeling_unispeech.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "unispeech/modeling_unispeech.py:UniSpeechModel": ["FloatTensor", "LongTensor", "ModelBaseModelOutput", "ModelConfig", "ModelEncoder", "ModelEncoderStableLayerNorm", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModel", "ModelPreTrainedModel", "Modelform_", "Modelon", "None", "Optional", "Parameter", "Tensor", "True", "__init__", "_compute_mask_indices", "_get_feature_vector_attention_mask", "_mask_hidden_states", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "do_stable_layer_norm", "dtype", "elif", "else", "encoder", "encoder_outputs", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "use_return_dict"], "unispeech/modeling_unispeech.py:UniSpeechForPreTraining": ["Dropout", "FloatTensor", "FutureWarning", "Linear", "Model", "ModelConfig", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelGumbelVectorQuantizer", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Please", "Tensor", "The", "Transformers", "__init__", "_freeze_parameters", "and", "attention_mask", "attentions", "auto_docstring", "be", "bernoulli", "bool", "cat", "class", "codevector_dim", "codevector_perplexity", "compute_contrastive_logits", "config", "cosine_similarity", "ctc_proj", "def", "deprecated", "device", "dim", "dropout", "dropout_features", "dtype", "else", "empty", "equivalent", "extract_features", "feat_quantizer_dropout", "feature_extractor", "fill_", "final_dropout", "float", "forward", "freeze_feature_encoder", "freeze_feature_extractor", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "int", "is", "logits", "loss", "masked_fill", "method", "negative_features", "nn", "not", "num_ctc_classes", "output_attentions", "output_hidden_states", "outputs", "post_init", "predicted_features", "prob_replace_matrix", "proj_codevector_dim", "project_hid", "project_q", "projected_quantized_states", "projected_states", "quantized_features", "quantizer", "r", "removed", "replace_prob", "return", "return_dict", "sampled_replace_matrix", "self", "set_gumbel_temperature", "size", "staticmethod", "super", "target_features", "temperature", "the", "to", "torch", "transformer_features", "transpose", "tuple", "type_as", "unsqueeze", "use", "use_return_dict", "v5", "warn", "warnings", "weight", "will"], "unispeech/modeling_unispeech.py:UniSpeechForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "unispeech/modeling_unispeech.py:UniSpeechForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "olmo/modeling_olmo.py:OlmoLayerNorm": ["F", "ModelLayerNorm", "Module", "None", "Tensor", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "int", "layer_norm", "nn", "normalized_shape", "orig_dtype", "return", "self", "super", "to", "torch"], "olmo/modeling_olmo.py:OlmoMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "olmo/modeling_olmo.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "olmo/modeling_olmo.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "olmo/modeling_olmo.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "olmo/modeling_olmo.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "k", "k_embed", "k_type", "position_ids", "q", "q_embed", "q_type", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "olmo/modeling_olmo.py:OlmoAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "clamp_", "class", "clip_qkv", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "max", "min", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "olmo/modeling_olmo.py:OlmoDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm", "ModelMLP", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "olmo/modeling_olmo.py:OlmoRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "olmo/modeling_olmo.py:OlmoPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "olmo/modeling_olmo.py:OlmoModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "olmo/modeling_olmo.py:OlmoForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "led/modeling_led.py:shift_tokens_right": ["ModeModel_input_ids", "Model_tokens_right", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "new_zeros", "pad_token_id", "raise", "return", "shape", "to", "torch"], "led/modeling_led.py:_prepare_4d_attention_mask_inverted": ["None", "Optional", "Tensor", "_prepare_4d_attention_mask_inverted", "bool", "bsz", "def", "dtype", "else", "expand", "expanded_attention_mask", "expanded_mask", "finfo", "if", "int", "inverted_mask", "is", "mask", "masked_fill", "min", "not", "return", "size", "src_len", "tgt_len", "to", "torch"], "led/modeling_led.py:LEDLearnedPositionalEmbedding": ["Embedding", "Model", "Size", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "embedding_dim", "forward", "input_ids_shape", "int", "long", "nn", "num_embeddings", "past_key_values_length", "positions", "return", "self", "seq_len", "super", "torch", "weight"], "led/modeling_led.py:LEDEncoderSelfAttention": ["False", "Given", "Head", "Linear", "Model", "Module", "None", "Sequence", "Size", "Tensor", "The", "True", "Unexpected", "ValueError", "__init__", "_chunk", "_compute_attn_output_with_global_indices", "_compute_global_attn_output_from_hidden", "_concat_with_global_key_attn_probs", "_get_global_attn_indices", "_mask_invalid_locations", "_pad_and_diagonalize", "_pad_and_transpose_last_two_dims", "_sliding_chunks_matmul_attn_probs_value", "_sliding_chunks_query_key_matmul", "a", "affected_seq_len", "an", "and", "arange", "as_strided", "as_tuple", "assert", "attention", "attention_mask", "attention_probs_dropout_prob", "attention_window", "attn_output", "attn_output_only_global", "attn_output_without_global", "attn_probs", "attn_probs_from_global_key", "attn_probs_only_global", "attn_probs_without_global", "attn_scores", "batch_size", "bcdh", "bcwd", "bcwh", "bcxd", "bcxy", "bcyd", "be", "beginning_input", "beginning_mask", "beginning_mask_2d", "blhd", "blhs", "bmm", "bool", "bshd", "but", "cat", "chunk", "chunk_size", "chunk_stride", "chunked_attn_probs", "chunked_hidden_states", "chunked_value", "chunked_value_size", "chunked_value_stride", "chunks_count", "class", "clone", "config", "context", "contiguous", "def", "del", "device", "diagonal_attention_scores", "diagonal_chunked_attention_scores", "diagonal_mask", "dim", "dims", "div", "dropout", "dtype", "einsum", "else", "embed_dim", "empty", "ending_input", "ending_mask", "even", "expand", "f", "finfo", "flip", "float", "float32", "float_mask", "for", "forward", "full_like", "functional", "getattr", "global_attn_hidden_states", "global_attn_output", "global_attn_probs", "global_attn_probs_float", "global_attn_scores", "global_key_attn_scores", "global_key_vectors", "global_query_vectors_only_global", "global_value_vectors", "has", "have", "head_dim", "heads", "hidden", "hidden_dim", "hidden_size", "hidden_states", "hidden_states_padded", "if", "in", "inf", "input_tensor", "int", "is", "is_global_attn", "is_index_global_attn", "is_index_global_attn_nonzero", "is_index_masked", "is_local_index_global_attn", "is_local_index_global_attn_nonzero", "is_local_index_no_global_attn_nonzero", "key", "key_global", "key_vectors", "key_vectors_only_global", "layer", "layer_head_mask", "layer_id", "len", "length", "list", "local_attn_probs", "long", "mask", "masked_fill", "math", "matmul", "max", "max_num_global_attn_indices", "min", "multiple", "narrow", "new_ones", "new_zeros", "nn", "nonzero", "nonzero_global_attn_output", "not", "num_attention_heads", "num_chunks", "num_global_attn_indices", "num_heads", "number", "of", "one_sided_attn_window_size", "onnx_export", "output_attentions", "outputs", "overlapping_chunks", "p", "pad", "padded_value", "padding", "positive", "query", "query_global", "query_vectors", "raise", "range", "remove_from_windowed_attention_mask", "reshape", "return", "rounding_mode", "self", "seq_len", "shape", "should", "single", "size", "softmax", "sqrt", "staticmethod", "stride", "sum", "super", "tensor", "the", "to", "torch", "total_num_heads", "training", "transpose", "tril", "trunc", "type_as", "unsqueeze", "value", "value_global", "value_vectors", "value_vectors_only_global", "view", "where", "window_overlap", "wrong"], "led/modeling_led.py:LEDEncoderAttention": ["False", "Linear", "Model", "ModelEncoderSelfAttention", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_output", "bool", "class", "config", "d_model", "def", "forward", "hidden_states", "is_global_attn", "is_index_global_attn", "is_index_masked", "layer_head_mask", "layer_id", "longformer_self_attn", "nn", "output", "output_attentions", "outputs", "return", "self", "self_outputs", "super", "torch", "tuple"], "led/modeling_led.py:LEDDecoderAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "class", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "else", "embed_dim", "f", "float", "for", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "mask", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "should", "single", "size", "softmax", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view", "weights"], "led/modeling_led.py:LEDEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEncoderAttention", "None", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_mask", "attn_outputs", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "int", "is_global_attn", "is_index_global_attn", "is_index_masked", "isinf", "isnan", "layer_head_mask", "layer_id", "max", "min", "nn", "or", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "led/modeling_led.py:LEDDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoderAttention", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_present_key_value", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "present_key_value", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "led/modeling_led.py:LEDClassificationHead": ["Dropout", "Linear", "Model", "Module", "Tensor", "__init__", "class", "def", "dense", "dropout", "float", "forward", "hidden_states", "inner_dim", "input_dim", "int", "nn", "num_classes", "out_proj", "p", "pooler_dropout", "return", "self", "super", "tanh", "torch"], "led/modeling_led.py:LEDPreTrainedModel": ["Embedding", "Linear", "Model", "ModelConfig", "None", "PreTrainedModel", "True", "_init_weights", "attention_mask", "base_model_prefix", "bias", "class", "config", "data", "def", "device", "dummy_inputs", "elif", "if", "init_std", "input_ids", "is", "isinstance", "mean", "module", "ne", "nn", "normal_", "not", "pad_token", "pad_token_id", "padding_idx", "property", "return", "self", "std", "supports_gradient_checkpointing", "tensor", "torch", "weight", "zero_"], "led/modeling_led.py:LEDEncoderBaseModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "led/modeling_led.py:LEDSeq2SeqModelOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "led/modeling_led.py:LEDSeq2SeqLMOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "led/modeling_led.py:LEDSeq2SeqSequenceClassifierOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "led/modeling_led.py:LEDSeq2SeqQuestionAnsweringModelOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "end_logits", "loss", "past_key_values", "r", "start_logits", "torch", "tuple"], "led/modeling_led.py:LEDEncoder": ["Embedding", "Expected", "False", "Given", "Input", "LayerNorm", "Model", "ModelBaseModelOutput", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "The", "ValueError", "You", "__init__", "_merge_to_attention_mask", "_pad_to_window_size", "_prepare_4d_attention_mask_inverted", "a", "all_attentions", "all_global_attentions", "an", "and", "any", "are", "at", "attention_mask", "attention_window", "attentions", "automatically", "batch_size", "be", "both", "but", "cannot", "cat", "class", "config", "d_model", "def", "device", "dim", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "equal", "even", "f", "flatten", "for", "forward", "from", "functional", "given", "global_attention_mask", "global_attentions", "gradient_checkpointing", "has", "have", "head_mask", "hidden_states", "i", "ids", "idx", "if", "in", "input_ids", "input_ids_padding", "input_shape", "inputs_embeds", "inputs_embeds_padding", "int", "is", "is_global_attn", "is_index_global_attn", "is_index_masked", "isinstance", "it", "item", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "long", "max", "max_encoder_position_embeddings", "max_source_positions", "multiple", "new_full", "nn", "not", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad", "pad_token_id", "padded", "padding_idx", "padding_len", "positive", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "self", "seq_len", "shape", "should", "size", "specified", "specify", "state", "super", "the", "time", "to", "torch", "training", "transpose", "tuple", "use_return_dict", "v", "value", "view", "vocab_size", "warning_once"], "led/modeling_led.py:LEDDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "Model", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "The", "Transformers", "True", "ValueError", "You", "__init__", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask_inverted", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "at", "attention_mask", "attentions", "attn_mask", "be", "both", "but", "cache_position", "cannot", "checkpointing", "class", "combined_attention_mask", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "global_attention_mask", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "max_decoder_position_embeddings", "max_target_positions", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "self", "should", "size", "specified", "specify", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "led/modeling_led.py:LEDModel": ["Cache", "Embedding", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelEncoderBaseModelOutput", "ModelPreTrainedModel", "ModelSeq2SeqModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "global_attention_mask", "global_attentions", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "nn", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "shared", "shift_tokens_right", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "led/modeling_led.py:LEDForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelSeq2SeqLMOutput", "None", "Optional", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "global_attention_mask", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "led/modeling_led.py:LEDForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "FutureWarning", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelSeq2SeqSequenceClassifierOutput", "No", "None", "NotImplementedError", "Optional", "Passing", "Tensor", "The", "Transformers", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "actual", "and", "attention_mask", "auto_docstring", "be", "bool", "class", "classification", "classification_head", "classifier_dropout", "config", "cross_attentions", "cross_attn_head_mask", "currently", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "deprecated", "device", "dtype", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "global_attention_mask", "have", "head_mask", "hidden_states", "how", "if", "in", "input", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "len", "logits", "long", "loss", "loss_fct", "method", "multi_label_classification", "must", "not", "num_labels", "number", "of", "on", "or", "original", "output", "output_attentions", "output_hidden_states", "outputs", "paper", "past_key_values", "perform", "post_init", "problem_type", "provided", "r", "raise", "regression", "removed", "return", "return_dict", "same", "self", "sentence_representation", "sequence", "single_label_classification", "size", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "transformers", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "version", "view", "warn", "warnings", "weight", "were", "will"], "led/modeling_led.py:LEDForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "ModelSeq2SeqQuestionAnsweringModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_global_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "global_attention_mask", "head_mask", "hidden_size", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "weight"], "bloom/modeling_bloom.py:build_alibi_tensor": ["Model_alibi_tensor", "None", "Tensor", "alibi", "arange", "arange_tensor", "attention_mask", "base", "batch_size", "cat", "closest_power_of_2", "cumsum", "def", "device", "dim", "dtype", "extra_base", "extra_powers", "float32", "floor", "if", "int", "int32", "log2", "math", "min", "num_heads", "num_remaining_heads", "pow", "powers", "reshape", "return", "seq_length", "shape", "slopes", "tensor", "to", "torch"], "bloom/modeling_bloom.py:dropout_add": ["F", "Model", "Model_add", "Tensor", "bool", "def", "float", "out", "p", "prob", "residual", "return", "torch", "training", "x"], "bloom/modeling_bloom.py:bloom_gelu_forward": ["Model_gelu_forward", "Tensor", "def", "return", "tanh", "torch", "x"], "bloom/modeling_bloom.py:bloom_gelu_back": ["Model_gelu_back", "Tensor", "def", "ff", "g", "return", "tanh", "tanh_out", "torch", "x"], "bloom/modeling_bloom.py:GeLUFunction": ["Function", "GeLUFunction", "Model_gelu_back", "Model_gelu_forward", "Tensor", "autograd", "backward", "class", "ctx", "def", "forward", "grad_output", "input", "return", "save_for_backward", "saved_tensors", "staticmethod", "tmp", "torch"], "bloom/modeling_bloom.py:BloomGelu": ["GeLUFunction", "ModelGelu", "Model_gelu_forward", "Module", "Tensor", "__init__", "apply", "class", "def", "else", "forward", "if", "nn", "return", "self", "super", "torch", "training", "x"], "bloom/modeling_bloom.py:BloomAttention": ["Cache", "Dropout", "F", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_merge_heads", "_reshape", "a", "alibi", "alpha", "and", "attention_dropout", "attention_mask", "attention_probs", "attention_probs_reshaped", "attention_scores", "attn_weights", "baddbmm", "batch1", "batch2", "batch_size", "batch_size_and_num_heads", "be", "beta", "bias", "bmm", "bool", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "context_layer", "creating", "def", "dense", "dim", "divisible", "dropout_add", "dtype", "during", "else", "errors", "f", "float32", "for", "forward", "fused_qkv", "got", "head_dim", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "i", "if", "in", "int", "inv_norm_factor", "is", "key_layer", "layer_idx", "layer_past", "lead", "linear", "logger", "make", "math", "must", "n_head", "nn", "not", "num_heads", "output_attentions", "output_tensor", "passing", "permute", "pretraining_tp", "provide", "q_length", "query_key_value", "query_layer", "raise", "range", "recommended", "reshape", "residual", "return", "self", "seq_length", "shape", "slices", "slow_but_exact", "softmax", "split_size", "sqrt", "super", "sure", "the", "this", "three_times_hidden_size", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "value_layer", "view", "warning_once", "weight", "when", "will", "without", "x", "zeros_like"], "bloom/modeling_bloom.py:BloomMLP": ["F", "Linear", "ModelConfig", "ModelGelu", "ModelMLP", "Module", "Tensor", "__init__", "and", "class", "config", "def", "dense_4h_to_h", "dense_h_to_4h", "dropout_add", "else", "for", "forward", "gelu_impl", "hidden_dropout", "hidden_size", "hidden_states", "i", "if", "in", "int", "intermediate_output", "linear", "nn", "output", "pretraining_tp", "range", "residual", "return", "self", "shape", "slices", "slow_but_exact", "super", "torch", "training", "weight", "zeros_like"], "bloom/modeling_bloom.py:BloomBlock": ["Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelBlock", "ModelConfig", "ModelMLP", "None", "Optional", "Tensor", "__init__", "alibi", "apply_residual_connection_post_layernorm", "attention_mask", "attention_output", "attn_weights", "bool", "cache_position", "class", "config", "def", "else", "eps", "forward", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "layer_idx", "layer_norm_epsilon", "layer_past", "layernorm_output", "mlp", "n_head", "num_heads", "output", "output_attentions", "post_attention_layernorm", "residual", "return", "self", "self_attention", "super", "torch", "use_cache"], "bloom/modeling_bloom.py:BloomPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelBlock", "ModelConfig", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "__init__", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "inputs", "is", "isinstance", "kwargs", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "bloom/modeling_bloom.py:BloomModel": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FutureWarning", "Got", "LayerNorm", "LongTensor", "Model", "ModelBlock", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "alibi", "all_hidden_states", "all_self_attentions", "and", "any", "arange", "arguments", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "block", "bool", "build_alibi_tensor", "cache_position", "can", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "def", "deprecated_arguments", "device", "diagonal", "dim", "dtype", "else", "embed_dim", "enumerate", "eps", "exactly", "expand", "f", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "functionality", "get_head_mask", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "h", "have", "head_mask", "hidden_size", "hidden_states", "i", "if", "ignore", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layer_past", "len", "ln_f", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "n_head", "n_layer", "new_embeddings", "nn", "no", "not", "npu", "num_heads", "num_hidden_layers", "of", "one", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "passing", "past_key_values", "past_key_values_length", "past_length", "past_seen_tokens", "pop", "position_ids", "post_init", "r", "raise", "range", "removed", "reshape", "return", "return_dict", "safely", "sdpa", "self", "seq_length", "seq_length_with_past", "sequence_length", "set_input_embeddings", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unexpected", "use_cache", "use_return_dict", "using_compilable_cache", "v", "v5", "vocab_size", "warn", "warning_once", "warnings", "will", "with", "word_embeddings", "word_embeddings_layernorm", "xpu"], "bloom/modeling_bloom.py:BloomForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "FutureWarning", "GenerationMixin", "Got", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_tied_weights_keys", "and", "arguments", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bias", "bool", "cache_position", "can", "cat", "class", "clone", "config", "contiguous_format", "def", "deprecated_arguments", "device", "diff", "dim", "dtype", "elif", "else", "f", "for", "forward", "functionality", "get_max_cache_shape", "have", "head_mask", "hidden_size", "hidden_states", "if", "ignore", "in", "input_ids", "inputs_embeds", "is", "isinstance", "items", "key", "kwargs", "labels", "len", "lm_head", "lm_logits", "logits", "loss", "loss_function", "memory_format", "model_inputs", "new_attn_mask", "new_embeddings", "nn", "no", "not", "num_items_in_batch", "or", "output", "output_attentions", "output_hidden_states", "passing", "past_key_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "raise", "removed", "return", "return_dict", "safely", "self", "seq_length", "set_output_embeddings", "shape", "super", "target_length", "to", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "update", "use_cache", "use_return_dict", "v5", "value", "vocab_size", "warn", "warnings", "weight", "will", "zeros"], "bloom/modeling_bloom.py:BloomForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FutureWarning", "Got", "Linear", "LongTensor", "MSELoss", "Model", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "You", "__class__", "__init__", "__name__", "and", "arange", "argmax", "arguments", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "can", "class", "config", "conjunction", "def", "defined", "deprecated_arguments", "detect", "device", "dtype", "elif", "else", "f", "forward", "functionality", "handle", "have", "head_mask", "hidden_size", "hidden_states", "if", "ignore", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "len", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "passing", "past_key_values", "pooled_logits", "pop", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "removed", "return", "return_dict", "safely", "score", "self", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "v5", "warn", "warning_once", "warnings", "will", "with"], "bloom/modeling_bloom.py:BloomForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "False", "FutureWarning", "Got", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "ValueError", "You", "__init__", "and", "arguments", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "can", "class", "classifier", "classifier_dropout", "config", "def", "deprecated_arguments", "device", "dropout", "elif", "else", "f", "forward", "functionality", "hasattr", "have", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "ignore", "in", "input_ids", "inputs_embeds", "is", "labels", "len", "logits", "loss", "loss_fct", "nn", "no", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "passing", "past_key_values", "pop", "position_ids", "post_init", "r", "raise", "removed", "return", "return_dict", "safely", "self", "seq_length", "shape", "super", "to", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "v5", "view", "warn", "warnings", "will"], "bloom/modeling_bloom.py:BloomForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "helium/modeling_helium.py:HeliumRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "helium/modeling_helium.py:HeliumRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "helium/modeling_helium.py:HeliumMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "helium/modeling_helium.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "helium/modeling_helium.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "helium/modeling_helium.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "helium/modeling_helium.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dim", "k", "k_embed", "position_ids", "q", "q_embed", "repeat_interleave", "return", "rotate_half", "shape", "sin", "unsqueeze", "unsqueeze_dim"], "helium/modeling_helium.py:HeliumAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "math", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sqrt", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "helium/modeling_helium.py:HeliumDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "helium/modeling_helium.py:HeliumPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "helium/modeling_helium.py:HeliumModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "helium/modeling_helium.py:HeliumForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "helium/modeling_helium.py:HeliumForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "helium/modeling_helium.py:HeliumForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "musicgen/modeling_musicgen.py:MusicgenUnconditionalInput": ["FloatTensor", "LongTensor", "ModelOutput", "ModelUnconditionalInput", "None", "Optional", "attention_mask", "class", "encoder_outputs", "float", "guidance_scale", "r", "torch", "tuple"], "musicgen/modeling_musicgen.py:shift_tokens_right": ["Make", "Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "attribute", "clone", "configuration", "decoder_start_token_id", "def", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "of", "pad_token_id", "raise", "return", "s", "set", "shape", "sure", "the", "to", "torch", "transpose"], "musicgen/modeling_musicgen.py:MusicgenSinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "Tensor", "__init__", "arange", "bsz", "cat", "class", "codebooks", "cos", "def", "detach", "device", "dim", "dtype", "emb", "emb_weights", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "index_select", "input_ids", "int", "int64", "log", "make_weights", "math", "nn", "no_grad", "num_embeddings", "num_positions", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sin", "size", "staticmethod", "super", "to", "torch", "unsqueeze", "view", "weights", "zeros"], "musicgen/modeling_musicgen.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "musicgen/modeling_musicgen.py:MusicgenAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "class", "config", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "past_key_value", "past_key_values", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "musicgen/modeling_musicgen.py:MusicgenDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelDecoderConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bias", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "ffn_dim", "final_layer_norm", "forward", "functional", "hidden_size", "hidden_states", "if", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "musicgen/modeling_musicgen.py:MusicgenPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelAttention", "ModelDecoderConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_factor", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "musicgen/modeling_musicgen.py:MusicgenDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelDecoder", "ModelDecoderConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Size", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask", "_prepare_4d_causal_attention_mask_for_sdpa", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "at", "attention_mask", "attentions", "attn_implementation", "attn_mask", "auto_docstring", "be", "bool", "both", "bsz", "but", "cache_position", "cannot", "checkpointing", "class", "codebook", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_dim", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "flash_attention_2", "flex_attention", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "int", "is", "is_causal", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "make_flex_block_causal_mask", "mask_name", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "num_codebooks", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "query_length", "r", "raise", "random", "range", "removed", "reshape", "return", "return_dict", "same", "scale_embedding", "sdpa", "self", "seq_len", "shape", "should", "size", "specified", "specify", "sqrt", "sum", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "uniform", "use_cache", "use_return_dict", "v", "v4", "vocab_size", "warning_once", "will", "with", "zip"], "musicgen/modeling_musicgen.py:MusicgenModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LongTensor", "ModelDecoder", "ModelDecoderConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_outputs", "def", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "last_hidden_state", "not", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "value"], "musicgen/modeling_musicgen.py:MusicgenForCausalLM": ["BaseStreamer", "Cache", "CausalLMOutputWithCrossAttentions", "ClassifierFreeGuidanceLogitsProcessor", "CrossEntropyLoss", "Ensure", "False", "FloatTensor", "GREEDY_SEARCH", "GenerationConfig", "GenerationMixin", "GenerationMode", "Got", "Linear", "LogitsProcessorList", "LongTensor", "ModelDecoderConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "SAMPLE", "StoppingCriteriaList", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_decoder_start_token_tensor", "_expand_inputs_for_generation", "_get_logits_processor", "_get_stopping_criteria", "_pad_token_tensor", "_prepare_attention_mask_for_generation", "_prepare_cache_for_generation", "_prepare_generated_length", "_prepare_model_inputs", "_prepare_special_tokens", "_sample", "_validate_generated_length", "_validate_model_kwargs", "activated", "and", "append", "apply_delay_pattern_mask", "attention_mask", "attentions", "audio_channels", "auto_docstring", "batch_size", "be", "beam", "bias", "bool", "bos_token_id", "bsz", "build_delay_pattern_mask", "by", "cache_position", "channel_codebooks", "class", "codebook", "codebook_labels", "codebook_logits", "config", "contiguous", "copy", "cpu", "cross_attentions", "cross_attn_head_mask", "de", "decoder", "decoder_pad_token_mask", "deepcopy", "def", "delay_pattern", "delay_pattern_mask", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "encoder_input_ids", "encoder_outputs", "expand_size", "first_codebook_ids", "first_start_id", "for", "forward", "generate", "generation", "generation_config", "generation_mode", "get", "get_decoder", "get_generation_mode", "get_input_embeddings", "get_output_embeddings", "greedy", "guidance_scale", "has_default_max_length", "has_default_min_length", "head", "head_mask", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_ids_length", "input_ids_seq_length", "input_ids_shifted", "inputs", "inputs_embeds", "inputs_tensor", "int", "is", "is_encoder_decoder", "kwargs", "kwargs_has_attention_mask", "labels", "len", "lm_heads", "lm_logits", "logits", "logits_processor", "long", "loss", "loss_fct", "mask", "masked_fill", "max_cache_length", "max_length", "min", "min_length", "mode", "model", "model_input_name", "model_kwargs", "new_embeddings", "nn", "no_grad", "nonzero", "not", "num_beams", "num_codebooks", "num_return_sequences", "of", "one", "ones", "or", "output", "output_attentions", "output_hidden_states", "output_ids", "outputs", "pad_token_id", "past_key_values", "pattern_mask", "post_init", "prefix_allowed_tokens_fn", "prepare_inputs_for_generation", "put", "r", "raise", "range", "repeat", "repeat_interleave", "requires_attention_mask", "reshape", "return", "return_dict", "return_dict_in_generate", "sampling", "search", "self", "seq_len", "sequences", "set_decoder", "set_input_embeddings", "set_output_embeddings", "setting", "shape", "shift_tokens_right", "should", "stack", "start_ids", "staticmethod", "stopping_criteria", "streamer", "super", "synced_gpus", "that", "to", "torch", "tril", "triu", "tuple", "update", "use_cache", "use_return_dict", "validate", "value", "view", "vocab_size", "where", "zeros"], "musicgen/modeling_musicgen.py:MusicgenForConditionalGeneration": ["Any", "AutoConfig", "AutoModel", "AutoModelForTextEncoding", "BaseModelOutput", "BaseStreamer", "BoolTensor", "Cache", "ClassifierFreeGuidanceLogitsProcessor", "Config", "Cross", "Decoder", "Either", "EncoderDecoderModel", "Ensure", "Expected", "False", "FloatTensor", "GREEDY_SEARCH", "GenerationConfig", "GenerationMixin", "GenerationMode", "GitHub", "Got", "Head", "If", "In", "Initializing", "LM", "Linear", "LogitsProcessorList", "LongTensor", "Model", "ModelConfig", "ModelForCausalLM", "ModelForConditionalGeneration", "ModelPreTrainedModel", "ModelUnconditionalInput", "None", "NotImplementedError", "Optional", "Please", "PreTrainedModel", "Resizing", "SAMPLE", "Seq2SeqLMOutput", "StoppingCriteriaList", "Tensor", "The", "True", "Union", "ValueError", "__class__", "__init__", "_attn_implementation", "_bos_token_tensor", "_decoder_start_token_tensor", "_dynamic_tied_weights_keys", "_expand_inputs_for_generation", "_from_config", "_get_decoder_start_token_id", "_get_logits_processor", "_get_stopping_criteria", "_hf_hook", "_maybe_initialize_input_ids_for_generation", "_modules", "_pad_token_tensor", "_prepare_attention_mask_for_generation", "_prepare_audio_encoder_kwargs_for_generation", "_prepare_decoder_input_ids_for_generation", "_prepare_generated_length", "_prepare_model_inputs", "_prepare_special_tokens", "_prepare_text_encoder_kwargs_for_generation", "_requires_grad", "_sample", "_tie_encoder_decoder_weights", "_validate_model_kwargs", "a", "activated", "add_cross_attention", "added", "all", "allows", "an", "and", "any", "append", "apply_delay_pattern_mask", "architecture", "are", "args", "argument", "as", "attention", "attention_mask", "attentions", "attributes", "audio", "audio_channels", "audio_codes", "audio_codes_left", "audio_codes_right", "audio_encoder", "audio_encoder_", "audio_encoder_model", "audio_encoder_outputs", "audio_encoder_outputs_left", "audio_encoder_outputs_right", "audio_encoder_pretrained_model_name_or_path", "audio_scales", "audio_scales_left", "audio_scales_right", "audio_values", "auto_docstring", "base_model_prefix", "batch_size", "be", "beam", "bool", "bos_token_id", "break", "bsz", "build_delay_pattern_mask", "but", "by", "cache_position", "cat", "causal", "channel", "channels", "chunk_length", "chunking", "class", "classmethod", "cls", "code", "codebooks", "codec_outputs_left", "codec_outputs_right", "com", "concatenate", "config", "config_class", "configuration", "copy", "cpu", "cross", "cross_attention_hidden_size", "cross_attentions", "cross_attn", "cross_attn_head_mask", "de", "decode", "decoder", "decoder_", "decoder_attention_mask", "decoder_attentions", "decoder_base_model_prefix", "decoder_config", "decoder_delay_pattern_mask", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_input_ids_start", "decoder_inputs_embeds", "decoder_model", "decoder_outputs", "decoder_pretrained_model_name_or_path", "decoder_signature", "decoder_start_token_id", "deepcopy", "def", "defined", "del", "device", "dict", "dim", "directly", "disabled", "discussion", "do", "dtype", "elif", "else", "embedding", "enc_to_dec_proj", "encode", "encoder", "encoder_accepts_wildcard", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_decoder", "encoder_hidden_states", "encoder_input_ids", "encoder_kwargs", "encoder_last_hidden_state", "encoder_outputs", "encoder_signature", "equal", "example", "expand_size", "f", "following", "for", "forward", "frame", "frames", "freeze_audio_encoder", "freeze_text_encoder", "from", "from_config", "from_pretrained", "from_sub_models_config", "from_sub_models_pretrained", "generate", "generation", "generation_config", "generation_mode", "get", "get_audio_encoder", "get_encoder", "get_generation_mode", "get_input_embeddings", "get_output_embeddings", "get_text_encoder", "get_unconditional_inputs", "github", "got", "greedy", "guidance_scale", "has", "has_default_max_length", "has_default_min_length", "hasattr", "have", "head_mask", "hidden", "hidden_size", "hidden_states", "https", "huggingface", "if", "in", "incompatible", "info", "initialize", "initialized", "input_ids", "input_ids_length", "input_ids_seq_length", "input_values", "inputs", "inputs_embeds", "inputs_tensor", "inspect", "int", "io_same_device", "irrelevant_prefix", "is", "is_decoder", "is_encoder_decoder", "isinstance", "issues", "it", "item", "items", "key", "keys", "kwargs", "kwargs_audio_encoder", "kwargs_decoder", "kwargs_has_attention_mask", "kwargs_text_encoder", "labels", "last_hidden_state", "layers", "len", "list", "logger", "logits", "logits_processor", "long", "loss", "main_input_name", "make", "mask", "max_length", "methods", "min_length", "mode", "model", "model_args", "model_input_name", "model_kwargs", "new_embeddings", "new_ones", "nn", "no", "no_grad", "not", "num_beams", "num_codebooks", "num_return_sequences", "num_samples", "objects", "of", "on", "one", "ones", "ones_like", "or", "order", "output_attentions", "output_hidden_states", "output_ids", "output_values", "output_values_left", "output_values_right", "outputs", "overwritten", "p", "pad_token_id", "padding_mask", "param", "parameters", "pass", "passed", "past_key_values", "pop", "prefix_allowed_tokens_fn", "prepare_decoder_input_ids_from_labels", "prepare_inputs_for_generation", "prepared", "provided", "put", "r", "raise", "randomly", "repeat", "repeat_interleave", "requires_attention_mask", "requires_grad", "reshape", "resize_token_embeddings", "respective", "return", "return_dict", "return_dict_in_generate", "return_unused_kwargs", "s", "sampling", "search", "see", "selected", "self", "seq_len", "sequences", "set", "set_output_embeddings", "setting", "shape", "shared", "shift_tokens_right", "should", "signature", "size", "specified", "stack", "startswith", "states", "stereo", "stopping_criteria", "str", "streamer", "super", "supported", "supports_gradient_checkpointing", "sure", "synced_gpus", "text", "text_encoder", "text_encoder_", "text_encoder_model", "text_encoder_pretrained_model_name_or_path", "that", "the", "three", "tie_encoder_decoder", "tie_weights", "tied_weights", "to", "to_dict", "torch", "transformers", "tuple", "type", "update", "use", "use_cache", "use_return_dict", "validate", "value", "values", "via", "warning", "when", "without", "wrapped", "zeros", "zeros_like"], "roc_bert/modeling_roc_bert.py:RoCBertEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Linear", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "and", "arange", "batch_size", "buffered_token_type_ids", "cat", "class", "clone", "concat_input", "config", "def", "denominator", "device", "dim", "dropout", "dtype", "else", "embedding_in", "embedding_pronunciation", "embedding_shape", "embeddings", "enable_pronunciation", "enable_shape", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "index", "input_dim", "input_ids", "input_pronunciation_ids", "input_shape", "input_shape_ids", "inputs_embeds", "is", "layer_norm_eps", "long", "map_inputs_layer", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "pronunciation_dim", "pronunciation_embed", "pronunciation_embed_dim", "pronunciation_vocab_size", "register_buffer", "return", "self", "seq_length", "shape", "shape_dim", "shape_embed", "shape_embed_dim", "shape_vocab_size", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "roc_bert/modeling_roc_bert.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "roc_bert/modeling_roc_bert.py:RoCBertSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "roc_bert/modeling_roc_bert.py:RoCBertCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "roc_bert/modeling_roc_bert.py:RoCBertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "roc_bert/modeling_roc_bert.py:RoCBertAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "roc_bert/modeling_roc_bert.py:RoCBertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "roc_bert/modeling_roc_bert.py:RoCBertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "roc_bert/modeling_roc_bert.py:RoCBertLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "roc_bert/modeling_roc_bert.py:RoCBertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "roc_bert/modeling_roc_bert.py:RoCBertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "roc_bert/modeling_roc_bert.py:RoCBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "roc_bert/modeling_roc_bert.py:RoCBertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "roc_bert/modeling_roc_bert.py:RoCBertOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "roc_bert/modeling_roc_bert.py:RoCBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLMPredictionHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "roc_bert/modeling_roc_bert.py:RoCBertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_pronunciation_embeddings", "get_seq_length", "get_shape_embeddings", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_pronunciation_ids", "input_shape", "input_shape_ids", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "pronunciation_embed", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "set_pronunciation_embeddings", "set_shape_embeddings", "shape", "shape_embed", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "roc_bert/modeling_roc_bert.py:RoCBertForPreTraining": ["CrossEntropyLoss", "MaskedLMOutput", "Model", "ModelForPreTraining", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "T", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "_", "__init__", "_tied_weights_keys", "attack_attention_mask", "attack_input_ids", "attack_input_pronunciation_ids", "attack_input_shape_ids", "attack_output", "attack_pooled_output", "attack_pooled_output_norm", "attack_token_type_ids", "attention_mask", "attentions", "auto_docstring", "batch_labels", "batch_size", "bias", "can_return_tuple", "class", "clone", "cls", "config", "contrastive_loss", "decoder", "def", "device", "dim", "else", "forward", "functional", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_pronunciation_ids", "input_shape_ids", "inputs_embeds", "is", "kwargs", "labels_attention_mask", "labels_input_ids", "labels_input_pronunciation_ids", "labels_input_shape_ids", "labels_output", "labels_pooled_output", "labels_pooled_output_norm", "labels_token_type_ids", "list", "logits", "loss", "loss_fct", "masked_lm_loss", "matmul", "new_embeddings", "nn", "normalize", "not", "outputs", "pad_token_id", "pooled_output", "pooled_output_norm", "position_ids", "post_init", "prediction_scores", "predictions", "r", "range", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "sim_matrix", "sim_matrix_target", "super", "target_inputs", "tensor", "token_type_ids", "torch", "tuple", "view", "vocab_size", "weight"], "roc_bert/modeling_roc_bert.py:RoCBertForMaskedLM": ["CrossEntropyLoss", "False", "If", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bi", "bias", "bool", "can_generate", "can_return_tuple", "cat", "class", "classmethod", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_pronunciation_ids", "input_shape", "input_shape_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "roc_bert/modeling_roc_bert.py:RoCBertForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "If", "Model", "ModelForCausalLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "RoCModelForCausalLM", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "cls", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_pronunciation_ids", "input_shape_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "list", "lm_loss", "logger", "logits", "loss", "loss_function", "model_inputs", "model_kwargs", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "roc_bert/modeling_roc_bert.py:RoCBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "input_pronunciation_ids", "input_shape_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "roc_bert/modeling_roc_bert.py:RoCBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "input_pronunciation_ids", "input_shape_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "roc_bert/modeling_roc_bert.py:RoCBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "input_pronunciation_ids", "input_shape_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "view"], "roc_bert/modeling_roc_bert.py:RoCBertForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "input_pronunciation_ids", "input_shape_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "bitnet/modeling_bitnet.py:BitNetRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "bitnet/modeling_bitnet.py:BitNetMLP": ["ACT2FN", "False", "Linear", "ModelConfig", "ModelMLP", "ModelRMSNorm", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "eps", "ffn_sub_norm", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "rms_norm_eps", "self", "super", "up_proj", "x"], "bitnet/modeling_bitnet.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "bitnet/modeling_bitnet.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "bitnet/modeling_bitnet.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "bitnet/modeling_bitnet.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "bitnet/modeling_bitnet.py:BitNetAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_sub_norm", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "bitnet/modeling_bitnet.py:BitNetDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "bitnet/modeling_bitnet.py:BitNetRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "bitnet/modeling_bitnet.py:BitNetPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "bitnet/modeling_bitnet.py:BitNetModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "bitnet/modeling_bitnet.py:BitNetForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "mask2former/modeling_mask2former.py:Mask2FormerPixelDecoderOutput": ["FloatTensor", "ModelOutput", "ModelPixelDecoderOutput", "Model_features", "None", "Optional", "attentions", "class", "multi_scale_features", "r", "torch", "tuple"], "mask2former/modeling_mask2former.py:Mask2FormerMaskedAttentionDecoderOutput": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "ModelModeledAttentionDecoderOutput", "Models_queries_logits", "None", "Optional", "attentions", "class", "hidden_states", "intermediate_hidden_states", "last_hidden_state", "r", "torch", "tuple"], "mask2former/modeling_mask2former.py:Mask2FormerPixelLevelModuleOutput": ["FloatTensor", "ModelOutput", "ModelPixelLevelModuleOutput", "None", "Optional", "class", "decoder_hidden_states", "decoder_last_hidden_state", "encoder_hidden_states", "encoder_last_hidden_state", "r", "torch", "tuple"], "mask2former/modeling_mask2former.py:Mask2FormerModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "Models_queries_logits", "None", "Optional", "attentions", "class", "encoder_hidden_states", "encoder_last_hidden_state", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "r", "torch", "transformer_decoder_hidden_states", "transformer_decoder_intermediate_states", "transformer_decoder_last_hidden_state", "tuple"], "mask2former/modeling_mask2former.py:Mask2FormerForUniversalSegmentationOutput": ["FloatTensor", "ModelForUniversalSegmentationOutput", "ModelOutput", "Models_queries_logits", "None", "Optional", "attentions", "auxiliary_logits", "class", "class_queries_logits", "dict", "encoder_hidden_states", "encoder_last_hidden_state", "list", "loss", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "r", "str", "torch", "transformer_decoder_hidden_states", "transformer_decoder_last_hidden_state", "tuple"], "mask2former/modeling_mask2former.py:sample_point": ["False", "Model_point", "Tensor", "True", "add_dim", "def", "dim", "functional", "grid_Model", "if", "input_features", "kwargs", "nn", "point_coordinates", "point_features", "return", "squeeze", "torch", "unsqueeze"], "mask2former/modeling_mask2former.py:dice_loss": ["Model_loss", "Tensor", "def", "denominator", "flatten", "inputs", "int", "labels", "loss", "num_masks", "numerator", "probs", "r", "return", "sigmoid", "sum"], "mask2former/modeling_mask2former.py:sigmoid_cross_entropy_loss": ["BCEWithLogitsLoss", "Model_cross_entropy_loss", "Tensor", "criterion", "cross_entropy_loss", "def", "inputs", "int", "labels", "loss", "mean", "nn", "none", "num_masks", "r", "reduction", "return", "sum", "torch"], "mask2former/modeling_mask2former.py:pair_wise_dice_loss": ["Model_wise_dice_loss", "None", "T", "Tensor", "def", "denominator", "flatten", "inputs", "labels", "loss", "matmul", "numerator", "return", "sigmoid", "sum", "torch"], "mask2former/modeling_mask2former.py:pair_wise_sigmoid_cross_entropy_loss": ["BCEWithLogitsLoss", "Model_wise_sigmoid_cross_entropy_loss", "T", "Tensor", "criterion", "cross_entropy_loss_neg", "cross_entropy_loss_pos", "def", "height_and_width", "inputs", "labels", "loss", "loss_neg", "loss_pos", "matmul", "nn", "none", "ones_like", "r", "reduction", "return", "shape", "torch", "zeros_like"], "mask2former/modeling_mask2former.py:Mask2FormerHungarianMatcher": ["All", "False", "ModelHungarianMatcher", "Model_labels", "Models_queries_logits", "Module", "None", "Tensor", "ValueError", "__init__", "align_corners", "and", "append", "array", "as_tensor", "assigned_indices", "batch_size", "be", "can", "class", "class_labels", "class_queries_logits", "cost_Model", "cost_class", "cost_dice", "cost_matrix", "costs", "cpu", "def", "device", "dtype", "float", "for", "forward", "i", "if", "in", "indices", "int", "int64", "j", "linear_sum_assignment", "list", "matched_indices", "maximum", "minimum", "nan_to_num", "nn", "no_grad", "np", "num_points", "pair_wise_dice_loss", "pair_wise_sigmoid_cross_entropy_loss", "point_coordinates", "pred_Model", "pred_coordinates", "pred_probs", "raise", "rand", "range", "repeat", "return", "sample_point", "self", "shape", "softmax", "squeeze", "super", "t", "target_Model", "target_coordinates", "tensor", "to", "torch", "tuple"], "mask2former/modeling_mask2former.py:Mask2FormerLoss": ["CrossEntropyLoss", "False", "ModelConfig", "ModelHungarianMatcher", "ModelLoss", "Model_labels", "Model_weight", "Models_queries_logits", "Module", "None", "Optional", "PartialState", "Tensor", "_", "__init__", "_get_predictions_permutation_indices", "_get_targets_permutation_indices", "_max_by_axis", "_pad_images_to_max_in_batch", "_shared_state", "abs", "align_corners", "arange", "array", "as_tensor", "aux_outputs", "auxiliary_predictions", "batch_indices", "batch_shape", "batch_size", "bool", "calculate_uncertainty", "cat", "clamp", "class", "class_labels", "class_queries_logits", "class_weight", "classes", "config", "copy_", "cost_Model", "cost_class", "cost_dice", "criterion", "def", "del", "device", "dice_loss", "dice_weight", "dict", "dim", "dtype", "empty_weight", "enumerate", "eos_coef", "f", "fill_value", "float", "for", "forward", "full", "full_like", "get_num_Models", "height", "i", "idx", "if", "importance_sample_ratio", "in", "index", "indices", "int", "int64", "is", "is_accelerate_available", "item", "items", "j", "k", "key", "lambda", "len", "list", "logits", "long", "loss_Model", "loss_Models", "loss_ce", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_labels", "losses", "matcher", "max", "max_size", "maxes", "min", "nn", "no_grad", "no_object_weight", "not", "np", "num_Models", "num_boxes", "num_labels", "num_points", "num_points_sampled", "num_processes", "num_queries", "num_random_points", "num_uncertain_points", "ones", "oversample_ratio", "padded_tensor", "padded_tensors", "padding_Model", "padding_Models", "point_coordinates", "point_labels", "point_logits", "point_uncertainties", "pred_Models", "pred_logits", "pred_logits_transposed", "predictions_indices", "rand", "reduce", "register_buffer", "requires_backends", "return", "sample_point", "sample_points_using_uncertainty", "scipy", "self", "shape", "shift", "sigmoid_cross_entropy_loss", "sizes", "squeeze", "src", "src_idx", "str", "sublist", "sum", "super", "target", "target_Models", "target_classes", "target_classes_o", "target_indices", "tensor", "tensors", "tgt", "tgt_idx", "topk", "torch", "train_num_points", "transpose", "tuple", "uncertainty_function", "uncertainty_scores", "update", "value", "view", "weight", "weight_dict", "width", "with", "world_size", "zeros", "zip"], "mask2former/modeling_mask2former.py:multi_scale_deformable_attention": ["False", "Model_scale_deformable_attention", "Tensor", "Union", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "contiguous", "def", "dim", "enumerate", "flatten", "for", "functional", "grid_sample", "height", "hidden_dim", "in", "level_id", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "view", "width", "zeros"], "mask2former/modeling_mask2former.py:Mask2FormerSinePositionEmbedding": ["False", "Model", "ModelSinePositionEmbedding", "Module", "None", "Optional", "Size", "Tensor", "True", "Union", "ValueError", "__init__", "and", "arange", "be", "bool", "cat", "class", "compile_compatible_method_lru_cache", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "else", "eps", "flatten", "float", "floor", "forward", "if", "int", "int64", "is", "math", "maxsize", "nn", "normalize", "not", "not_Model", "num_pos_feats", "passed", "permute", "pi", "pos", "pos_x", "pos_y", "raise", "return", "rounding_mode", "scale", "self", "shape", "should", "sin", "stack", "str", "super", "temperature", "to", "torch", "x_embed", "y_embed", "zeros"], "mask2former/modeling_mask2former.py:Mask2FormerPixelDecoderEncoderMultiscaleDeformableAttention": ["CUDA", "DeformableDetrMultiscaleDeformableAttention", "False", "Last", "Linear", "Make", "ModelPixelDecoderEncoderMultiscaleDeformableAttention", "Modeled_fill", "Module", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_Model", "attention_weights", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "d", "d_model", "def", "device", "dim", "dim_per_head", "dimension", "divisible", "dtype", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_Model", "encoder_hidden_states", "f", "float", "for", "forward", "functional", "got", "head", "height", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "long", "make", "more", "multi_scale_deformable_attention", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes_list", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "total_elements", "value", "value_proj", "view", "warn", "warnings", "which", "width", "with", "with_pos_embed"], "mask2former/modeling_mask2former.py:Mask2FormerPixelDecoderEncoderLayer": ["False", "LayerNorm", "Linear", "ModelConfig", "ModelPixelDecoderEncoderLayer", "ModelPixelDecoderEncoderMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "any", "attention_Model", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "def", "dropout", "dtype", "embed_dim", "encoder_attention_Model", "encoder_feedforward_dim", "encoder_hidden_states", "fc1", "fc2", "feature_size", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "level_start_index", "max", "min", "n_levels", "n_points", "nn", "num_attention_heads", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "relu", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "spatial_shapes_list", "super", "torch", "training", "transpose"], "mask2former/modeling_mask2former.py:Mask2FormerPixelDecoderEncoderOnly": ["BaseModelOutput", "ModelConfig", "ModelPixelDecoderEncoderLayer", "ModelPixelDecoderEncoderOnly", "Module", "ModuleList", "None", "_", "__init__", "all_attentions", "all_hidden_states", "append", "attention_Model", "attentions", "cat", "class", "config", "def", "device", "dropout", "dtype", "else", "encoder_layer", "encoder_layers", "enumerate", "for", "forward", "get_reference_points", "height", "hidden_states", "i", "if", "ij", "in", "indexing", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "level_start_index", "linspace", "lvl", "meshgrid", "nn", "not", "output_attentions", "output_hidden_states", "position_embeddings", "r", "range", "ref", "ref_x", "ref_y", "reference_points", "reference_points_list", "reshape", "return", "return_dict", "self", "spatial_shapes_list", "stack", "staticmethod", "super", "torch", "transpose", "use_return_dict", "valid_ratios", "width"], "mask2former/modeling_mask2former.py:Mask2FormerPixelDecoder": ["Conv2d", "False", "GroupNorm", "Model", "ModelConfig", "ModelPixelDecoder", "ModelPixelDecoderEncoderOnly", "ModelPixelDecoderOutput", "ModelSinePositionEmbedding", "Model_dim", "Model_feature_size", "Model_features", "Model_projection", "Models", "Models_flat", "Module", "ModuleList", "None", "Parameter", "ReLU", "Sequential", "Tensor", "True", "_", "__init__", "adapter_", "add_module", "align_corners", "append", "as_tensor", "attention_Model", "attentions", "batch_size", "bias", "bilinear", "bool", "cat", "class", "common_stride", "config", "cumsum", "current_fpn", "def", "device", "dim", "dtype", "else", "embed", "encoder", "encoder_output", "encoder_outputs", "enumerate", "f", "feature", "feature_channels", "feature_dim", "feature_size", "feature_strides", "features", "flatten", "float32", "for", "forward", "functional", "get_valid_ratio", "height", "i", "idx", "if", "in", "in_channels", "input_embeds", "input_embeds_flat", "input_projections", "input_projections_list", "inputs_embeds", "int", "interpolate", "is", "kernel_size", "last_hidden_state", "lateral_conv", "lateral_convolutions", "lateral_convs", "layer_", "level", "level_embed", "level_pos_embed_flat", "level_start_index", "level_start_index_list", "log2", "long", "min", "mode", "multi_scale_features", "new_zeros", "nn", "normalize", "not", "np", "num_cur_levels", "num_feature_levels", "num_fpn_levels", "num_pos_feats", "num_pos_features", "out", "output_attentions", "output_conv", "output_convolutions", "output_convs", "output_hidden_states", "outputs", "padding", "position_embedding", "position_embeddings", "prod", "range", "return", "return_dict", "self", "shape", "size", "spatial_shapes", "spatial_shapes_list", "split", "split_sizes", "stack", "stride", "sum", "super", "to", "torch", "transformer_feature_strides", "transformer_in_channels", "transpose", "tuple", "valid_height", "valid_ratio", "valid_ratio_height", "valid_ratio_width", "valid_ratios", "valid_width", "view", "width", "x", "zeros"], "mask2former/modeling_mask2former.py:Mask2FormerPixelLevelModule": ["False", "ModelConfig", "ModelPixelDecoder", "ModelPixelLevelModule", "ModelPixelLevelModuleOutput", "Model_features", "Module", "None", "Tensor", "__init__", "backbone_features", "bool", "channels", "class", "config", "decoder", "decoder_hidden_states", "decoder_last_hidden_state", "decoder_output", "def", "else", "encoder", "encoder_hidden_states", "encoder_last_hidden_state", "feature_channels", "feature_maps", "forward", "if", "load_backbone", "multi_scale_features", "nn", "output_hidden_states", "pixel_values", "return", "self", "super", "tuple"], "mask2former/modeling_mask2former.py:Mask2FormerAttention": ["Attention", "False", "Linear", "Model", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_shape", "and", "attention_Model", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "int", "is", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_position_embeddings", "key_value_states", "key_value_states_original", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "permute", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed"], "mask2former/modeling_mask2former.py:Mask2FormerMaskedAttentionDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelModeledAttentionDecoderLayer", "MultiheadAttention", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_Model", "attn_Model", "bool", "class", "config", "cross_attn", "cross_attn_layer_norm", "cross_attn_weights", "def", "dim_feedforward", "dropout", "else", "embed_dim", "encoder_attention_Model", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "forward_post", "forward_pre", "functional", "hidden_dim", "hidden_states", "if", "int", "is", "is_decoder", "key", "key_padding_Model", "level_index", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "p", "pos", "position_embeddings", "pre_norm", "query", "query_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "tensor", "torch", "training", "value", "with_pos_embed"], "mask2former/modeling_mask2former.py:Mask2FormerMaskedAttentionDecoder": ["False", "LayerNorm", "ModelConfig", "ModelModelPredictor", "ModelModeledAttentionDecoder", "ModelModeledAttentionDecoderLayer", "ModelModeledAttentionDecoderOutput", "Model_feature_size", "Model_predictor", "Models_queries_logits", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_hidden_states", "and", "attention_Model", "attentions", "bool", "class", "config", "continue", "decoder_layer", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_attention_Model", "encoder_hidden_states", "enumerate", "feature_size_list", "for", "forward", "gradient_checkpointing", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "inputs_embeds", "intermediate", "intermediate_Model_predictions", "intermediate_hidden_states", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layernorm", "layers", "level_index", "list", "multi_stage_positional_embeddings", "nn", "not", "num_attention_heads", "num_feature_levels", "num_heads", "output_attentions", "output_hidden_states", "outputs", "pixel_embeddings", "predicted_Model", "query_position_embeddings", "r", "rand", "range", "return", "return_dict", "self", "shape", "sum", "super", "to", "torch", "training", "transpose", "tuple", "unsqueeze", "use_return_dict", "v", "where"], "mask2former/modeling_mask2former.py:Mask2FormerPredictionBlock": ["Linear", "ModelPredictionBlock", "Module", "None", "Tensor", "__init__", "activation", "add_module", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_dim", "input", "int", "layer", "layers", "nn", "out_dim", "return", "self", "str", "super"], "mask2former/modeling_mask2former.py:Mask2FormerMLPPredictionHead": ["Identity", "ModelMLPPredictionHead", "ModelPredictionBlock", "Module", "ReLU", "Tensor", "__init__", "activation", "add_module", "append", "class", "def", "else", "enumerate", "for", "forward", "hidden_dim", "hidden_state", "i", "if", "in", "in_dim", "in_dims", "input", "input_dim", "int", "layer", "layers", "nn", "num_layers", "out_dim", "out_dims", "output_dim", "return", "self", "str", "super", "zip"], "mask2former/modeling_mask2former.py:Mask2FormerMaskPredictor": ["False", "ModelMLPPredictionHead", "ModelModelPredictor", "Model_embedder", "Model_embeddings", "Model_feature_size", "Module", "None", "Optional", "Tensor", "__init__", "align_corners", "attention_Model", "attention_Model_target_size", "bchw", "bilinear", "bool", "bqc", "bqhw", "class", "def", "detach", "einsum", "flatten", "forward", "functional", "hidden_size", "int", "interpolate", "mode", "nn", "num_heads", "outputs", "outputs_Model", "pixel_embeddings", "repeat", "return", "self", "sigmoid", "size", "super", "torch", "transpose", "unsqueeze"], "mask2former/modeling_mask2former.py:Mask2FormerTransformerModule": ["Conv2d", "Embedding", "False", "ModelConfig", "ModelModeledAttentionDecoder", "ModelModeledAttentionDecoderOutput", "ModelSinePositionEmbedding", "ModelTransformerModule", "Model_features", "Module", "None", "Sequential", "Tensor", "True", "_", "__init__", "append", "batch_size", "bool", "class", "config", "decoder", "decoder_output", "def", "device", "dtype", "else", "encoder_hidden_states", "enforce_input_projection", "feature_size_list", "flatten", "for", "forward", "hidden_dim", "i", "if", "in", "in_features", "input_projections", "inputs_embeds", "int", "kernel_size", "level_embed", "list", "multi_scale_features", "multi_stage_features", "multi_stage_positional_embeddings", "nn", "normalize", "num_feature_levels", "num_pos_feats", "num_queries", "or", "output_attentions", "output_hidden_states", "permute", "pixel_embeddings", "position_embedder", "queries_embedder", "queries_features", "query_embeddings", "query_features", "query_position_embeddings", "range", "repeat", "return", "return_dict", "self", "shape", "size_list", "super", "unsqueeze", "weight"], "mask2former/modeling_mask2former.py:Mask2FormerPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "GroupNorm", "LayerNorm", "Linear", "ModelConfig", "ModelModeledAttentionDecoderLayer", "ModelPixelDecoder", "ModelPixelDecoderEncoderMultiscaleDeformableAttention", "ModelPreTrainedModel", "ModelTransformerModule", "Module", "None", "Parameter", "PreTrainedModel", "Sequential", "True", "_init_weights", "abs", "arange", "attention_weights", "base_model_prefix", "bias", "class", "config", "constant_", "cos", "cross_attn", "data", "def", "dim", "dtype", "elif", "fill_", "float", "for", "gain", "grid_init", "hasattr", "i", "if", "in", "in_proj_bias", "init", "init_std", "init_xavier_std", "input_projection", "input_projections", "int64", "is", "isinstance", "keepdim", "level_embed", "main_input_name", "math", "max", "mean", "model", "module", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "output_proj", "p", "padding_idx", "parameters", "pi", "pixel_values", "range", "reference_points", "repeat", "sampling_offsets", "self", "sin", "stack", "std", "thetas", "torch", "value_proj", "view", "weight", "with", "xavier_std", "xavier_uniform_", "zero_"], "mask2former/modeling_mask2former.py:Mask2FormerModel": ["ModelConfig", "ModelModel", "ModelModelOutput", "ModelPixelLevelModule", "ModelPreTrainedModel", "ModelTransformerModule", "Model_features", "Models_queries_logits", "None", "Optional", "Tensor", "True", "_", "__init__", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "decoder_hidden_states", "decoder_last_hidden_state", "def", "device", "else", "encoder_hidden_states", "encoder_last_hidden_state", "feature_size", "for", "forward", "height", "hidden_states", "if", "in", "in_features", "intermediate_hidden_states", "is", "last_hidden_state", "main_input_name", "multi_scale_features", "not", "ones", "output", "output_attentions", "output_hidden_states", "pixel_Model", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "pixel_level_module", "pixel_level_module_output", "pixel_values", "post_init", "return", "return_dict", "self", "shape", "super", "torch", "transformer_decoder_hidden_states", "transformer_decoder_intermediate_states", "transformer_decoder_last_hidden_state", "transformer_module", "transformer_module_output", "tuple", "use_return_dict", "v", "values", "width"], "mask2former/modeling_mask2former.py:Mask2FormerForUniversalSegmentation": ["Linear", "ModelConfig", "ModelForUniversalSegmentation", "ModelForUniversalSegmentationOutput", "ModelLoss", "ModelModel", "ModelPreTrainedModel", "Model_labels", "Model_weight", "Models_queries_logits", "None", "Optional", "Tensor", "True", "__init__", "and", "append", "attentions", "auto_docstring", "aux_binary_Models", "aux_classes", "auxiliary_logits", "auxiliary_predictions", "bool", "class", "class_labels", "class_prediction", "class_predictor", "class_queries_logits", "class_weight", "classes", "config", "criterion", "decoder_output", "def", "dice_weight", "dict", "else", "encoder_hidden_states", "encoder_last_hidden_state", "float", "for", "forward", "get_auxiliary_logits", "get_loss", "get_loss_dict", "hidden_dim", "if", "in", "is", "items", "key", "list", "loss", "loss_Model", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_key", "main_input_name", "model", "nn", "not", "num_labels", "or", "output", "output_Models", "output_attentions", "output_auxiliary_logits", "output_hidden_states", "outputs", "pixel_Model", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "pixel_values", "post_init", "r", "return", "return_dict", "self", "str", "sum", "super", "torch", "transformer_decoder_hidden_states", "transformer_decoder_intermediate_states", "transformer_decoder_last_hidden_state", "transpose", "tuple", "use_auxiliary_loss", "use_return_dict", "v", "values", "weight", "weight_dict", "zip"], "granitemoe/modeling_granitemoe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "granitemoe/modeling_granitemoe.py:GraniteMoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "granitemoe/modeling_granitemoe.py:GraniteMoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "granitemoe/modeling_granitemoe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "granitemoe/modeling_granitemoe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "granitemoe/modeling_granitemoe.py:GraniteMoeParallelExperts": ["F", "ModelParallelExperts", "Module", "None", "Parameter", "__init__", "append", "cat", "class", "def", "dim", "empty", "expert_size", "for", "forward", "i", "in", "input_list", "input_size", "inputs", "int", "linear", "nn", "num_experts", "output_list", "output_size", "range", "results", "return", "self", "split", "super", "torch", "weight"], "granitemoe/modeling_granitemoe.py:GraniteMoeTopKGating": ["False", "Linear", "ModelTopKGating", "Module", "_", "__init__", "batch_gates", "batch_index", "bias", "class", "def", "device", "dim", "div", "dtype", "expert_size", "flatten", "float", "forward", "gates", "hidden_states", "index_sorted_experts", "input_size", "int", "layer", "logits", "long", "nn", "num_experts", "return", "rounding_mode", "scatter", "self", "size", "softmax", "sort", "sum", "super", "tolist", "top_k", "top_k_experts", "top_k_gates", "top_k_indices", "top_k_logits", "topk", "torch", "trunc", "type_as", "zeros"], "granitemoe/modeling_granitemoe.py:GraniteMoeMoE": ["ACT2FN", "ModelConfig", "ModelMoE", "ModelParallelExperts", "ModelTopKGating", "Module", "None", "_", "__init__", "activation", "batch_gates", "batch_index", "bsz", "chunk", "chunked_hidden_states", "class", "config", "def", "device", "dim", "dtype", "emb_size", "expert_inputs", "expert_outputs", "expert_size", "forward", "hidden_act", "hidden_size", "hidden_states", "index_add", "input_linear", "input_size", "intermediate_size", "layer_input", "layer_output", "length", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "output_linear", "reshape", "return", "router", "router_logits", "self", "size", "super", "top_k", "torch", "view", "zeros"], "granitemoe/modeling_granitemoe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "granitemoe/modeling_granitemoe.py:GraniteMoeAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_multiplier", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "errors", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "return", "scaling", "self", "sin", "size", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "granitemoe/modeling_granitemoe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "granitemoe/modeling_granitemoe.py:GraniteMoeDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "block_sparse_moe", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "new_name", "num_local_experts", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "residual_multiplier", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "granitemoe/modeling_granitemoe.py:GraniteMoePreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelParallelExperts", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "data", "def", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "granitemoe/modeling_granitemoe.py:GraniteMoeModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "embedding_multiplier", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "head_dim", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "list", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "max_position_embeddings", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_attention_heads", "num_heads", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embedding_type", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "return_dict", "rms_norm_eps", "rope", "rope_theta", "rotary_emb", "router_logits", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "vocab_size", "warning_once", "with", "xpu"], "granitemoe/modeling_granitemoe.py:GraniteMoeForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "class", "config", "def", "device", "else", "float", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "list", "lm_head", "load_balancing_loss_func", "logits", "logits_scaling", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "falcon_h1/modeling_falcon_h1.py:FalconHybridMambaAttentionDynamicCache": ["Any", "False", "LongTensor", "ModelConfig", "ModelHybridMambaAttentionDynamicCache", "None", "Optional", "Tensor", "_", "__init__", "append", "batch_size", "beam_idx", "cache_kwargs", "cache_position", "cat", "clamp", "class", "config", "conv_kernel_size", "conv_state", "conv_states", "def", "device", "devices", "dict", "dim", "dims", "dtype", "elif", "else", "float16", "for", "get_seq_length", "has_previous_state", "hidden_size", "i", "if", "in", "index_select", "int", "intermediate_size", "is", "is_compileable", "key_cache", "key_states", "layer_idx", "len", "list", "mamba_d_conv", "mamba_d_head", "mamba_d_ssm", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "new_conv_state", "not", "num_hidden_layers", "range", "reorder_cache", "reset", "return", "roll", "self", "seqlen_offset", "shape", "shifts", "ssm_states", "str", "to", "torch", "transformer_layers", "tuple", "update", "update_conv_state", "value_cache", "value_states", "zero_", "zeros"], "falcon_h1/modeling_falcon_h1.py:FalconH1RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "falcon_h1/modeling_falcon_h1.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "falcon_h1/modeling_falcon_h1.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "falcon_h1/modeling_falcon_h1.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "falcon_h1/modeling_falcon_h1.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "falcon_h1/modeling_falcon_h1.py:FalconH1Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_multiplier", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "falcon_h1/modeling_falcon_h1.py:FalconH1RMSNormGated": ["F", "ModelRMSNormGated", "Module", "None", "Parameter", "True", "__init__", "and", "batch_size", "class", "def", "dim", "dtype", "else", "eps", "float32", "forward", "gate", "hidden_size", "hidden_states", "if", "input_dtype", "int", "is", "keepdim", "len", "mean", "n_groups", "nn", "norm_before_gate", "not", "ones", "pow", "return", "rsqrt", "self", "seq_len", "shape", "silu", "squeeze", "super", "to", "torch", "variance", "variance_epsilon", "view", "weight"], "falcon_h1/modeling_falcon_h1.py:pad_tensor_by_size": ["Model", "Model_shape", "Model_size", "Model_tensor_by_size", "Tensor", "constant", "def", "else", "functional", "if", "input_tensor", "int", "len", "mode", "nn", "return", "shape", "torch", "value"], "falcon_h1/modeling_falcon_h1.py:reshape_into_chunks": ["Model", "Model_into_chunks", "chunk_size", "def", "else", "if", "input_tensor", "len", "pad_size", "pad_tensor_by_size", "return", "shape"], "falcon_h1/modeling_falcon_h1.py:segment_sum": ["Model_sum", "None", "bool", "chunk_size", "cumsum", "def", "device", "diagonal", "dim", "dtype", "expand", "inf", "input_tensor", "mask", "masked_fill", "ones", "return", "size", "tensor_segsum", "torch", "tril"], "falcon_h1/modeling_falcon_h1.py:apply_mask_to_padding_states": ["Model_mask_to_padding_states", "None", "and", "attention_mask", "def", "dtype", "hidden_states", "if", "is", "not", "return", "shape", "to"], "falcon_h1/modeling_falcon_h1.py:FalconH1Mixer": ["A", "ACT2FN", "AILab", "A_cumsum", "A_log", "B", "B_decay", "C", "C_reshaped", "C_times_states", "Conv1d", "D", "D_residual", "Dao", "F", "Falling", "False", "G", "GPU", "G_intermediate", "L", "Linear", "LongTensor", "M", "M_intermediate", "Model", "ModelConfig", "ModelHybridMambaAttentionDynamicCache", "ModelMixer", "ModelRMSNormGated", "Module", "None", "Optional", "Parameter", "Tensor", "The", "To", "True", "Y_diag", "Y_off", "_", "__init__", "a", "act", "activation", "and", "apply_mask_to_padding_states", "arange", "attention_mask", "available", "back", "batch_size", "be", "because", "bias", "bmm", "cache_device", "cache_params", "cache_position", "cat", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk_size", "clamp", "class", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_dim", "conv_kernel_size", "conv_states", "copy_", "cuda", "cuda_kernels_forward", "cumsum", "dA", "dB", "dBx", "d_mlp", "d_to_remove", "decay_chunk", "decay_states", "def", "device", "dim", "dims", "dt", "dt_bias", "dt_limit", "dt_limit_kwargs", "dt_softplus", "dtype", "else", "eps", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "groups_time_state_size", "has_previous_state", "head_dim", "headdim", "hidden_act", "hidden_size", "hidden_states", "hidden_states_B_C", "hidden_states_B_C_transposed", "hidden_states_reshaped", "https", "if", "implementation", "in", "in_channels", "in_proj", "inf", "input_states", "install", "int", "intermediate_size", "is", "is_fast_path_available", "kernel_size", "layer_idx", "layer_norm_epsilon", "log", "logger", "mamba", "mamba_chunk_scan_combined", "mamba_chunk_size", "mamba_conv_bias", "mamba_d_conv", "mamba_d_head", "mamba_d_ssm", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "mamba_norm_before_gate", "mamba_proj_bias", "mamba_rms_norm", "mamba_split_conv1d_scan_combined", "model", "mup_vector", "n_groups", "naive", "new_states", "ngroups", "nn", "norm", "norm_before_gate", "not", "num_heads", "of", "on", "one", "ones", "or", "out", "out_channels", "out_proj", "outproj_bias", "outproj_weight", "output_size", "pad", "pad_size", "pad_tensor_by_size", "padding", "path", "permute", "previous_states", "projected_states", "projection_size", "projectors_bias", "repeat_interleave", "reshape", "reshape_into_chunks", "return", "return_final_states", "rms_norm_eps", "rmsnorm_eps", "rmsnorm_weight", "roll", "running", "scan_output", "segment_sum", "selective_state_update", "self", "seq_idx", "seq_len", "shape", "shifts", "silu", "softplus", "spaces", "split", "squeeze", "ssm_in_multiplier", "ssm_multipliers", "ssm_state", "ssm_state_size", "ssm_states", "ssm_states_reshaped", "state", "state_decay_out", "state_decay_out_permuted", "states", "sum", "super", "swish", "t", "the", "time_step", "time_step_limit", "time_step_max", "time_step_min", "to", "torch", "torch_forward", "training", "transpose", "type", "update_conv_state", "use_bias", "use_conv_bias", "use_precomputed_states", "used", "variance_epsilon", "view", "warning_once", "weight", "when", "will", "with", "x", "x0", "y", "z", "z0", "zeros_like", "zxbcdt_multipliers"], "falcon_h1/modeling_falcon_h1.py:FalconH1MLP": ["ACT2FN", "Linear", "ModelConfig", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_multiplier", "down_proj", "forward", "gate_multiplier", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "mlp_multipliers", "nn", "return", "self", "super", "up_proj", "x", "y"], "falcon_h1/modeling_falcon_h1.py:FalconH1RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "falcon_h1/modeling_falcon_h1.py:FalconH1DecoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelHybridMambaAttentionDynamicCache", "ModelMLP", "ModelMixer", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_hidden_states", "attention_in_multiplier", "attention_mask", "attention_out_multiplier", "attn_out_multiplier", "bool", "cache_params", "cache_position", "channels_attn", "class", "config", "def", "deprecate_kwarg", "eps", "feed_forward", "forward", "head_dim", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mamba", "mamba_attention_mask", "mamba_hidden_states", "new_name", "num_attention_heads", "num_key_value_heads", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "pre_ff_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "ssm_out_multiplier", "super", "torch", "tuple", "use_cache", "version"], "falcon_h1/modeling_falcon_h1.py:FalconH1PreTrainedModel": ["Exception", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "Skipping", "True", "_init_weights", "_is_stateful", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "and", "as", "base_model_prefix", "bias", "class", "config", "continue", "data", "def", "due", "e", "elif", "else", "error", "except", "f", "fill_", "for", "if", "in", "init", "initializer_range", "layernorm", "lower", "mean", "model", "module", "name", "named_parameters", "normal_", "not", "param", "past_key_values", "print", "recurse", "requires_grad", "self", "std", "supports_gradient_checkpointing", "to", "try", "weight", "zero_"], "falcon_h1/modeling_falcon_h1.py:compute_mup_vector": ["Model_mup_vector", "None", "config", "def", "else", "groups_time_state_size", "hidden_size", "if", "int", "intermediate_size", "is", "mamba_d_ssm", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "mup_vector", "not", "num_heads", "ones", "return", "ssm_multipliers", "torch", "vector_shape", "zxbcdt_multipliers"], "falcon_h1/modeling_falcon_h1.py:FalconH1Model": ["AttentionMaskConverter", "BaseModelOutputWithPast", "Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoderLayer", "ModelHybridMambaAttentionDynamicCache", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "_update_mamba_mask", "a", "all", "all_hidden_states", "all_self_attns", "an", "and", "append", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "cache", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "compute_mup_vector", "config", "cuda", "decoder_layer", "decoder_layers", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "embedding_multiplier", "eps", "exactly", "expand", "fill_value", "final_layernorm", "finfo", "flash_attention_2", "for", "forward", "full", "get_seq_length", "gradient", "gradient_checkpointing", "has_previous_state", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "initialized", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer", "layer_idx", "layer_outputs", "layers", "lm_head_multiplier", "logger", "mamba", "mamba_attention_mask", "mamba_mask", "mask_length", "masked_fill", "min", "min_dtype", "mup_vector", "must", "next_cache", "nn", "no", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_attention_mask", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "persistent", "position_embeddings", "position_ids", "post_init", "provided", "raise", "range", "register_buffer", "requires", "reshape", "return", "returned", "rms_norm_eps", "rotary_emb", "sdpa", "self", "sequence_length", "shape", "so", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "vocab_size", "warning_once", "was", "will", "with", "xpu"], "falcon_h1/modeling_falcon_h1.py:FalconH1ForCausalLM": ["CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelHybridMambaAttentionDynamicCache", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "contiguous", "conv1d", "cumsum", "def", "device", "devices", "dtype", "elif", "else", "empty_past_kv", "for", "forward", "hidden_size", "hidden_states", "i", "if", "in", "input_ids", "inputs_embeds", "int", "is", "is_torchdynamo_compiling", "isinstance", "items", "key", "kwargs", "labels", "layers", "lm_head", "lm_head_multiplier", "logits", "logits_to_keep", "long", "loss", "loss_function", "mamba", "masked_fill_", "model", "model_inputs", "nn", "not", "num_hidden_layers", "num_logits_to_keep", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "range", "return", "self", "shape", "slice", "slice_indices", "super", "torch", "tuple", "update", "use_cache", "value", "vocab_size", "weight"], "table_transformer/modeling_table_transformer.py:TableTransformerDecoderOutput": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "ModelDecoderOutput", "None", "Optional", "class", "intermediate_hidden_states", "r", "torch"], "table_transformer/modeling_table_transformer.py:TableTransformerModelOutput": ["FloatTensor", "ModelModelOutput", "None", "Optional", "Seq2SeqModelOutput", "class", "intermediate_hidden_states", "r", "torch"], "table_transformer/modeling_table_transformer.py:TableTransformerObjectDetectionOutput": ["FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "table_transformer/modeling_table_transformer.py:TableTransformerFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "table_transformer/modeling_table_transformer.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "table_transformer/modeling_table_transformer.py:TableTransformerConvEncoder": ["Either", "False", "ModelConvEncoder", "Module", "None", "Tensor", "True", "ValueError", "__init__", "and", "append", "backbone", "backbone_config", "backbone_kwargs", "backbone_model_type", "be", "bool", "channels", "class", "config", "copy", "create_model", "def", "dilation", "elif", "else", "feature_info", "feature_map", "feature_maps", "features", "features_only", "float", "for", "forward", "functional", "get", "getattr", "if", "in", "in_chans", "intermediate_channel_sizes", "interpolate", "is", "kwargs", "layer2", "layer3", "layer4", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "num_channels", "or", "out", "out_indices", "output_stride", "parameter", "pixel_mask", "pixel_values", "pop", "pretrained", "provided", "raise", "replace_batch_norm", "requires_backends", "requires_grad_", "resnet", "return", "self", "shape", "should", "size", "stage", "super", "the", "timm", "to", "torch", "use_pretrained_backbone", "use_timm_backbone", "with"], "table_transformer/modeling_table_transformer.py:TableTransformerConvModel": ["ModelConvModel", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "table_transformer/modeling_table_transformer.py:TableTransformerSinePositionEmbedding": ["False", "ModelSinePositionEmbedding", "Module", "No", "None", "True", "ValueError", "__init__", "and", "arange", "be", "cat", "class", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "flatten", "float", "float32", "floor", "forward", "if", "int64", "is", "mask", "math", "nn", "normalize", "not", "passed", "permute", "pi", "pixel", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "provided", "raise", "return", "rounding_mode", "scale", "self", "should", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "table_transformer/modeling_table_transformer.py:TableTransformerLearnedPositionEmbedding": ["Embedding", "ModelLearnedPositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "table_transformer/modeling_table_transformer.py:build_position_encoding": ["ModelLearnedPositionEmbedding", "ModelSinePositionEmbedding", "Model_position_encoding", "Not", "True", "ValueError", "config", "d_model", "def", "elif", "else", "f", "if", "learned", "n_steps", "normalize", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "table_transformer/modeling_table_transformer.py:TableTransformerAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "is_cross_attention", "k_proj", "key_states", "key_value_states", "key_value_states_original", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "object_queries", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "spatial_position_embeddings", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "table_transformer/modeling_table_transformer.py:TableTransformerEncoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "max", "min", "nn", "num_heads", "object_queries", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "table_transformer/modeling_table_transformer.py:TableTransformerDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "key_value_states", "nn", "not", "num_heads", "object_queries", "output_attentions", "outputs", "p", "query_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_position_embeddings", "super", "torch", "training"], "table_transformer/modeling_table_transformer.py:TableTransformerPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "Linear", "ModelConfig", "ModelConvEncoder", "ModelDecoderLayer", "ModelEncoderLayer", "ModelLearnedPositionEmbedding", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "column_embeddings", "config", "data", "def", "elif", "if", "init", "init_std", "is", "isinstance", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "r", "row_embeddings", "self", "std", "uniform_", "weight", "zero_"], "table_transformer/modeling_table_transformer.py:TableTransformerEncoder": ["BaseModelOutput", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "True", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "attention_mask", "attentions", "class", "config", "d_model", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "for", "forward", "functional", "hidden_states", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layernorm", "layers", "nn", "not", "object_queries", "output_attentions", "output_hidden_states", "p", "post_init", "r", "rand", "range", "return", "return_dict", "self", "super", "to_drop", "torch", "training", "tuple", "use_return_dict", "v"], "table_transformer/modeling_table_transformer.py:TableTransformerDecoder": ["False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "_prepare_4d_attention_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attention_mask", "attentions", "auxiliary_loss", "class", "combined_attention_mask", "config", "continue", "cross_attentions", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "input_shape", "inputs_embeds", "intermediate", "intermediate_hidden_states", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layernorm", "layers", "nn", "not", "object_queries", "output_attentions", "output_hidden_states", "post_init", "query_position_embeddings", "r", "rand", "range", "return", "return_dict", "self", "size", "stack", "super", "tgt_len", "torch", "training", "tuple", "use_return_dict", "v"], "table_transformer/modeling_table_transformer.py:TableTransformerModel": ["Backbone", "BaseModelOutput", "Conv2d", "Embedding", "False", "FloatTensor", "ModelConfig", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_size", "bool", "build_position_encoding", "class", "config", "conv_encoder", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "device", "does", "downsampled", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "feature_map", "features", "flatten", "flattened_features", "flattened_mask", "for", "forward", "freeze_backbone", "get_encoder", "height", "hidden_states", "if", "in", "input_projection", "inputs_embeds", "intermediate_channel_sizes", "intermediate_hidden_states", "is", "isinstance", "kernel_size", "last_hidden_state", "len", "mask", "model", "name", "named_parameters", "nn", "not", "num_channels", "num_queries", "object_queries", "ones", "output_attentions", "output_hidden_states", "param", "permute", "pixel", "pixel_mask", "pixel_values", "position_embeddings_list", "post_init", "projected_feature_map", "queries", "query_position_embeddings", "r", "raise", "repeat", "requires_grad_", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "unfreeze_backbone", "unsqueeze", "use_return_dict", "weight", "width", "zeros_like"], "table_transformer/modeling_table_transformer.py:TableTransformerForObjectDetection": ["FloatTensor", "Linear", "ModelConfig", "ModelForObjectDetection", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "bbox_predictor", "bool", "class", "class_labels_classifier", "config", "cross_attentions", "d_model", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "def", "device", "dict", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "hidden_dim", "if", "input_dim", "inputs_embeds", "intermediate", "intermediate_hidden_states", "is", "labels", "last_hidden_state", "list", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_labels", "num_layers", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "outputs_class", "outputs_coord", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "r", "return", "return_dict", "self", "sequence_output", "sigmoid", "super", "torch", "tuple", "use_return_dict"], "table_transformer/modeling_table_transformer.py:TableTransformerMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "speecht5/modeling_speecht5.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "speecht5/modeling_speecht5.py:shift_spectrograms_right": ["Model_spectrograms_right", "Modeled_input_values", "None", "Optional", "Tensor", "attention_mask", "clone", "def", "if", "input_values", "int", "is", "masked_fill_", "new_zeros", "not", "reduction_factor", "return", "shape", "torch"], "speecht5/modeling_speecht5.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "speecht5/modeling_speecht5.py:SpeechT5NoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "speecht5/modeling_speecht5.py:SpeechT5LayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "speecht5/modeling_speecht5.py:SpeechT5GroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "speecht5/modeling_speecht5.py:SpeechT5SinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "cos", "create_position_ids_from_input_ids", "cumsum", "def", "detach", "device", "dim", "dtype", "emb", "emb_weights", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "speecht5/modeling_speecht5.py:SpeechT5PositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "speecht5/modeling_speecht5.py:SpeechT5ScaledPositionalEncoding": ["Dropout", "False", "ModelScaledPositionalEncoding", "Module", "Parameter", "__init__", "alpha", "arange", "class", "cos", "def", "dim", "div_term", "dropout", "dtype", "emb", "exp", "float", "forward", "int64", "log", "math", "max_len", "nn", "p", "pe", "persistent", "position", "register_buffer", "return", "self", "sin", "size", "super", "tensor", "torch", "unsqueeze", "zeros"], "speecht5/modeling_speecht5.py:SpeechT5RelativePositionalEncoding": ["Embedding", "ModelRelativePositionalEncoding", "Module", "None", "__init__", "arange", "class", "def", "device", "dim", "dtype", "forward", "hidden_states", "long", "max_length", "nn", "pe_k", "pos_seq", "return", "self", "seq_len", "shape", "super", "to", "torch"], "speecht5/modeling_speecht5.py:SpeechT5SamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "speecht5/modeling_speecht5.py:SpeechT5FeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "speecht5/modeling_speecht5.py:SpeechT5FeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "speecht5/modeling_speecht5.py:SpeechT5SpeechEncoderPrenet": ["FloatTensor", "LongTensor", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModelEncoderPrenet", "ModelPositionalConvEmbedding", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_conv_out_length", "_freeze_parameters", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_mask_hidden_states", "and", "apply_spec_augment", "arange", "attention_mask", "batch_size", "bool", "class", "config", "conv_kernel", "conv_stride", "cumsum", "def", "device", "dim", "div", "dtype", "elif", "else", "expand", "extract_features", "feature_encoder", "feature_projection", "feature_vector_length", "flip", "floor", "for", "forward", "freeze_feature_encoder", "getattr", "hidden_size", "hidden_states", "if", "in", "input_length", "input_lengths", "input_values", "int", "is", "kernel_size", "long", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "max_Model_positions", "min_masks", "ne", "nn", "non_padded_lengths", "not", "or", "output_lengths", "pad_token_id", "padding_mask", "pos_conv_embed", "pos_sinusoidal_embed", "positional_conv_embedding", "positional_sinusoidal_embeddings", "return", "rounding_mode", "self", "sequence_length", "shape", "size", "stride", "super", "tensor", "to", "torch", "training", "transpose", "uniform_", "zeros", "zip"], "speecht5/modeling_speecht5.py:SpeechT5SpeechDecoderPrenet": ["Linear", "ModelModelDecoderPrenet", "ModelScaledPositionalEncoding", "Model_decoder_prenet_dropout", "Model_decoder_prenet_layers", "Model_decoder_prenet_units", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "_consistent_dropout", "all_masks", "bernoulli", "cat", "class", "config", "def", "dim", "else", "encode_positions", "expand", "final_layer", "for", "forward", "functional", "hidden_size", "i", "if", "in", "input_values", "inputs_embeds", "is", "layer", "layers", "mask", "max_Model_positions", "nn", "normalize", "not", "num_mel_bins", "p", "positional_dropout", "range", "relu", "repeat", "return", "self", "size", "speaker_embedding_dim", "speaker_embeddings", "speaker_embeds_layer", "super", "torch", "unsqueeze", "where"], "speecht5/modeling_speecht5.py:SpeechT5BatchNormConvLayer": ["BatchNorm1d", "Conv1d", "Dropout", "False", "ModelBatchNormConvLayer", "Model_decoder_postnet_dropout", "Model_decoder_postnet_kernel", "Model_decoder_postnet_layers", "Model_decoder_postnet_units", "Module", "None", "Tanh", "__init__", "activation", "batch_norm", "bias", "class", "config", "conv", "def", "dropout", "else", "forward", "hidden_states", "if", "in_conv_dim", "is", "kernel_size", "layer_id", "nn", "not", "num_mel_bins", "out_conv_dim", "padding", "return", "self", "stride", "super"], "speecht5/modeling_speecht5.py:SpeechT5SpeechDecoderPostnet": ["Linear", "ModelBatchNormConvLayer", "ModelModelDecoderPostnet", "Model_decoder_postnet_layers", "Module", "ModuleList", "Tensor", "__init__", "class", "config", "def", "feat_out", "for", "forward", "hidden_size", "hidden_states", "i", "in", "layer", "layer_output", "layers", "logits", "nn", "num_mel_bins", "outputs_after_postnet", "outputs_before_postnet", "postnet", "prob_out", "range", "reduction_factor", "return", "self", "size", "super", "torch", "transpose", "view"], "speecht5/modeling_speecht5.py:SpeechT5TextEncoderPrenet": ["Embedding", "EmbeddingAccessMixin", "ModelScaledPositionalEncoding", "ModelTextEncoderPrenet", "Module", "Tensor", "__init__", "class", "config", "def", "embed_tokens", "encode_positions", "forward", "hidden_size", "input_ids", "inputs_embeds", "max_text_positions", "nn", "pad_token_id", "positional_dropout", "return", "self", "super", "torch", "vocab_size"], "speecht5/modeling_speecht5.py:SpeechT5TextDecoderPrenet": ["Cache", "Dropout", "Embedding", "EmbeddingAccessMixin", "LongTensor", "ModelSinusoidalPositionalEmbedding", "ModelTextDecoderPrenet", "Module", "None", "Optional", "Tensor", "ValueError", "You", "__init__", "attention_mask", "class", "config", "decoder_input_ids", "def", "dropout", "else", "embed_positions", "embed_scale", "embed_tokens", "forward", "get_seq_length", "have", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "isinstance", "math", "max_text_positions", "nn", "not", "pad_token_id", "past_key_values", "past_key_values_length", "positional_dropout", "positions", "raise", "return", "scale_embedding", "self", "shape", "size", "specify", "sqrt", "super", "to", "torch", "view", "vocab_size"], "speecht5/modeling_speecht5.py:SpeechT5TextDecoderPostnet": ["EmbeddingAccessMixin", "False", "Linear", "ModelTextDecoderPostnet", "Module", "Tensor", "__init__", "bias", "class", "config", "def", "forward", "get_output_embeddings", "hidden_size", "hidden_states", "lm_head", "new_embeddings", "nn", "return", "self", "set_output_embeddings", "super", "torch", "vocab_size"], "speecht5/modeling_speecht5.py:SpeechT5Attention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "class", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "else", "embed_dim", "f", "float", "for", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "mask", "matmul", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "position_bias", "proj_shape", "q_proj", "query_states", "raise", "rel_pos_bias", "reshape", "reshape_q", "return", "scaling", "self", "self_attention_cache", "should", "single", "size", "softmax", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view", "weights"], "speecht5/modeling_speecht5.py:SpeechT5FeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "speecht5/modeling_speecht5.py:SpeechT5EncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelFeedForward", "None", "Optional", "Tensor", "__init__", "attention", "attention_dropout", "attention_mask", "attn_weights", "bool", "class", "config", "def", "dropout", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_head_mask", "layer_norm", "layer_norm_eps", "nn", "num_heads", "output_attentions", "outputs", "position_bias", "residual", "return", "self", "super", "torch"], "speecht5/modeling_speecht5.py:SpeechT5DecoderLayer": ["Cache", "Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFeedForward", "None", "Optional", "Tensor", "True", "__init__", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm_eps", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "use_cache", "version"], "speecht5/modeling_speecht5.py:SpeechT5PreTrainedModel": ["BatchNorm1d", "Conv1d", "Embedding", "GroupNorm", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelFeatureProjection", "ModelPositionalConvEmbedding", "ModelPreTrainedModel", "ModelScaledPositionalEncoding", "Module", "None", "PreTrainedModel", "True", "_init_weights", "a", "alpha", "b", "base_model_prefix", "bias", "class", "config", "constant_", "conv", "data", "def", "elif", "fill_", "groups", "hasattr", "if", "in_channels", "in_features", "init", "initializer_range", "input_values", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "main_input_name", "masked_spec_embed", "math", "mean", "module", "nn", "normal_", "not", "padding_idx", "projection", "self", "sqrt", "std", "supports_gradient_checkpointing", "uniform_", "weight", "zero_"], "speecht5/modeling_speecht5.py:SpeechT5Encoder": ["BaseModelOutput", "Dropout", "False", "FloatTensor", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelRelativePositionalEncoding", "ModuleList", "None", "Optional", "Tensor", "The", "Union", "ValueError", "_", "__init__", "_prepare_4d_attention_mask", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "be", "bool", "but", "class", "config", "def", "dropout", "dropout_probability", "dtype", "else", "embed_positions", "encoder_attention_heads", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_max_relative_position", "enumerate", "eps", "f", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "idx", "if", "in", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "len", "nn", "not", "or", "output_attentions", "output_hidden_states", "position_bias", "post_init", "raise", "rand", "range", "return", "return_dict", "self", "should", "size", "skip_the_layer", "specified", "super", "synced_gpus", "torch", "training", "tuple", "use_return_dict", "v"], "speecht5/modeling_speecht5.py:SpeechT5EncoderWithSpeechPrenet": ["BaseModelOutput", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelEncoderWithModelPrenet", "ModelModelEncoderPrenet", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "bool", "class", "config", "def", "forward", "head_mask", "hidden_states", "input_values", "output_attentions", "output_hidden_states", "outputs", "post_init", "prenet", "return", "return_dict", "self", "super", "torch", "tuple", "wrapped_encoder"], "speecht5/modeling_speecht5.py:SpeechT5EncoderWithTextPrenet": ["BaseModelOutput", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelEncoderWithTextPrenet", "ModelPreTrainedModel", "ModelTextEncoderPrenet", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "bool", "class", "config", "def", "forward", "get_input_embeddings", "head_mask", "hidden_states", "input_values", "output_attentions", "output_hidden_states", "outputs", "post_init", "prenet", "return", "return_dict", "self", "set_input_embeddings", "super", "torch", "tuple", "value", "wrapped_encoder"], "speecht5/modeling_speecht5.py:SpeechT5EncoderWithoutPrenet": ["BaseModelOutput", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelEncoderWithoutPrenet", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "bool", "class", "config", "def", "forward", "head_mask", "hidden_states", "input_values", "output_attentions", "output_hidden_states", "post_init", "return", "return_dict", "self", "super", "torch", "tuple", "wrapped_encoder"], "speecht5/modeling_speecht5.py:SpeechT5Decoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "attention_mask", "attentions", "attn_mask", "be", "bool", "but", "cache_position", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "dropout_probability", "dtype", "e", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "g", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_shape", "instance", "instead", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "pass", "past_key_values", "past_key_values_length", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "self", "should", "size", "skip_the_layer", "specified", "super", "synced_gpus", "tgt_len", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "warning_once", "will", "with", "zip"], "speecht5/modeling_speecht5.py:SpeechT5DecoderWithSpeechPrenet": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderWithModelPrenet", "ModelModelDecoderPrenet", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_head_mask", "decoder_hidden_states", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "head_mask", "hidden_states", "input_values", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "prenet", "return", "return_dict", "self", "speaker_embeddings", "super", "torch", "tuple", "use_cache", "wrapped_decoder"], "speecht5/modeling_speecht5.py:SpeechT5DecoderWithTextPrenet": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderWithTextPrenet", "ModelPreTrainedModel", "ModelTextDecoderPrenet", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_head_mask", "decoder_hidden_states", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "head_mask", "hidden_states", "input_values", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "prenet", "return", "return_dict", "self", "set_input_embeddings", "super", "torch", "tuple", "use_cache", "value", "wrapped_decoder"], "speecht5/modeling_speecht5.py:SpeechT5DecoderWithoutPrenet": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderWithoutPrenet", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_head_mask", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "head_mask", "hidden_states", "input_values", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "return", "return_dict", "self", "super", "torch", "tuple", "use_cache", "wrapped_decoder"], "speecht5/modeling_speecht5.py:SpeechT5GuidedMultiheadAttentionLoss": ["BoolTensor", "FloatTensor", "ModelConfig", "ModelGuidedMultiheadAttentionLoss", "Module", "Tensor", "__init__", "_make_guided_attention_mask", "_make_guided_attention_masks", "arange", "attentions", "class", "config", "def", "device", "enumerate", "exp", "float", "for", "forward", "grid_x", "grid_y", "guided_attention_loss_scale", "guided_attention_loss_sigma", "guided_attn_masks", "idx", "ilen", "in", "indexing", "input_length", "input_lengths", "input_masks", "len", "loss", "losses", "masked_select", "masks", "mean", "meshgrid", "nn", "olen", "output_length", "output_lengths", "output_masks", "return", "scale", "self", "shape", "sigma", "staticmethod", "sum", "super", "to", "torch", "unsqueeze", "xy", "zeros", "zip"], "speecht5/modeling_speecht5.py:SpeechT5SpectrogramLoss": ["BCEWithLogitsLoss", "FloatTensor", "L1Loss", "LongTensor", "ModelConfig", "ModelGuidedMultiheadAttentionLoss", "ModelSpectrogramLoss", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn", "attn_criterion", "attn_loss", "bce_criterion", "bce_loss", "cat", "class", "config", "cross_attentions", "def", "device", "dim", "for", "forward", "guided_attention_loss_num_heads", "if", "in", "input_masks", "l1_criterion", "l1_loss", "labels", "logits", "loss", "masked_select", "masks", "nn", "ones", "output_masks", "outputs_after_postnet", "outputs_before_postnet", "padding_mask", "pos_weight", "reduction_factor", "return", "self", "size", "stop_labels", "super", "tensor", "to", "torch", "use_guided_attention_loss", "x"], "speecht5/modeling_speecht5.py:SpeechT5Model": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderWithModelPrenet", "ModelDecoderWithTextPrenet", "ModelDecoderWithoutPrenet", "ModelEncoderWithModelPrenet", "ModelEncoderWithTextPrenet", "ModelEncoderWithoutPrenet", "ModelModel", "ModelPreTrainedModel", "Module", "None", "NotImplementedError", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_get_feature_vector_attention_mask", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_args", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_values", "decoder_outputs", "def", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "freeze_feature_encoder", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_values", "is", "isinstance", "last_hidden_state", "len", "nn", "not", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "prenet", "r", "raise", "return", "return_dict", "self", "set_input_embeddings", "shape", "speaker_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "value"], "speecht5/modeling_speecht5.py:SpeechT5ForSpeechToText": ["Cache", "CrossEntropyLoss", "FloatTensor", "GenerationMixin", "LongTensor", "Model", "ModelConfig", "ModelDecoderWithTextPrenet", "ModelEncoderWithModelPrenet", "ModelForModeloText", "ModelModel", "ModelPreTrainedModel", "ModelTextDecoderPostnet", "Model_encoder", "None", "Optional", "Please", "Seq2SeqLMOutput", "Tensor", "True", "Union", "ValueError", "You", "__class__", "__init__", "_tied_weights_keys", "a", "are", "as", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "config", "configuration", "cross_attentions", "cross_attn_head_mask", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_input_values", "decoder_start_token_id", "def", "define", "does", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "follows", "forward", "freeze_feature_encoder", "from_pretrained", "get_decoder", "get_encoder", "get_output_embeddings", "head", "head_mask", "if", "input_values", "instantiate", "is", "labels", "language", "lm_head", "logits", "loss", "loss_fct", "model", "new_embeddings", "not", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "post_init", "prenet", "r", "raise", "return", "return_dict", "s", "self", "set_output_embeddings", "shift_tokens_right", "size", "super", "text_decoder", "text_decoder_postnet", "that", "the", "to", "torch", "trying", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "vocabulary", "weight", "with", "your"], "speecht5/modeling_speecht5.py:_generate_speech": ["False", "FloatTensor", "LongTensor", "Model", "ModelEncoderWithSpeechPrenet", "ModelPreTrainedModel", "Module", "None", "Optional", "True", "Union", "ValueError", "_generate_speech", "_get_feature_vector_attention_mask", "append", "attention_mask", "batch_first", "bool", "break", "bsz", "cat", "config", "continue", "cross_attentions", "decoder", "decoder_hidden_states", "decoder_out", "def", "dim", "else", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_out", "feat_out", "flatten", "float", "for", "hidden_states", "i", "idx", "if", "in", "input_values", "int", "is", "isinstance", "last_decoder_output", "last_hidden_state", "len", "max", "maxlen", "maxlenratio", "meet_index", "meet_indexes", "meet_thresholds", "minlen", "minlenratio", "model", "new_spectrogram", "new_zeros", "nn", "not", "num_mel_bins", "output_attentions", "output_cross_attentions", "output_sequence", "outputs", "pad_sequence", "pad_token_id", "past_key_values", "postnet", "prenet", "prob", "prob_out", "raise", "range", "reduction_factor", "result_spectrogram", "return", "return_dict", "return_output_lengths", "rnn", "shape", "sigmoid", "size", "speaker_embeddings", "spectrogram", "spectrogram_lengths", "spectrograms", "spectrum", "speech_decoder_postnet", "squeeze", "stack", "sum", "threshold", "tolist", "torch", "transpose", "tuple", "use_cache", "utils", "view", "vocoder", "waveform_lengths", "waveforms", "where", "while", "wrapped_decoder"], "speecht5/modeling_speecht5.py:SpeechT5ForTextToSpeech": ["Cache", "False", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoderWithModelPrenet", "ModelEncoderWithTextPrenet", "ModelForTextToModel", "ModelModel", "ModelModelDecoderPostnet", "ModelPreTrainedModel", "ModelSpectrogramLoss", "Model_decoder", "Model_decoder_postnet", "Module", "None", "Optional", "Please", "Seq2SeqSpectrogramOutput", "Tensor", "The", "True", "Union", "ValueError", "You", "__class__", "__init__", "_generate_Model", "a", "are", "as", "attention_mask", "auto_docstring", "batch", "batch_size", "be", "bool", "cache_position", "can_generate", "class", "classmethod", "cls", "config", "configuration", "criterion", "cross_attentions", "cross_attn_head_mask", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_values", "def", "define", "dimension", "does", "either", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "first", "float", "follows", "forward", "from_pretrained", "generate", "generate_Model", "get_decoder", "get_encoder", "head", "head_mask", "if", "input_ids", "input_values", "instantiate", "is", "kwargs", "labels", "language", "logits", "loss", "main_input_name", "maxlenratio", "minlenratio", "model", "must", "nn", "no_grad", "not", "of", "or", "output", "output_attentions", "output_cross_attentions", "output_hidden_states", "outputs", "outputs_after_postnet", "outputs_before_postnet", "past_key_values", "post_init", "r", "raise", "reduction_factor", "repeat", "return", "return_dict", "return_output_lengths", "s", "same", "self", "shift_spectrograms_right", "size", "speaker_embeddings", "spectrogram", "stop_labels", "super", "text_encoder", "that", "the", "threshold", "to", "torch", "trying", "tuple", "use_cache", "use_guided_attention_loss", "use_return_dict", "vocab_size", "vocabulary", "vocoder", "with", "your"], "speecht5/modeling_speecht5.py:SpeechT5ForSpeechToSpeech": ["Cache", "False", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoderWithModelPrenet", "ModelEncoderWithModelPrenet", "ModelForModeloModel", "ModelModel", "ModelModelDecoderPostnet", "ModelPreTrainedModel", "Model_decoder", "Model_decoder_postnet", "Model_encoder", "Module", "None", "Optional", "Seq2SeqSpectrogramOutput", "Tensor", "True", "Union", "_", "__init__", "_generate_Model", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_values", "def", "device", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "float", "forward", "freeze_feature_encoder", "generate_Model", "get_decoder", "get_encoder", "head_mask", "if", "input_values", "is", "labels", "logits", "loss", "maxlenratio", "minlenratio", "nn", "no_grad", "not", "output", "output_attentions", "output_cross_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "prenet", "r", "reduction_factor", "return", "return_dict", "return_output_lengths", "self", "shift_spectrograms_right", "speaker_embeddings", "spectrogram", "stop_labels", "super", "threshold", "torch", "tuple", "use_cache", "use_return_dict", "vocoder", "zeros"], "speecht5/modeling_speecht5.py:HifiGanResidualBlock": ["Conv1d", "ModelGanResidualBlock", "Module", "ModuleList", "_", "__init__", "apply_weight_norm", "channels", "class", "conv1", "conv2", "convs1", "convs2", "def", "dilation", "for", "forward", "functional", "get_padding", "hasattr", "hidden_states", "i", "if", "in", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "nn", "padding", "parametrizations", "range", "remove_weight_norm", "residual", "return", "self", "stride", "super", "utils", "weight_norm", "zip"], "speecht5/modeling_speecht5.py:SpeechT5HifiGan": ["Conv1d", "ConvTranspose1d", "FloatTensor", "HifiGanResidualBlock", "ModelHifiGan", "ModelHifiGanConfig", "Module", "ModuleList", "None", "PreTrainedModel", "__init__", "_init_weights", "append", "apply_weight_norm", "auto_docstring", "bias", "channels", "class", "config", "conv_post", "conv_pre", "custom_intro", "data", "def", "dilation", "dim", "else", "enumerate", "for", "forward", "functional", "hasattr", "hidden_states", "i", "if", "in", "initializer_range", "is", "is_batched", "isinstance", "j", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "main_input_name", "mean", "model_in_dim", "module", "nn", "normal_", "normalize_before", "not", "num_kernels", "num_upsamples", "ones", "padding", "parametrizations", "post_init", "r", "range", "register_buffer", "remove_weight_norm", "res_state", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "scale", "self", "spectrogram", "squeeze", "std", "stride", "super", "tanh", "torch", "transpose", "unsqueeze", "upsample_initial_channel", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "utils", "view", "waveform", "weight", "weight_norm", "zero_", "zeros", "zip"], "hiera/modeling_hiera.py:HieraEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "reshaped_hidden_states", "torch", "tuple"], "hiera/modeling_hiera.py:HieraModelOutput": ["BoolTensor", "FloatTensor", "LongTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "bool_masked_pos", "class", "hidden_states", "ids_restore", "last_hidden_state", "pooler_output", "r", "reshaped_hidden_states", "torch", "tuple"], "hiera/modeling_hiera.py:HieraForImageClassificationOutput": ["FloatTensor", "ImageClassifierOutput", "ModelForImageClassificationOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "hiera/modeling_hiera.py:HieraForPreTrainingOutput": ["BoolTensor", "FloatTensor", "LongTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "bool_masked_pos", "class", "hidden_states", "ids_restore", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "hiera/modeling_hiera.py:HieraPatchEmbeddings": ["BoolTensor", "Conv2d", "False", "FloatTensor", "LongTensor", "ModelPatchEmbeddings", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "argsort", "batch_size", "be", "bool", "bool_masked_pos", "but", "class", "config", "def", "device", "dim", "dimensions", "else", "embed_dim", "embeddings", "f", "flatten", "float", "for", "forward", "functional", "gather", "got", "i", "ids_restore", "ids_shuffle", "if", "image", "image_size", "in", "index", "input", "int", "interpolate", "is", "is_mae", "kernel_size", "len", "len_keep", "mask_ratio", "mask_spatial_shape", "masked_conv", "masked_unit_size", "math", "nn", "noise", "num_channels", "num_windows", "number", "of", "padding", "patch_padding", "patch_size", "patch_stride", "pixel_values", "prod", "projection", "raise", "rand", "random_masking", "return", "s", "self", "shape", "should", "size", "spatial_dims", "stride", "super", "target_size", "the", "to", "tokens_spatial_shape", "torch", "transpose", "tuple", "view", "zeros", "zip"], "hiera/modeling_hiera.py:HieraEmbeddings": ["BoolTensor", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "align_corners", "and", "bicubic", "bool", "bool_masked_pos", "class", "config", "def", "dim", "else", "embed_dim", "embeddings", "for", "forward", "functional", "get_position_embedding", "height", "i", "ids_restore", "if", "image_size", "in", "int", "interpolate", "interpolate_pos_encoding", "is_mae", "is_tracing", "jit", "mask_spatial_shape", "masked_unit_size", "math", "mode", "new_height", "new_width", "nn", "noise", "not", "num_patches", "num_positions", "num_tokens", "patch_embeddings", "patch_stride", "permute", "pixel_values", "pos_embeds", "position_embeddings", "prod", "reshape", "return", "s", "self", "shape", "size", "sqrt_num_positions", "super", "tokens_spatial_shape", "torch", "torch_int", "tuple", "view", "width", "zeros", "zip"], "hiera/modeling_hiera.py:HieraMaskUnitAttention": ["False", "FloatTensor", "Linear", "ModelMaskUnitAttention", "Module", "None", "Optional", "Tensor", "_", "__init__", "attn_output", "attn_weights", "batch_size", "bool", "class", "def", "dim", "else", "forward", "head_dim", "head_mask", "hidden_size", "hidden_size_output", "hidden_states", "if", "int", "is", "key", "max", "nn", "not", "num_heads", "num_windows", "output_attentions", "permute", "proj", "qkv", "query", "query_stride", "reshape", "return", "scale", "self", "seq_len", "shape", "softmax", "super", "torch", "transpose", "tuple", "unbind", "use_mask_unit_attn", "value", "values", "view", "window_size"], "hiera/modeling_hiera.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "hiera/modeling_hiera.py:HieraDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "hiera/modeling_hiera.py:HieraMlp": ["ACT2FN", "Linear", "ModelMlp", "Module", "None", "Tensor", "__init__", "activation_fn", "class", "config", "def", "dim", "fc1", "fc2", "forward", "hidden_act", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "hiera/modeling_hiera.py:HieraLayer": ["False", "FloatTensor", "Identity", "LayerNorm", "Linear", "ModelDropPath", "ModelLayer", "ModelMaskUnitAttention", "ModelMlp", "Module", "None", "Optional", "Tensor", "_", "__init__", "attn", "attn_weights", "batch_size", "bool", "class", "config", "def", "dim", "drop_path", "else", "eps", "float", "forward", "head_mask", "hidden_size", "hidden_size_output", "hidden_states", "hidden_states_norm", "if", "int", "layer_norm_eps", "layernorm_after", "layernorm_before", "max", "mlp", "nn", "num_heads", "output_attentions", "proj", "query_stride", "residual", "return", "self", "seq_len", "shape", "super", "torch", "tuple", "use_mask_unit_attn", "values", "view", "window_size"], "hiera/modeling_hiera.py:HieraStage": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelLayer", "ModelStage", "ModuleList", "None", "Optional", "Tensor", "__init__", "and", "attn_weights", "bool", "class", "config", "def", "depth", "drop_path", "else", "enumerate", "float", "for", "forward", "head_mask", "hidden_size", "hidden_size_output", "hidden_states", "i", "if", "in", "int", "is", "layer_head_mask", "layer_module", "layers", "list", "masked_unit_attention", "nn", "not", "num_heads", "or", "output_attentions", "previous_stage_used_masked_attention", "query_stride", "range", "return", "self", "stage_num", "super", "torch", "tuple", "use_mask_unit_attn", "window_size"], "hiera/modeling_hiera.py:undo_windowing": ["Model_windowing", "Tensor", "batch_size", "def", "for", "hidden_size", "hidden_states", "in", "int", "list", "mask_unit_shape", "mu", "num_mask_units", "permute", "reshape", "return", "s", "shape", "torch", "view", "zip"], "hiera/modeling_hiera.py:HieraEncoder": ["BaseModelOutput", "BoolTensor", "False", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelEncoderOutput", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "append", "attentions", "batch_size", "bool", "bool_masked_pos", "class", "config", "cpu", "cumsum", "cumulative_depths", "def", "depth", "depths", "device", "dpr", "drop_path", "drop_path_rate", "else", "embed_dim", "embed_dim_multiplier", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_size_output", "hidden_states", "i", "idx_stage", "if", "image_size", "in", "int", "is", "item", "last_hidden_state", "layer_head_mask", "layer_outputs", "len", "linspace", "mask_unit_shape", "masked_unit_area", "masked_unit_attention", "masked_unit_size", "math", "nn", "not", "num_dim", "num_heads", "num_query_pool", "output_attentions", "output_hidden_states", "patch_stride", "permute", "prod", "query_pool_layer", "query_stride", "query_stride_area", "query_strides", "range", "reroll", "reshape", "reshaped_hidden_states", "return", "return_dict", "s", "schedule", "self", "seq_len", "shape", "size", "stage", "stage_ends", "stage_idx", "stage_module", "stage_num", "stage_size", "stages", "strides", "sum", "super", "tensor", "tolist", "torch", "total_depth", "tuple", "undo_windowing", "unroll_schedule", "use_mask_unit_attn", "v", "view", "window_size", "x", "zip"], "hiera/modeling_hiera.py:unroll": ["Model", "Tensor", "_", "batch_size", "current_size", "def", "flatten", "for", "hidden_size", "hidden_states", "i", "image_shape", "in", "int", "item", "len", "list", "math", "new_shape", "num_dims", "pair", "patch_stride", "permute", "prod", "range", "reshape", "return", "s", "schedule", "shape", "size", "strides", "torch", "tuple", "view", "zip"], "hiera/modeling_hiera.py:HieraPreTrainedModel": ["Conv1d", "Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoder", "ModelEmbeddings", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "constant_", "decoder_position_embeddings", "def", "elif", "if", "init", "initializer_range", "is", "isinstance", "layer_norm_init", "main_input_name", "mask_token", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight"], "hiera/modeling_hiera.py:HieraPooler": ["AdaptiveAvgPool1d", "LayerNorm", "ModelConfig", "ModelPooler", "Module", "Tensor", "__init__", "class", "config", "def", "depths", "embed_dim", "embed_dim_multiplier", "eps", "flatten", "forward", "hidden_states", "int", "layer_norm_eps", "layernorm", "len", "nn", "num_features", "pooled_output", "pooler", "return", "self", "super", "torch", "transpose"], "hiera/modeling_hiera.py:HieraModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPatchEmbeddings", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "batch_size", "bool", "bool_masked_pos", "class", "config", "def", "depths", "dict", "else", "embed_dim", "embed_dim_multiplier", "embedding_output", "embeddings", "encoder", "encoder_outputs", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "head_outputs", "heads", "heads_to_prune", "hidden_size", "hidden_states", "ids_restore", "if", "image_shape", "in", "int", "interpolate_pos_encoding", "is", "is_mae", "items", "last_hidden_state", "layer", "len", "list", "mask_unit_area", "masked_unit_size", "math", "noise", "not", "num_features", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_stride", "pixel_values", "pooled_output", "pooler", "pooler_output", "positions", "post_init", "prod", "prune_heads", "query_stride", "r", "raise", "reshaped_hidden_states", "return", "return_dict", "schedule", "self", "sequence_output", "shape", "specify", "super", "tile", "to", "torch", "tuple", "unroll", "unroll_schedule", "unsqueeze", "use_return_dict", "view"], "hiera/modeling_hiera.py:HieraDecoder": ["BoolTensor", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoder", "ModelStage", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "attn_weights", "batch_size", "bool", "bool_masked_pos", "class", "config", "decoder_block", "decoder_depth", "decoder_embeddings", "decoder_hidden_size", "decoder_hidden_states", "decoder_norm", "decoder_num_heads", "decoder_position_embeddings", "decoder_pred", "def", "depth", "depths", "device", "drop_path", "dtype", "embed_dim", "embed_dim_multiplier", "encoder_hidden_states", "eps", "expand", "flatten", "float", "for", "forward", "head_mask", "hidden_size", "hidden_size_output", "hidden_states", "i", "image_size", "in", "int", "layer_norm_eps", "len", "mask_token", "mask_tokens", "mask_unit_height", "mask_unit_spatial_shape_final", "mask_unit_width", "masked_unit_size", "math", "nn", "num_channels", "num_features", "num_heads", "num_mask_units", "num_query_pool", "output_attentions", "patch_stride", "pred_dim", "pred_stride", "prod", "query_stride", "reshape", "return", "s", "self", "shape", "super", "tokens_spatial_shape", "tokens_spatial_shape_final", "torch", "tuple", "undo_windowing", "use_mask_unit_attn", "view", "window_size", "zeros", "zip"], "hiera/modeling_hiera.py:HieraMultiScaleHead": ["Conv2d", "Identity", "ModelConfig", "ModelMultiScaleHead", "Module", "ModuleList", "Tensor", "__init__", "append", "apply_fusion_head", "batch_size", "class", "config", "current_masked_unit_size", "def", "depths", "embed_dim", "embed_dim_multiplier", "feature_map", "feature_maps", "for", "forward", "head", "hidden_size", "hidden_states", "i", "idx", "if", "in", "int", "isinstance", "kernel", "kernel_size", "len", "list", "mask_unit_height", "mask_unit_height_final", "mask_unit_spatial_shape_final", "mask_unit_width", "mask_unit_width_final", "masked_unit_size", "multi_scale_fusion_heads", "nn", "num_mask_units", "num_query_pool", "permute", "query_stride", "range", "reshape", "return", "s", "self", "shape", "stage_dimensions", "stride", "super", "torch", "zip"], "hiera/modeling_hiera.py:HieraForPreTraining": ["BoolTensor", "False", "FloatTensor", "LayerNorm", "Model", "ModelConfig", "ModelDecoder", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelMultiScaleHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "decoder", "def", "dim", "else", "encoder_norm", "eps", "feature_maps", "flatten", "forward", "forward_loss", "fused_hidden_states", "get_pixel_label_2d", "head_mask", "hidden_states", "ids_restore", "ids_to_restore", "if", "interpolate_pos_encoding", "is", "is_mae", "keepdim", "label", "layer_norm_eps", "logits", "loss", "mean", "multiscale_fusion", "nn", "noise", "normalize_pixel_loss", "not", "num_features", "num_query_pool", "output", "output_attentions", "output_hidden_states", "outputs", "permute", "pixel_values", "post_init", "pred_stride", "r", "reshaped_hidden_states", "return", "return_dict", "self", "size", "super", "torch", "tuple", "unfold", "use_return_dict", "var"], "hiera/modeling_hiera.py:HieraForImageClassification": ["False", "Identity", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelForImageClassificationOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_states", "if", "interpolate_pos_encoding", "is", "is_mae", "labels", "logits", "loss", "loss_function", "nn", "not", "num_features", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "post_init", "r", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "hiera/modeling_hiera.py:HieraBackbone": ["BackboneMixin", "BackboneOutput", "False", "LayerNorm", "ModelBackbone", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "ModuleDict", "None", "Optional", "Tensor", "True", "_", "__init__", "_init_backbone", "_out_features", "attentions", "batch_size", "bool", "channels", "class", "config", "contiguous", "def", "depths", "else", "embed_dim", "embed_dim_multiplier", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "get_input_embeddings", "head_mask", "height", "hidden_state", "hidden_states", "hidden_states_norms", "i", "if", "in", "int", "is", "is_mae", "len", "nn", "not", "num_channels", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "outputs", "patch_embeddings", "permute", "pixel_values", "post_init", "range", "return", "return_dict", "self", "shape", "stage", "stage_names", "super", "torch", "use_return_dict", "view", "width", "zip"], "canine/modeling_canine.py:CanineModelOutputWithPooling": ["FloatTensor", "ModelModelOutputWithPooling", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "torch", "tuple"], "canine/modeling_canine.py:CanineEmbeddings": ["Dropout", "Embedding", "Expected", "False", "FloatTensor", "HashBucketCodepointEmbedder_", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "ValueError", "_PRIMES", "__init__", "_embed_hash_buckets", "_hash_bucket_tensors", "absolute", "append", "arange", "be", "cat", "char_position_embeddings", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embedding_shards", "embedding_size", "embeddings", "enumerate", "eps", "expand", "f", "for", "forward", "getattr", "hash_bucket_ids", "hash_bucket_tensors", "hashed", "hidden_dropout_prob", "hidden_size", "i", "if", "in", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "len", "long", "max_position_embeddings", "must", "name", "nn", "not", "num_buckets", "num_hash_buckets", "num_hash_functions", "num_hashes", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "prime", "primes", "raise", "range", "register_buffer", "result_tensors", "return", "self", "seq_length", "setattr", "shard_embedding_size", "shard_embeddings", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "zeros"], "canine/modeling_canine.py:CharactersToMolecules": ["ACT2FN", "Conv1d", "LayerNorm", "ModelToMolecules", "Module", "Tensor", "__init__", "activation", "cat", "char_encoding", "class", "cls_encoding", "config", "conv", "def", "dim", "downsampled", "downsampled_truncated", "downsampling_rate", "eps", "forward", "hidden_act", "hidden_size", "in_channels", "kernel_size", "layer_norm_eps", "nn", "out_channels", "result", "return", "self", "stride", "super", "torch", "transpose"], "canine/modeling_canine.py:ConvProjection": ["ACT2FN", "ConstantPad1d", "Dropout", "LayerNorm", "Model", "Model1d", "ModelForMaskedLM", "ModelProjection", "Module", "None", "NotImplementedError", "Optional", "Tensor", "__init__", "activation", "class", "config", "currently", "def", "dropout", "else", "eps", "final_char_seq", "final_seq_char_positions", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "if", "in_channels", "inputs", "is", "kernel_size", "layer_norm_eps", "nn", "not", "out_channels", "pad", "pad_beg", "pad_end", "pad_total", "query_seq", "raise", "result", "return", "self", "stride", "super", "supported", "torch", "transpose", "upsampling_kernel_size"], "canine/modeling_canine.py:CanineSelfAttention": ["Dropout", "Embedding", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "_", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bhld", "bhlr", "bhrd", "bool", "class", "config", "context_layer", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "f", "finfo", "float", "forward", "from_tensor", "functional", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "long", "lrd", "math", "matmul", "max_position_embeddings", "min", "multiple", "ndim", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "to", "to_tensor", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view"], "canine/modeling_canine.py:CanineSelfOutput": ["Dropout", "FloatTensor", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch", "tuple"], "canine/modeling_canine.py:CanineAttention": ["Check", "Expected", "False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "ValueError", "__init__", "all_head_size", "always_attend_to_first_position", "and", "append", "attend_from_chunk_stride", "attend_from_chunk_width", "attend_to_chunk_stride", "attend_to_chunk_width", "attention_head_size", "attention_mask", "attention_mask_chunk", "attention_output", "attention_output_chunks", "attention_outputs_chunk", "attention_probs_chunks", "bool", "cat", "cause", "chunk_end", "chunk_start", "class", "cls_attention_mask", "cls_position", "config", "def", "dense", "dim", "else", "f", "find_pruneable_heads_and_indices", "first_position_attends_to_all", "for", "forward", "from_chunks", "from_end", "from_seq_length", "from_start", "from_tensor", "from_tensor_chunk", "get", "have", "head_mask", "heads", "hidden_states", "if", "in", "index", "int", "key", "len", "local", "min", "nn", "not", "num_attention_heads", "number", "of", "output", "output_attentions", "outputs", "positions", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "raise", "range", "return", "same", "self", "self_outputs", "sequence", "set", "shape", "skipped", "strides", "super", "to", "to_chunks", "to_end", "to_seq_length", "to_start", "to_tensor", "to_tensor_chunk", "torch", "tuple", "union", "value", "would", "zip"], "canine/modeling_canine.py:CanineIntermediate": ["ACT2FN", "FloatTensor", "Linear", "ModelIntermediate", "Module", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "canine/modeling_canine.py:CanineOutput": ["Dropout", "FloatTensor", "LayerNorm", "Linear", "ModelOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch", "tuple"], "canine/modeling_canine.py:CanineLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "__init__", "always_attend_to_first_position", "apply_chunking_to_forward", "attend_from_chunk_stride", "attend_from_chunk_width", "attend_to_chunk_stride", "attend_to_chunk_width", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "first_position_attends_to_all", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "layer_output", "local", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "canine/modeling_canine.py:CanineEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "True", "Union", "_", "__init__", "all_hidden_states", "all_self_attentions", "always_attend_to_first_position", "attend_from_chunk_stride", "attend_from_chunk_width", "attend_to_chunk_stride", "attend_to_chunk_width", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "first_position_attends_to_all", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "local", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "canine/modeling_canine.py:CaninePooler": ["FloatTensor", "Linear", "ModelPooler", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch", "tuple"], "canine/modeling_canine.py:CaninePredictionHeadTransform": ["ACT2FN", "FloatTensor", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn", "tuple"], "canine/modeling_canine.py:CanineLMPredictionHead": ["False", "FloatTensor", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "tuple", "vocab_size", "zeros"], "canine/modeling_canine.py:CanineOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch", "tuple"], "canine/modeling_canine.py:CaninePreTrainedModel": ["Conv1d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "canine/modeling_canine.py:CanineModel": ["CharactersToMolecules", "ConvProjection", "False", "FloatTensor", "LongTensor", "MaxPool1d", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutputWithPooling", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_create_3d_attention_mask_from_input_mask", "_downsample_attention_mask", "_prune_heads", "_repeat_molecules", "add_pooling_layer", "all_hidden_states", "all_self_attentions", "always_attend_to_first_position", "and", "at", "attend_from_chunk_stride", "attend_from_chunk_width", "attend_to_chunk_stride", "attend_to_chunk_width", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "broadcast_ones", "cannot", "cat", "char_attention_mask", "char_embeddings", "char_seq_len", "char_seq_length", "chars_to_molecules", "class", "concat", "config", "copy", "deep_encoder_hidden_states", "deep_encoder_self_attentions", "deepcopy", "def", "device", "dim", "downsampling_rate", "dtype", "either", "elif", "else", "encoder", "encoder_outputs", "extended_attention_mask", "extended_molecule_attention_mask", "final_char_encoder", "final_chars_encoder_outputs", "first_position_attends_to_all", "float", "float32", "for", "forward", "from_seq_length", "from_tensor", "get_extended_attention_mask", "get_head_mask", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "init_chars_encoder_outputs", "init_molecule_encoding", "initial_char_encoder", "input_char_embeddings", "input_char_encoding", "input_ids", "input_shape", "inputs_embeds", "int", "is", "items", "kernel_size", "last_hidden_state", "last_molecule", "layer", "local", "local_transformer_stride", "long", "mask", "molecule_attention_mask", "molecule_sequence_output", "molecules", "molecules_without_extra_cls", "nn", "not", "num_hidden_layers", "ones", "or", "output", "output_attentions", "output_hidden_states", "poolable_char_mask", "pooled_molecule_mask", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "projection", "prune_heads", "r", "raise", "rate", "remainder_length", "remainder_repeated", "repeat_interleave", "repeated", "repeated_molecules", "repeats", "reshape", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "shallow_config", "shape", "size", "specify", "squeeze", "stride", "super", "the", "time", "to", "to_mask", "to_seq_length", "token_type_ids", "torch", "tuple", "use_return_dict", "v", "warn_if_padding_and_no_attention_mask", "zeros"], "canine/modeling_canine.py:CanineForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "canine/modeling_canine.py:CanineForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "canine/modeling_canine.py:CanineForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "canine/modeling_canine.py:CanineForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp_", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "xlm_roberta/modeling_xlm_roberta.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "Model", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "Model", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaSelfOutput": ["Dropout", "LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaAttention": ["Cache", "False", "FloatTensor", "Model", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaIntermediate": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaOutput": ["Dropout", "LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "Model", "ModelAttention", "ModelIntermediate", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaLMHead": ["LayerNorm", "Linear", "Model", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLMHead", "ModelLayer", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "roberta", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "Model", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaPooler": ["Linear", "Model", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "Model", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "Model", "ModelLMHead", "ModelLMHeadModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "device", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "device", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaClassificationHead": ["Dropout", "Linear", "Model", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "roberta", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "view"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaForMultipleChoice": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "roberta", "self", "shape", "size", "super", "to", "token_type_ids", "torch", "tuple", "view"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "roberta", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "view"], "xlm_roberta/modeling_xlm_roberta.py:XLMRobertaForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "roberta", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "zoedepth/modeling_zoedepth.py:ZoeDepthDepthEstimatorOutput": ["FloatTensor", "ModelDepthEstimatorOutput", "ModelOutput", "None", "Optional", "attentions", "class", "domain_logits", "hidden_states", "loss", "predicted_depth", "r", "torch", "tuple"], "zoedepth/modeling_zoedepth.py:ZoeDepthReassembleStage": ["ACT2FN", "Linear", "ModelReassembleLayer", "ModelReassembleStage", "Module", "ModuleList", "Sequential", "Tensor", "_", "__init__", "add", "append", "backbone_hidden_size", "batch_size", "cat", "channels", "class", "cls_token", "config", "contiguous", "def", "dim", "elif", "enumerate", "expand_as", "factor", "flatten", "for", "forward", "hidden_act", "hidden_size", "hidden_state", "hidden_states", "if", "in", "layers", "list", "neck_hidden_size", "neck_hidden_sizes", "nn", "num_channels", "out", "patch_height", "patch_width", "permute", "project", "readout", "readout_projects", "readout_type", "reassemble_factors", "reshape", "return", "self", "sequence_length", "shape", "split", "stage_idx", "super", "torch", "total_batch_size", "unsqueeze", "zip"], "zoedepth/modeling_zoedepth.py:ZoeDepthReassembleLayer": ["Conv2d", "ConvTranspose2d", "Identity", "ModelReassembleLayer", "Module", "__init__", "backbone_hidden_size", "channels", "class", "config", "def", "elif", "factor", "forward", "hidden_size", "hidden_state", "if", "in_channels", "int", "kernel_size", "nn", "out_channels", "padding", "projection", "resize", "return", "self", "stride", "super"], "zoedepth/modeling_zoedepth.py:ZoeDepthFeatureFusionStage": ["ModelConfig", "ModelFeatureFusionLayer", "ModelFeatureFusionStage", "Module", "ModuleList", "None", "_", "__init__", "append", "class", "config", "def", "else", "for", "forward", "fused_hidden_state", "fused_hidden_states", "hidden_state", "hidden_states", "if", "in", "is", "layer", "layers", "len", "neck_hidden_sizes", "nn", "range", "return", "self", "super", "zip"], "zoedepth/modeling_zoedepth.py:ZoeDepthPreActResidualLayer": ["BatchNorm2d", "Conv2d", "ModelPreActResidualLayer", "Module", "None", "ReLU", "Tensor", "__init__", "activation1", "activation2", "batch_norm1", "batch_norm2", "batch_norm_eps", "bias", "class", "config", "convolution1", "convolution2", "def", "else", "eps", "forward", "fusion_hidden_size", "hidden_state", "if", "is", "kernel_size", "nn", "not", "padding", "residual", "return", "self", "stride", "super", "torch", "use_batch_norm", "use_batch_norm_in_fusion_residual", "use_bias_in_fusion_residual"], "zoedepth/modeling_zoedepth.py:ZoeDepthFeatureFusionLayer": ["Conv2d", "False", "ModelConfig", "ModelFeatureFusionLayer", "ModelPreActResidualLayer", "Module", "None", "Optional", "Tensor", "True", "__init__", "align_corners", "bias", "bilinear", "bool", "class", "config", "def", "forward", "functional", "fusion_hidden_size", "hidden_state", "if", "interpolate", "is", "kernel_size", "mode", "nn", "not", "projection", "residual", "residual_layer1", "residual_layer2", "return", "scale_factor", "self", "shape", "size", "super", "torch"], "zoedepth/modeling_zoedepth.py:ZoeDepthNeck": ["Conv2d", "False", "ModelConfig", "ModelFeatureFusionStage", "ModelNeck", "ModelReassembleStage", "Module", "ModuleList", "None", "Tensor", "The", "TypeError", "ValueError", "__init__", "a", "and", "append", "backbone_config", "be", "bias", "channel", "class", "config", "convs", "def", "else", "enumerate", "equal", "feature", "features", "for", "forward", "fusion_hidden_size", "fusion_stage", "hidden", "hidden_states", "i", "if", "in", "is", "isinstance", "kernel_size", "len", "list", "model_type", "neck", "neck_hidden_sizes", "nn", "not", "number", "of", "or", "output", "padding", "patch_height", "patch_width", "raise", "reassemble_stage", "return", "self", "should", "sizes", "states", "super", "swinv2", "tensors", "the", "to", "torch", "tuple"], "zoedepth/modeling_zoedepth.py:ZoeDepthRelativeDepthEstimationHead": ["Conv2d", "ModelRelativeDepthEstimationHead", "Module", "None", "ReLU", "Tensor", "True", "Upsample", "__init__", "add_projection", "align_corners", "bilinear", "class", "config", "conv1", "conv2", "conv3", "def", "dim", "features", "forward", "fusion_hidden_size", "head_in_index", "hidden_states", "if", "is", "kernel_size", "list", "mode", "nn", "not", "num_relative_features", "padding", "predicted_depth", "projection", "return", "scale_factor", "self", "squeeze", "stride", "super", "torch", "upsample"], "zoedepth/modeling_zoedepth.py:log_binom": ["Model", "Model_binom", "def", "eps", "k", "n", "return", "torch"], "zoedepth/modeling_zoedepth.py:LogBinomialSoftmax": ["False", "Model", "ModelBinomialSoftmax", "Model_binom", "Module", "__init__", "act", "arange", "clamp", "class", "def", "dim", "eps", "forward", "if", "k", "k_idx", "k_minus_1", "n_classes", "ndim", "nn", "one_minus_probabilities", "persistent", "probabilities", "register_buffer", "return", "self", "softmax", "super", "temperature", "tensor", "torch", "unsqueeze", "view", "y"], "zoedepth/modeling_zoedepth.py:ZoeDepthConditionalLogBinomialSoftmax": ["Conv2d", "GELU", "LogBinomialSoftmax", "ModelConditionalLogBinomialSoftmax", "Module", "Sequential", "Softplus", "__init__", "act", "bottleneck", "bottleneck_factor", "class", "concat", "condition_dim", "condition_feature", "config", "def", "dim", "forward", "in_features", "kernel_size", "log_binomial_transform", "main_feature", "max_temp", "min_temp", "mlp", "n_classes", "nn", "p_eps", "padding", "probabilities", "probabilities_and_temperature", "return", "self", "softmax", "stride", "super", "temperature", "torch", "unsqueeze"], "zoedepth/modeling_zoedepth.py:ZoeDepthSeedBinRegressor": ["Conv2d", "ModelSeedBinRegressor", "Module", "ReLU", "Softplus", "True", "__init__", "act1", "act2", "bin_centers", "bin_centers_type", "bin_edges", "bin_widths", "bin_widths_normed", "bottleneck_features", "class", "config", "constant", "conv1", "conv2", "cumsum", "def", "dim", "else", "forward", "functional", "if", "in_features", "inplace", "keepdim", "max_depth", "min_depth", "mlp_dim", "mode", "n_bins", "nn", "normed", "pad", "return", "self", "sum", "super", "torch", "value", "x"], "zoedepth/modeling_zoedepth.py:inv_attractor": ["Model_attractor", "alpha", "def", "div", "dx", "float", "gamma", "int", "pow", "return"], "zoedepth/modeling_zoedepth.py:ZoeDepthAttractorLayer": ["Conv2d", "False", "ModelAttractorLayer", "Module", "None", "ReLU", "True", "_", "__init__", "act1", "act2", "align_corners", "alpha", "attractor_alpha", "attractor_gamma", "attractor_kind", "attractors", "attractors_normed", "batch_size", "bilinear", "bin_centers", "bin_embedding_dim", "bin_new_centers", "class", "clip", "config", "conv1", "conv2", "def", "delta_c", "device", "dim", "else", "for", "forward", "func", "functional", "gemma", "height", "i", "if", "in", "in_features", "inplace", "interpolate", "inv_attractor", "is", "kind", "max_depth", "mean", "memory_efficient", "min_depth", "mlp_dim", "mode", "n_attractors", "n_bins", "nn", "not", "prev_bin", "prev_bin_embedding", "range", "return", "self", "shape", "sort", "sum", "super", "torch", "unsqueeze", "view", "width", "x", "zeros_like"], "zoedepth/modeling_zoedepth.py:ZoeDepthAttractorLayerUnnormed": ["Conv2d", "ModelAttractorLayerUnnormed", "Module", "None", "ReLU", "Softplus", "True", "__init__", "act1", "act2", "align_corners", "alpha", "attractor_alpha", "attractor_kind", "attractors", "bilinear", "bin_centers", "bin_embedding_dim", "bin_new_centers", "class", "config", "conv1", "conv2", "def", "delta_c", "device", "dim", "else", "for", "forward", "func", "functional", "gamma", "height", "i", "if", "in", "in_features", "inplace", "interpolate", "inv_attractor", "is", "kind", "max_depth", "mean", "memory_efficient", "min_depth", "mlp_dim", "mode", "n_attractors", "n_bins", "nn", "not", "prev_bin", "prev_bin_embedding", "range", "return", "self", "shape", "sum", "super", "torch", "unsqueeze", "width", "x", "zeros_like"], "zoedepth/modeling_zoedepth.py:ZoeDepthProjector": ["Conv2d", "ModelProjector", "Module", "ReLU", "Tensor", "True", "__init__", "act", "class", "conv1", "conv2", "def", "forward", "hidden_state", "in_features", "inplace", "mlp_dim", "nn", "out_features", "return", "self", "super", "torch"], "zoedepth/modeling_zoedepth.py:ZoeDepthMultiheadAttention": ["Dropout", "False", "FloatTensor", "Linear", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_scores", "batch_size", "bool", "class", "context_layer", "contiguous", "def", "dim", "dropout", "else", "f", "forward", "functional", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "keys", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "out_proj", "output_attentions", "outputs", "permute", "queries", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "values", "view"], "zoedepth/modeling_zoedepth.py:ZoeDepthTransformerEncoderLayer": ["ACT2FN", "Dropout", "LayerNorm", "Linear", "ModelMultiheadAttention", "ModelTransformerEncoderLayer", "Module", "None", "Optional", "Tensor", "__init__", "activation", "attention_mask", "class", "config", "def", "dropout", "dropout1", "dropout2", "forward", "hidden_size", "intermediate_size", "keys", "linear1", "linear2", "nn", "norm1", "norm2", "num_attention_heads", "patch_transformer_hidden_size", "patch_transformer_intermediate_size", "patch_transformer_num_attention_heads", "queries", "relu", "return", "self", "self_attn", "src", "src2", "src_mask", "super", "torch", "values"], "zoedepth/modeling_zoedepth.py:ZoeDepthPatchTransformerEncoder": ["Conv2d", "ModelPatchTransformerEncoder", "ModelTransformerEncoderLayer", "Module", "ModuleList", "_", "__init__", "arange", "batch_size", "bottleneck_features", "cat", "class", "config", "cos", "cpu", "def", "device", "dim", "div_term", "dtype", "embedding_convPxP", "embedding_dim", "embeddings", "exp", "flatten", "float32", "for", "forward", "functional", "i", "in", "in_channels", "index", "kernel_size", "log", "nn", "num_patch_transformer_layers", "pad", "padding", "patch_transformer_hidden_size", "permute", "pos_encoding", "position", "positional_encoding_1d", "range", "repeat", "return", "self", "sequence_length", "shape", "sin", "stride", "super", "tensor", "torch", "transformer_encoder", "unsqueeze", "x"], "zoedepth/modeling_zoedepth.py:ZoeDepthMLPClassifier": ["Linear", "ModelMLPClassifier", "Module", "None", "ReLU", "__init__", "activation", "class", "def", "domain_logits", "forward", "hidden_features", "hidden_state", "in_features", "linear1", "linear2", "nn", "out_features", "return", "self", "super"], "zoedepth/modeling_zoedepth.py:ZoeDepthMultipleMetricDepthEstimationHeads": ["Attractor", "Conv2d", "IndexError", "ModelAttractorLayer", "ModelAttractorLayerUnnormed", "ModelConditionalLogBinomialSoftmax", "ModelMLPClassifier", "ModelMultipleMetricDepthEstimationHeads", "ModelPatchTransformerEncoder", "ModelProjector", "ModelSeedBinRegressor", "Module", "ModuleDict", "ModuleList", "True", "ValueError", "_", "__init__", "align_corners", "argmax", "attractor", "attractors", "bilinear", "bin", "bin_centers", "bin_centers_type", "bin_configurations", "bin_configurations_name", "bin_configurationss", "bin_embedding", "bin_embedding_dim", "bottleneck", "bottleneck_factor", "bottleneck_features", "class", "conditional_log_binomial", "conf", "config", "configuration", "conv2", "def", "dim", "domain_logits", "domain_vote", "elif", "else", "embedding", "except", "f", "feature", "feature_blocks", "for", "forward", "found", "functional", "fusion_hidden_size", "hybrid2", "i", "if", "in", "in_features", "interpolate", "item", "keepdim", "kernel_size", "last", "last_in", "len", "max_depth", "min_depth", "mlp_classifier", "mlp_dim", "mode", "n_attractors", "n_bins", "name", "names", "nn", "normed", "not", "num_attractors", "num_relative_features", "out", "out_features", "outconv_activation", "padding", "patch_transformer", "prev_bin", "prev_bin_embedding", "projector", "projectors", "raise", "range", "relative_depth", "return", "seed_bin_centers", "seed_bin_regressor", "seed_bin_regressors", "seed_projector", "self", "shape", "softmax", "softplus", "squeeze", "stride", "sum", "super", "torch", "try", "x", "zip"], "zoedepth/modeling_zoedepth.py:ZoeDepthMetricDepthEstimationHead": ["Attractor", "Conv2d", "ModelAttractorLayer", "ModelAttractorLayerUnnormed", "ModelConditionalLogBinomialSoftmax", "ModelMetricDepthEstimationHead", "ModelProjector", "ModelSeedBinRegressor", "Module", "ModuleList", "None", "True", "_", "__init__", "align_corners", "attractor", "attractors", "bilinear", "bin", "bin_centers", "bin_centers_type", "bin_configuration", "bin_configurations", "bin_embedding", "bin_embedding_dim", "bottleneck", "bottleneck_features", "cat", "class", "clone", "conditional_log_binomial", "config", "conv2", "def", "dim", "elif", "else", "feature", "feature_blocks", "for", "forward", "functional", "fusion_hidden_size", "hybrid2", "i", "if", "in", "in_features", "interpolate", "keepdim", "kernel_size", "last", "last_in", "max_depth", "min_depth", "mode", "n_attractors", "n_bins", "n_classes", "nn", "normed", "num_attractors", "num_relative_features", "out", "out_features", "outconv_activation", "padding", "prev_bin", "prev_bin_embedding", "projector", "projectors", "range", "relative_conditioning", "relative_depth", "return", "seed_bin_centers", "seed_bin_regressor", "seed_projector", "self", "shape", "size", "softplus", "stride", "sum", "super", "torch", "unsqueeze", "x", "zip"], "zoedepth/modeling_zoedepth.py:ZoeDepthPreTrainedModel": ["Conv2d", "ConvTranspose2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "zoedepth/modeling_zoedepth.py:ZoeDepthForDepthEstimation": ["DepthEstimatorOutput", "FloatTensor", "LongTensor", "Model", "ModelDepthEstimatorOutput", "ModelForDepthEstimation", "ModelMetricDepthEstimationHead", "ModelMultipleMetricDepthEstimationHeads", "ModelNeck", "ModelPreTrainedModel", "ModelRelativeDepthEstimationHead", "None", "NotImplementedError", "Optional", "Tensor", "Training", "Union", "ValueError", "_", "__init__", "and", "assumes", "attentions", "attributes", "auto_docstring", "backbone", "backbone_hidden_size", "bin_configurations", "bool", "bottleneck", "class", "config", "def", "dim", "domain_logits", "else", "feature_blocks", "feature_maps", "features", "forward", "forward_with_filtered_kwargs", "hasattr", "have", "height", "hidden_size", "hidden_states", "if", "implemented", "is", "labels", "len", "load_backbone", "loss", "metric_depth", "metric_head", "neck", "not", "out", "outconv_activation", "output", "output_attentions", "output_hidden_states", "outputs", "patch_height", "patch_size", "patch_width", "pixel_values", "post_init", "predicted_depth", "r", "raise", "relative_depth", "relative_head", "return", "return_dict", "s", "self", "shape", "squeeze", "super", "the", "to", "torch", "tuple", "use_return_dict", "width", "yet"], "groupvit/modeling_groupvit.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "groupvit/modeling_groupvit.py:groupvit_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "groupvit/modeling_groupvit.py:hard_softmax": ["Model_softmax", "Tensor", "True", "def", "detach", "dim", "index", "int", "keepdim", "legacy_contiguous_format", "logits", "max", "memory_format", "ret", "return", "scatter_", "softmax", "torch", "y_Model", "y_soft", "zeros_like"], "groupvit/modeling_groupvit.py:gumbel_softmax": ["False", "Model", "Model_dist", "Model_softmax", "Models", "Tensor", "True", "bool", "def", "detach", "device", "dim", "distributions", "dtype", "else", "float", "hard", "if", "index", "int", "keepdim", "legacy_contiguous_format", "logits", "max", "memory_format", "ret", "return", "sample", "scatter_", "shape", "softmax", "tau", "tensor", "torch", "y_hard", "y_soft", "zeros_like"], "groupvit/modeling_groupvit.py:resize_attention_map": ["False", "Model_attention_map", "align_corners", "attentions", "batch_size", "bilinear", "def", "else", "feat_height", "feat_width", "functional", "groups", "height", "if", "int", "interpolate", "mode", "nn", "np", "reshape", "return", "round", "scale", "shape", "size", "width"], "groupvit/modeling_groupvit.py:get_grouping_from_attentions": ["Model_grouping_from_attentions", "None", "append", "attentions", "attn_maps", "attn_masks", "contiguous", "cur_attn_map", "def", "else", "final_grouping", "for", "hw_shape", "if", "in", "is", "no_grad", "permute", "prev_attn_masks", "resize_attention_map", "return", "torch", "with"], "groupvit/modeling_groupvit.py:GroupViTCrossAttentionLayer": ["LayerNorm", "ModelAttention", "ModelCrossAttentionLayer", "ModelMLP", "ModelVisionConfig", "Module", "__init__", "attn", "class", "config", "def", "encoder_hidden_states", "eps", "forward", "hidden_size", "key", "layer_norm_eps", "mlp", "nn", "norm2", "norm_post", "query", "return", "self", "super", "x"], "groupvit/modeling_groupvit.py:GroupViTAssignAttention": ["False", "Linear", "ModelAssignAttention", "ModelVisionConfig", "Module", "True", "__init__", "and", "assign_eps", "attn", "class", "config", "def", "dim", "else", "forward", "functional", "get_attn", "gumbel", "gumbel_softmax", "hard", "hard_softmax", "hidden_size", "if", "k_proj", "keepdim", "key", "nn", "out", "proj", "q_proj", "query", "raw_attn", "return", "scale", "self", "soft_attn", "softmax", "sum", "super", "training", "transpose", "v_proj", "value"], "groupvit/modeling_groupvit.py:GroupViTTokenAssign": ["Iterable", "LayerNorm", "ModelAssignAttention", "ModelCrossAttentionLayer", "ModelMLP", "ModelMixerMLP", "ModelTokenAssign", "ModelVisionConfig", "Model_tokens", "Module", "__init__", "abc", "assign", "assign_mlp_ratio", "attention", "channels_dim", "class", "collections", "config", "def", "else", "eps", "for", "forward", "hidden_size", "if", "image_tokens", "in", "int", "isinstance", "layer_norm_eps", "mlp_channels", "mlp_inter", "new_image_tokens", "nn", "norm_new_x", "norm_post_tokens", "norm_tokens", "norm_x", "num_Model_token", "num_output_Model", "pre_assign_attn", "project_Model_token", "projected_Model_tokens", "return", "self", "super", "tokens_dim", "x"], "groupvit/modeling_groupvit.py:GroupViTModelOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "segmentation_logits", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "groupvit/modeling_groupvit.py:GroupViTPatchEmbeddings": ["Conv2d", "False", "Input", "Iterable", "ModelPatchEmbeddings", "Module", "Tensor", "Union", "ValueError", "__init__", "abc", "batch_size", "bool", "class", "collections", "def", "doesn", "else", "embed_dim", "f", "flatten", "forward", "height", "if", "image", "image_size", "int", "interpolate_pos_encoding", "isinstance", "kernel_size", "match", "model", "nn", "not", "num_channels", "num_patches", "or", "patch_size", "pixel_values", "projection", "raise", "return", "self", "shape", "size", "stride", "super", "t", "torch", "transpose", "tuple", "width", "x"], "groupvit/modeling_groupvit.py:GroupViTVisionEmbeddings": ["Dropout", "False", "LayerNorm", "ModelPatchEmbeddings", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "class", "config", "def", "dim", "dropout", "else", "embed_dim", "embeddings", "eps", "forward", "functional", "height", "hidden_size", "if", "image_size", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "layer_norm_eps", "layernorm", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "view", "width", "zeros"], "groupvit/modeling_groupvit.py:GroupViTTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "groupvit/modeling_groupvit.py:GroupViTStage": ["False", "FloatTensor", "LayerNorm", "ModelEncoderLayer", "ModelMixerMLP", "ModelStage", "ModelTokenAssign", "ModelVisionConfig", "Model_projector", "Model_token", "Module", "ModuleList", "None", "Optional", "Parameter", "Sequential", "Tensor", "_", "__init__", "and", "attention", "attention_mask", "bool", "cat", "cat_x", "causal_attention_mask", "class", "concat_x", "config", "def", "depth", "dim", "downsample", "else", "eps", "expand", "for", "forward", "hidden_size", "hidden_states", "if", "in", "int", "is", "layer", "layer_norm_eps", "layer_out", "layers", "nn", "not", "num_Model_token", "num_output_Model", "num_prev_Model_token", "output_attentions", "outputs", "prev_Model_token", "property", "range", "return", "self", "size", "split_x", "super", "torch", "tuple", "with_Model_token", "x", "zeros"], "groupvit/modeling_groupvit.py:GroupViTMLP": ["ACT2FN", "Linear", "ModelMLP", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "__init__", "activation_fn", "class", "config", "def", "else", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "int", "intermediate_size", "is", "nn", "not", "output_size", "return", "self", "super", "torch"], "groupvit/modeling_groupvit.py:GroupViTMixerMLP": ["ModelMLP", "ModelMixerMLP", "class", "def", "forward", "return", "self", "super", "transpose", "x"], "groupvit/modeling_groupvit.py:GroupViTAttention": ["Attention", "False", "FloatTensor", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_shape", "and", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bmm", "bool", "bsz", "but", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "dim", "divisible", "dropout", "else", "embed_dim", "encoder_hidden_states", "f", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "k_proj", "key_states", "mask", "must", "nn", "not", "num_attention_heads", "num_heads", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scale", "self", "seq_len", "should", "size", "softmax", "src_len", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights"], "groupvit/modeling_groupvit.py:GroupViTEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "groupvit/modeling_groupvit.py:GroupViTPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "ModelPreTrainedModel", "ModelTextEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "init_range", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "token_embedding", "v_proj", "weight", "zero_"], "groupvit/modeling_groupvit.py:GroupViTVisionEncoder": ["BaseModelOutput", "False", "ModelStage", "ModelVisionConfig", "ModelVisionEncoder", "Model_tokens", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "__init__", "all_Modelings", "all_hidden_states", "and", "attentions", "bool", "class", "config", "def", "depth", "depths", "else", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer_outputs", "len", "nn", "not", "num_Model_token", "num_Model_tokens", "num_output_Model", "num_output_Models", "num_prev_Model_token", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "stage", "stages", "super", "torch", "tuple", "use_return_dict", "v"], "groupvit/modeling_groupvit.py:GroupViTTextEncoder": ["BaseModelOutput", "False", "ModelEncoderLayer", "ModelTextConfig", "ModelTextEncoder", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "groupvit/modeling_groupvit.py:GroupViTTextTransformer": ["BaseModelOutputWithPooling", "LayerNorm", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eos_token_id", "eps", "final_layer_norm", "forward", "have", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "raise", "return", "return_dict", "self", "shape", "size", "specify", "super", "to", "torch", "tuple", "use_return_dict", "view"], "groupvit/modeling_groupvit.py:GroupViTTextModel": ["BaseModelOutputWithPooling", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "tuple", "value"], "groupvit/modeling_groupvit.py:GroupViTVisionTransformer": ["BaseModelOutputWithPooling", "FloatTensor", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "dim", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "is", "last_hidden_state", "layer_norm_eps", "layernorm", "mean", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "groupvit/modeling_groupvit.py:GroupViTVisionModel": ["BaseModelOutputWithPooling", "FloatTensor", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "main_input_name", "output_attentions", "output_hidden_states", "patch_embeddings", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "vision_model"], "groupvit/modeling_groupvit.py:GroupViTModel": ["BaseModelOutputWithPooling", "BatchNorm1d", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextTransformer", "ModelVisionConfig", "ModelVisionTransformer", "Model_loss", "Modeling", "None", "Optional", "Parameter", "ReLU", "Sequential", "Tensor", "True", "TypeError", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "be", "bias", "bool", "but", "class", "config", "def", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "flatten_Modeling", "forward", "get_Modeling_from_attentions", "get_image_features", "get_text_features", "hidden_size", "if", "image_Model_embeds", "image_embeds", "image_features", "inplace", "input_ids", "is", "isinstance", "keepdim", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_image_Model", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "output", "output_attentions", "output_hidden_states", "output_segmentation", "permute", "pixel_values", "pooler_output", "position_ids", "post_init", "projection_dim", "projection_intermediate_dim", "r", "raise", "reshape", "return", "return_dict", "return_loss", "seg_logits", "segmentation_logits", "self", "shape", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "use_return_dict", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "mt5/modeling_mt5.py:MT5LayerNorm": ["Model", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "mt5/modeling_mt5.py:MT5DenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "mt5/modeling_mt5.py:MT5DenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "mt5/modeling_mt5.py:MT5LayerFF": ["DenseReluDense", "Dropout", "Model", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelLayerNorm", "Module", "__init__", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is_gated_act", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "mt5/modeling_mt5.py:MT5Attention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Please", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "if", "index", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "len", "list", "log", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "union", "unsqueeze", "update", "use_cache", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "mt5/modeling_mt5.py:MT5LayerSelfAttention": ["Dropout", "False", "Model", "ModelAttention", "ModelLayerNorm", "Module", "None", "Optional", "SelfAttention", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "mt5/modeling_mt5.py:MT5LayerCrossAttention": ["Dropout", "EncDecAttention", "False", "Model", "ModelAttention", "ModelLayerNorm", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "mt5/modeling_mt5.py:MT5Block": ["False", "GradientCheckpointingLayer", "Model", "ModelLayerCrossAttention", "ModelLayerFF", "ModelLayerSelfAttention", "ModuleList", "None", "Optional", "True", "__init__", "and", "any", "append", "attention_mask", "attention_outputs", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "encoder_attention_mask", "encoder_decoder_position_bias", "encoder_hidden_states", "finfo", "float16", "forward", "has_relative_attention_bias", "hidden_states", "if", "int", "is", "is_decoder", "isinf", "key_value_states", "layer", "layer_head_mask", "layer_idx", "max", "min", "new_name", "nn", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "return_dict", "self", "self_attention_outputs", "super", "torch", "use_cache", "version", "where"], "mt5/modeling_mt5.py:MT5ClassificationHead": ["Dropout", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "classifier_dropout", "config", "d_model", "def", "dense", "dropout", "forward", "hidden_states", "nn", "num_labels", "out_proj", "p", "return", "self", "super", "tanh", "torch"], "mt5/modeling_mt5.py:MT5PreTrainedModel": ["DUMMY_INPUTS", "DUMMY_MASK", "In", "Model", "ModelAttention", "ModelBlock", "ModelClassificationHead", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelEncoderModel", "ModelForConditionalGeneration", "ModelForQuestionAnswering", "ModelForTokenClassification", "ModelLayerNorm", "ModelModel", "None", "PreTrainedModel", "See", "True", "ValueError", "_can_compile_fullgraph", "_init_weights", "_keep_in_fp32_modules", "_no_split_modules", "_shift_right", "and", "base_model_prefix", "be", "bias", "cat", "class", "classifier", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "def", "defined", "dense", "dim", "docs", "dummy_inputs", "elif", "else", "factor", "fill_", "for", "full", "has", "has_relative_attention_bias", "hasattr", "if", "information", "initializer_factor", "input_ids", "input_mask", "is", "is_parallelizable", "is_torch_fx_proxy", "isinstance", "it", "k", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "normal_", "not", "num_heads", "o", "out_proj", "pad_token_id", "property", "q", "qa_outputs", "raise", "relative_attention_bias", "return", "self", "set", "shape", "shared", "shifted_input_ids", "std", "supports_gradient_checkpointing", "tensor", "the", "tie_word_embeddings", "to", "torch", "transformer", "usually", "v", "weight", "wi", "wi_0", "wi_1", "wo", "zero_"], "mt5/modeling_mt5.py:MT5Stack": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "DEPARALLELIZE_DOCSTRING", "Dropout", "DynamicCache", "EncoderDecoderCache", "False", "FutureWarning", "Like", "Model", "ModelBlock", "ModelLayerNorm", "ModelPreTrainedModel", "ModuleList", "None", "PARALLELIZE_DOCSTRING", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "a", "add_start_docstrings", "all_attentions", "all_cross_attentions", "all_hidden_states", "also", "and", "any", "arange", "as", "assert_device_map", "at", "attention_mask", "attentions", "balanced", "batch_size", "be", "block", "bool", "both", "but", "cache_position", "call", "can", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cpu", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "cuda_device", "d_model", "decoder", "decoder_", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "diagonal", "dictionary", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "empty_cache", "encoder_attention_mask", "encoder_batch_size", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "first_device", "flash_attention_2", "flex_attention", "for", "forward", "from_pretrained", "full", "get_device_map", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_relative_attention_bias", "have", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "instance", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_torchdynamo_compiling", "is_training", "isinstance", "it", "items", "k", "keys", "kwargs", "last_device", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "len", "load", "logger", "long", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "max", "min", "min_dtype", "model", "model_parallel", "module_name", "needs", "new_embeddings", "nn", "not", "npu", "num_layers", "of", "ones", "only", "or", "output_attentions", "output_hidden_states", "own", "padding_mask", "parallelize", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_bias", "post_init", "provide", "raise", "range", "removed", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set", "set_device", "set_input_embeddings", "shape", "should", "size", "so", "specify", "staticmethod", "str", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "used", "using_compilable_cache", "v", "v5", "valid", "view", "warn", "warning_once", "warnings", "will", "with", "xpu", "you", "your"], "mt5/modeling_mt5.py:MT5Model": ["BaseModelOutput", "BoolTensor", "Cache", "DEPARALLELIZE_DOCSTRING", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "Like", "LongTensor", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "PARALLELIZE_DOCSTRING", "Seq2SeqModelOutput", "T5Model", "Tensor", "Transformers", "True", "Union", "You", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_prune_heads", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "block", "bool", "but", "cache_position", "call", "can", "class", "config", "copy", "cpu", "cross_attentions", "cross_attn_head_mask", "cuda", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "elif", "else", "embed_tokens", "empty_cache", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "first_device", "for", "forward", "from_pretrained", "get_device_map", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instance", "is", "is_decoder", "isinstance", "it", "items", "last_hidden_state", "layer", "len", "load", "model", "model_parallel", "model_type", "module_name", "needs", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "of", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "post_init", "provide", "prune_heads", "r", "range", "relative_attention_bias", "removed", "return", "return_dict", "self", "set_device", "set_input_embeddings", "shared", "should", "so", "super", "the", "tie_encoder_decoder", "to", "torch", "tuple", "use_cache", "use_return_dict", "v5", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "mt5/modeling_mt5.py:MT5ForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "Cache", "CrossEntropyLoss", "DEPARALLELIZE_DOCSTRING", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Like", "Linear", "LongTensor", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "PARALLELIZE_DOCSTRING", "Seq2SeqLMOutput", "T5ForConditionalGeneration", "Tensor", "Transformers", "True", "Union", "You", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "bias", "block", "bool", "but", "cache_position", "call", "can", "class", "config", "copy", "cpu", "cross_attentions", "cross_attn_head_mask", "cuda", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "elif", "else", "embed_tokens", "empty_cache", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "first_device", "for", "forward", "from_pretrained", "get_device_map", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "instance", "is", "is_decoder", "isinstance", "it", "labels", "last_hidden_state", "layer", "len", "lm_head", "lm_logits", "load", "logits", "loss", "loss_fct", "model", "model_dim", "model_parallel", "model_type", "module_name", "needs", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "of", "output", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provide", "r", "range", "relative_attention_bias", "removed", "return", "return_dict", "self", "sequence_output", "set_device", "set_input_embeddings", "shared", "should", "size", "so", "super", "the", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "v5", "view", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "mt5/modeling_mt5.py:MT5EncoderModel": ["BaseModelOutput", "DEPARALLELIZE_DOCSTRING", "Embedding", "False", "FloatTensor", "FutureWarning", "Like", "LongTensor", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "PARALLELIZE_DOCSTRING", "SelfAttention", "T5EncoderModel", "Transformers", "True", "Union", "You", "__init__", "_prune_heads", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "auto_docstring", "balanced", "be", "block", "bool", "but", "call", "can", "class", "config", "cpu", "cuda", "d_model", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "else", "embed_tokens", "empty_cache", "encoder", "encoder_config", "encoder_outputs", "for", "forward", "from_pretrained", "get_device_map", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "instance", "is", "is_encoder_decoder", "it", "items", "layer", "len", "load", "model", "model_parallel", "model_type", "module_name", "needs", "new_embeddings", "nn", "not", "of", "output_attentions", "output_hidden_states", "own", "parallelize", "post_init", "provide", "prune_heads", "r", "range", "removed", "return", "return_dict", "self", "set_input_embeddings", "shared", "should", "so", "super", "the", "to", "torch", "tuple", "use_cache", "use_return_dict", "v5", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "mt5/modeling_mt5.py:MT5ForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "EncDecAttention", "False", "FloatTensor", "If", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Please", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "_", "__class__", "__init__", "__name__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tied_weights_keys", "and", "are", "attention_mask", "auto_docstring", "batch_size", "be", "block", "bool", "cannot", "class", "classification_head", "config", "cross_attentions", "cross_attn_head_mask", "currently", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_size", "if", "input", "input_ids", "inputs_embeds", "int", "is", "labels", "layer", "len", "list", "logits", "long", "loss", "loss_fct", "model_parallel", "multi_label_classification", "must", "no", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pass", "passed", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "relative_attention_bias", "return", "return_dict", "same", "self", "sentence_representation", "sequence_output", "shape", "single_label_classification", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "transformer", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "mt5/modeling_mt5.py:MT5ForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelConfig", "ModelEncoderModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "embed_tokens", "encoder", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "r", "return", "return_dict", "self", "super", "torch", "transformer", "tuple", "use_return_dict", "view", "weight"], "mt5/modeling_mt5.py:MT5ForQuestionAnswering": ["BaseModelOutput", "BoolTensor", "CrossEntropyLoss", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "If", "Linear", "LongTensor", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Please", "Seq2SeqQuestionAnsweringModelOutput", "Tensor", "True", "Union", "ValueError", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tied_weights_keys", "and", "are", "attention_mask", "attentions", "auto_docstring", "be", "block", "bool", "cannot", "clamp", "class", "config", "contiguous", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "device", "dim", "either", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "last_hidden_state", "layer", "len", "logits", "loss", "loss_fct", "model_dim", "model_parallel", "new_embeddings", "nn", "no", "not", "num_decoder_layers", "num_labels", "num_layers", "or", "output", "output_attentions", "output_hidden_states", "pass", "passed", "past_key_values", "post_init", "qa_outputs", "r", "raise", "relative_attention_bias", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "shared", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "tie_encoder_decoder", "to", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "vocab_size", "warn", "warnings", "weight"], "mgp_str/modeling_mgp_str.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "mgp_str/modeling_mgp_str.py:MgpstrDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "mgp_str/modeling_mgp_str.py:MgpstrModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "a3_attentions", "attentions", "class", "hidden_states", "logits", "r", "torch", "tuple"], "mgp_str/modeling_mgp_str.py:MgpstrEmbeddings": ["Conv2d", "Dropout", "Input", "Iterable", "ModelConfig", "ModelEmbeddings", "Module", "Parameter", "ValueError", "__init__", "abc", "batch_size", "cat", "channel", "class", "cls_token", "cls_tokens", "collections", "config", "def", "dim", "distilled", "doesn", "drop_rate", "else", "embedding_output", "expand", "f", "flatten", "forward", "grid_size", "height", "hidden_size", "if", "image", "image_size", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "num_tokens", "or", "p", "patch_embeddings", "patch_size", "pixel_values", "pos_drop", "pos_embed", "proj", "raise", "return", "self", "shape", "size", "stride", "super", "t", "torch", "transpose", "width", "zeros"], "mgp_str/modeling_mgp_str.py:MgpstrMlp": ["Dropout", "GELU", "Linear", "ModelConfig", "ModelMlp", "Module", "__init__", "act", "class", "config", "def", "drop", "drop_rate", "fc1", "fc2", "forward", "hidden_features", "hidden_size", "hidden_states", "nn", "or", "return", "self", "super"], "mgp_str/modeling_mgp_str.py:MgpstrAttention": ["Dropout", "Linear", "ModelAttention", "ModelConfig", "Module", "__init__", "attention_probs", "attn_drop", "attn_drop_rate", "batch_size", "bias", "channel", "class", "config", "context_layer", "def", "dim", "drop_rate", "forward", "head_dim", "hidden_size", "hidden_states", "key", "nn", "num", "num_attention_heads", "num_heads", "permute", "proj", "proj_drop", "qkv", "qkv_bias", "query", "reshape", "return", "scale", "self", "shape", "softmax", "super", "transpose", "value"], "mgp_str/modeling_mgp_str.py:MgpstrLayer": ["Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelMlp", "Module", "None", "__init__", "attention_output", "attn", "class", "config", "def", "drop_path", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "int", "is", "layer_norm_eps", "layer_output", "mlp", "mlp_hidden_dim", "mlp_ratio", "nn", "norm1", "norm2", "not", "outputs", "return", "self", "self_attention_outputs", "super"], "mgp_str/modeling_mgp_str.py:MgpstrEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "None", "Sequential", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "blk", "blocks", "class", "config", "cpu", "def", "device", "dpr", "drop_path", "drop_path_rate", "else", "enumerate", "for", "forward", "hidden_states", "i", "if", "in", "is", "item", "last_hidden_state", "layer_outputs", "linspace", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v", "x"], "mgp_str/modeling_mgp_str.py:MgpstrA3Module": ["Conv2d", "F", "False", "LayerNorm", "ModelA3Module", "ModelConfig", "Module", "Sequential", "__init__", "a3_out", "attentions", "bias", "class", "config", "def", "dim", "einsum", "eps", "feat", "flatten", "forward", "groups", "hidden_size", "hidden_states", "id", "kernel_size", "layer_norm_eps", "max_token_length", "nn", "norm", "return", "sd", "selected", "self", "si", "softmax", "stride", "super", "tokenLearner", "token_norm", "torch", "transpose", "unsqueeze"], "mgp_str/modeling_mgp_str.py:MgpstrPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "not", "pos_embed", "self", "std", "trunc_normal_", "weight", "zero_"], "mgp_str/modeling_mgp_str.py:MgpstrModel": ["BaseModelOutput", "FloatTensor", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Union", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "forward", "get_input_embeddings", "have", "hidden_states", "if", "is", "last_hidden_state", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "proj", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "mgp_str/modeling_mgp_str.py:MgpstrForSceneTextRecognition": ["FloatTensor", "Linear", "Model", "ModelA3Module", "ModelConfig", "ModelForSceneTextRecognition", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "a3_attentions", "all_a3_attentions", "all_logits", "attentions", "auto_docstring", "bool", "bpe_a3_module", "bpe_a3_out", "bpe_attention", "bpe_head", "bpe_logits", "char_a3_module", "char_a3_out", "char_attention", "char_head", "char_logits", "class", "config", "def", "else", "for", "forward", "hidden_size", "hidden_states", "if", "in", "is", "logits", "main_input_name", "mgp_outputs", "nn", "not", "num_bpe_labels", "num_character_labels", "num_labels", "num_wordpiece_labels", "output", "output_a3_attentions", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict", "wp_a3_module", "wp_a3_out", "wp_attention", "wp_head", "wp_logits"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Embeddings": ["Dropout", "Embedding", "False", "IndexError", "LayerNorm", "ModelEmbeddings", "Module", "The", "__init__", "_calc_spatial_position_embeddings", "arange", "as", "bbox", "be", "cat", "class", "config", "coordinate", "coordinate_size", "def", "dim", "dropout", "e", "eps", "except", "expand", "from", "h_position_embeddings", "hidden_dropout_prob", "hidden_size", "layer_norm_eps", "left_position_embeddings", "lower_position_embeddings", "max_2d_position_embeddings", "max_position_embeddings", "nn", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "raise", "range", "register_buffer", "return", "right_position_embeddings", "self", "shape_size", "should", "spatial_position_embeddings", "super", "token_type_embeddings", "torch", "try", "type_vocab_size", "upper_position_embeddings", "values", "vocab_size", "w_position_embeddings", "within", "word_embeddings", "x_position_embeddings", "y_position_embeddings"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2SelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "Parameter", "The", "ValueError", "_", "__init__", "_sz", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "chunk", "class", "compute_qkv", "config", "context_layer", "contiguous", "def", "dim", "dropout", "dtype", "else", "embedding_size", "f", "fast_qkv", "finfo", "float", "float32", "forward", "functional", "has_relative_attention_bias", "has_spatial_attention_bias", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "k", "key", "key_layer", "masked_fill_", "math", "matmul", "min", "multiple", "ndimension", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "q", "q_bias", "qkv", "qkv_linear", "query", "query_layer", "raise", "rel_2d_pos", "rel_pos", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "to", "torch", "transpose", "type_as", "v", "v_bias", "value", "value_layer", "view", "zeros"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Attention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "attention_mask", "attention_output", "class", "config", "def", "forward", "head_mask", "hidden_states", "nn", "output", "output_attentions", "outputs", "rel_2d_pos", "rel_pos", "return", "self", "self_outputs", "super"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2SelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Intermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Output": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Layer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "outputs", "rel_2d_pos", "rel_pos", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "layoutlmv2/modeling_layoutlmv2.py:relative_position_bucket": ["Model_position", "Model_position_bucket", "True", "abs", "bidirectional", "def", "else", "float", "full_like", "if", "is_small", "log", "long", "math", "max", "max_distance", "max_exact", "min", "n", "num_buckets", "ret", "return", "to", "torch", "val_if_large", "where", "zeros_like"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Encoder": ["BaseModelOutput", "False", "Linear", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "_calculate_1d_position_embeddings", "_calculate_2d_position_embeddings", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bbox", "bias", "class", "config", "contiguous", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "has_relative_attention_bias", "has_spatial_attention_bias", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "max_distance", "max_rel_2d_pos", "max_rel_pos", "nn", "no_grad", "not", "num_attention_heads", "num_buckets", "num_hidden_layers", "output_attentions", "output_hidden_states", "permute", "position_coord_x", "position_coord_y", "position_ids", "range", "rel_2d_pos", "rel_2d_pos_bins", "rel_pos", "rel_pos_bias", "rel_pos_bins", "rel_pos_mat", "rel_pos_x", "rel_pos_x_2d_mat", "rel_pos_x_bias", "rel_pos_y", "rel_pos_y_2d_mat", "rel_pos_y_bias", "relative_position_bucket", "return", "return_dict", "self", "super", "t", "torch", "tuple", "unsqueeze", "v", "weight", "with"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2PreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fast_qkv", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "q_bias", "self", "std", "v_bias", "visual_segment_embedding", "weight", "zero_"], "layoutlmv2/modeling_layoutlmv2.py:my_convert_sync_batchnorm": ["FrozenBatchNorm2d", "None", "Parameter", "SyncBatchNorm", "True", "_BatchNorm", "add_module", "affine", "batchnorm", "bias", "child", "convert_sync_batchnorm", "def", "del", "detectron2", "device", "dtype", "eps", "for", "if", "in", "isinstance", "layers", "long", "module", "module_output", "modules", "my_convert_sync_batchnorm", "name", "named_children", "nn", "num_batches_tracked", "num_features", "process_group", "return", "running_mean", "running_var", "tensor", "torch", "track_running_stats", "weight"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2VisualBackbone": ["AdaptiveAvgPool2d", "AvgPool2d", "FPN", "False", "META_ARCHITECTURE", "META_ARCH_REGISTRY", "MODEL", "Make", "ModelVisualBackbone", "Module", "PIXEL_MEAN", "PIXEL_STD", "RuntimeError", "Tensor", "__init__", "and", "append", "are_deterministic_algorithms_enabled", "assert", "backbone", "backbone_stride", "be", "by", "can", "ceil", "cfg", "channels", "class", "config", "contiguous", "cuda", "def", "detectron2", "device_count", "distributed", "divided", "else", "features", "flatten", "for", "forward", "get", "get_detectron2_config", "get_rank", "get_world_size", "i", "if", "image_feature_pool_shape", "images", "images_input", "in", "input_shape", "instead", "is", "is_available", "is_initialized", "is_tensor", "isinstance", "len", "list", "logger", "math", "meta_arch", "model", "modeling", "my_convert_sync_batchnorm", "new_group", "nn", "node_global_ranks", "node_rank", "node_size", "nodes", "not", "num_channels", "number", "of", "out_feature_key", "output_shape", "p2", "persistent", "pixel_mean", "pixel_std", "pool", "process_group", "processes", "properly", "raise", "range", "ranks", "register_buffer", "return", "self", "self_rank", "set", "start_dim", "stride", "super", "sure", "sync_bn_groups", "synchronize_batch_norm", "tensor", "the", "torch", "transpose", "up", "using", "view", "warning", "world_size"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Pooler": ["Linear", "ModelPooler", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2Model": ["BaseModelOutputWithPooling", "Dropout", "Embedding", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "ModelVisualBackbone", "None", "Optional", "Parameter", "Size", "Union", "ValueError", "You", "__init__", "_calc_img_embeddings", "_calc_spatial_position_embeddings", "_calc_text_embeddings", "_calc_visual_bbox", "_get_input_shape", "and", "arange", "at", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "both", "cannot", "cat", "class", "config", "def", "detectron2", "device", "dim", "div", "dropout", "dtype", "either", "elif", "else", "embeddings", "encoder", "encoder_outputs", "eps", "expand", "expand_as", "extended_attention_mask", "final_attention_mask", "final_bbox", "final_emb", "final_position_ids", "final_shape", "finfo", "floor", "forward", "get_input_embeddings", "has_visual_segment_embedding", "have", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image", "image_feature_pool_shape", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "layer_norm_eps", "list", "long", "min", "next", "nn", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "parameters", "pooled_output", "pooler", "pooler_output", "position_embeddings", "position_ids", "post_init", "r", "raise", "repeat", "requires_backends", "return", "return_dict", "rounding_mode", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "spatial_position_embeddings", "specify", "stack", "super", "text_Model_emb", "the", "time", "to", "token_type_embeddings", "token_type_ids", "torch", "transpose", "tuple", "unsqueeze", "use_return_dict", "value", "view", "visual", "visual_LayerNorm", "visual_attention_mask", "visual_bbox", "visual_bbox_x", "visual_bbox_y", "visual_dropout", "visual_emb", "visual_embeddings", "visual_position_ids", "visual_proj", "visual_segment_embedding", "visual_shape", "weight", "word_embeddings", "zeros", "zeros_like"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2ForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Size", "Union", "ValueError", "You", "__init__", "_calc_img_embeddings", "_calc_visual_bbox", "and", "arange", "at", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "both", "cannot", "cat", "class", "classifier", "cls_final_output", "config", "def", "device", "dim", "dropout", "dtype", "either", "elif", "else", "embeddings", "final_image_embeddings", "final_shape", "forward", "get_input_embeddings", "have", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image", "image_feature_pool_shape", "initial_image_embeddings", "input_ids", "input_shape", "inputs_embeds", "int", "is", "labels", "list", "logits", "long", "loss", "loss_fct", "mean", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_final_image_embeddings", "pooled_initial_image_embeddings", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "repeat", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "single_label_classification", "size", "specify", "squeeze", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "visual_bbox", "visual_position_ids", "visual_shape", "warn_if_padding_and_no_attention_mask", "word_embeddings"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2ForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "class", "classifier", "config", "def", "dropout", "else", "embeddings", "forward", "get_input_embeddings", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image", "input_ids", "input_shape", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "seq_length", "sequence_output", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "word_embeddings"], "layoutlmv2/modeling_layoutlmv2.py:LayoutLMv2ForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "True", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "embeddings", "end_logits", "end_loss", "end_positions", "forward", "get_input_embeddings", "has_visual_segment_embedding", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "image", "input_ids", "input_shape", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "seq_length", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "word_embeddings"], "mllama/modeling_mllama.py:_prepare_cross_attention_mask": ["None", "Tensor", "_", "_prepare_cross_attention_mask", "any", "batch_size", "bool", "cross_attention_mask", "def", "dim", "dtype", "finfo", "full_text_row_masked_out_mask", "int", "inverted_cross_attn_mask", "masked_fill", "min", "negative_inf_value", "num_vision_tokens", "repeat_interleave", "return", "shape", "str", "text_total_length", "to", "torch", "tuple", "type_as", "unsqueeze", "view"], "mllama/modeling_mllama.py:_prepare_aspect_ratio_attention_mask": ["Tensor", "_prepare_aspect_ratio_attention_mask", "aspect_ratio_mask", "attention_mask", "batch_size", "def", "dtype", "finfo", "int", "max_num_tiles", "min", "num_patches", "pad_patches", "repeat", "reshape", "return", "shape", "target_length", "to", "torch", "transpose", "unsqueeze", "view"], "mllama/modeling_mllama.py:MllamaPrecomputedAspectRatioEmbedding": ["Embedding", "ModelPrecomputedAspectRatioEmbedding", "ModelVisionConfig", "Module", "Parameter", "Tensor", "True", "__init__", "aspect_ratio_ids", "bool", "class", "config", "def", "embedding", "embeddings", "forward", "gate", "hidden_size", "hidden_state", "if", "is_gated", "max_aspect_ratio_id", "max_num_tiles", "nn", "reshape", "return", "self", "super", "tanh", "torch", "zeros"], "mllama/modeling_mllama.py:MllamaPrecomputedPositionEmbedding": ["Embedding", "ModelPrecomputedPositionEmbedding", "ModelVisionConfig", "Module", "Parameter", "Tensor", "__init__", "aspect_ratio_ids", "batch_size", "class", "config", "def", "embedding", "forward", "gate", "gated_position_embedding", "gated_tile_position_embedding", "hidden_size", "hidden_state", "image_size", "max_aspect_ratio_id", "max_num_tiles", "nn", "num_patches", "patch_size", "position_embedding", "randn", "reshape", "return", "scale", "self", "shape", "super", "tanh", "tile_embedding", "tile_position_embedding", "torch", "view", "zeros"], "mllama/modeling_mllama.py:MllamaVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "mllama/modeling_mllama.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "mllama/modeling_mllama.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "mllama/modeling_mllama.py:MllamaVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "_", "__init__", "_attn_implementation", "attention_heads", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "embed_dim", "forward", "head_dim", "hidden_size", "hidden_state", "if", "k_proj", "key", "kv_seq_len", "kwargs", "nn", "num_heads", "num_key_value_groups", "o_proj", "q_proj", "q_seq_len", "query", "reshape", "return", "scaling", "self", "shape", "super", "torch", "transpose", "tuple", "v_proj", "value", "view"], "mllama/modeling_mllama.py:MllamaVisionEncoderLayer": ["False", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoderLayer", "ModelVisionMLP", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "attention_heads", "attention_mask", "attn_weights", "bool", "class", "config", "def", "eps", "forward", "gate_attn", "gate_ffn", "hidden_size", "hidden_state", "if", "input_layernorm", "intermediate_size", "is_gated", "math", "mlp", "nn", "norm_eps", "num_attention_heads", "ones", "pi", "post_attention_layernorm", "residual", "return", "self", "self_attn", "super", "tanh", "torch"], "mllama/modeling_mllama.py:MllamaVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "class", "config", "def", "encoder_layer", "encoder_states", "for", "forward", "gradient_checkpointing", "hidden_state", "hidden_states", "in", "is_gated", "last_hidden_state", "layers", "nn", "num_layers", "r", "range", "return", "self", "super", "torch"], "mllama/modeling_mllama.py:MllamaTextRMSNorm": ["ModelTextRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "mllama/modeling_mllama.py:MllamaTextCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Cross", "False", "Linear", "LongTensor", "ModelTextConfig", "ModelTextCrossAttention", "ModelTextRMSNorm", "Module", "None", "Optional", "Tensor", "ValueError", "_", "__init__", "_attn_implementation", "attention", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "bsz", "cache_position", "cached", "can", "class", "config", "contiguous", "cross_attention_states", "cross_attn_states", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "elif", "else", "eps", "find", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_norm", "k_proj", "key", "key_states", "keys", "kwargs", "layer", "layer_idx", "layers", "neither", "new_name", "nn", "nor", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "q_len", "q_norm", "q_proj", "query_states", "raise", "reshape", "return", "rms_norm_eps", "scaling", "self", "size", "super", "t", "torch", "training", "transpose", "tuple", "update", "use_cache", "v_proj", "value_states", "values", "version", "view"], "mllama/modeling_mllama.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "mllama/modeling_mllama.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "mllama/modeling_mllama.py:MllamaTextSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelTextConfig", "ModelTextSelfAttention", "Module", "None", "Tensor", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_len", "q_proj", "query_states", "reshape", "return", "rope_theta", "scaling", "self", "sin", "size", "super", "torch", "training", "transpose", "update", "use_cache", "v_proj", "value_states", "version", "view"], "mllama/modeling_mllama.py:MllamaTextMLP": ["ACT2FN", "False", "Linear", "ModelTextMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "mllama/modeling_mllama.py:MllamaSelfAttentionDecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelSelfAttentionDecoderLayer", "ModelTextConfig", "ModelTextMLP", "ModelTextRMSNorm", "ModelTextSelfAttention", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "cross_attention_mask", "cross_attention_states", "def", "deprecate_kwarg", "eps", "forward", "full_text_row_masked_out_mask", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "mllama/modeling_mllama.py:MllamaCrossAttentionDecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "GradientCheckpointingLayer", "LongTensor", "ModelCrossAttentionDecoderLayer", "ModelTextConfig", "ModelTextCrossAttention", "ModelTextMLP", "ModelTextRMSNorm", "None", "Optional", "Parameter", "Tensor", "Unpack", "__init__", "attention_mask", "attn_weights", "bool", "cache_position", "class", "config", "cross_attention_mask", "cross_attention_states", "cross_attn", "cross_attn_attn_gate", "cross_attn_mlp_gate", "def", "deprecate_kwarg", "eps", "forward", "full_text_row_masked_out_mask", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is", "kwargs", "layer_idx", "mlp", "new_name", "nn", "not", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "super", "tanh", "torch", "tuple", "use_cache", "version", "zeros"], "mllama/modeling_mllama.py:MllamaRotaryEmbedding": ["False", "ModelRotaryEmbedding", "ModelTextConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "mllama/modeling_mllama.py:MllamaPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Conv2d", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelCrossAttentionDecoderLayer", "ModelPreTrainedModel", "ModelPrecomputedAspectRatioEmbedding", "ModelPrecomputedPositionEmbedding", "ModelSelfAttentionDecoderLayer", "ModelTextCrossAttention", "ModelTextRMSNorm", "ModelTextSelfAttention", "ModelVisionEncoderLayer", "ModelVisionModel", "None", "OutputRecorder", "PreTrainedModel", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_can_record_outputs", "_ignore_causal_mask_sdpa", "_init_weights", "_no_split_modules", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "and", "any", "arange", "attention_mask", "attentions", "base_model_prefix", "batch_size", "bias", "bool", "cache_position", "causal_mask", "class", "class_embedding", "clone", "config", "cross_attn", "cross_attn_attn_gate", "cross_attn_mlp_gate", "cuda", "data", "def", "device", "diagonal", "dim", "dtype", "elif", "else", "embedding", "expand", "fill_", "fill_value", "finfo", "flash_attention_2", "flex_attention", "full", "gate", "gate_attn", "gate_ffn", "get_max_cache_shape", "get_seq_length", "get_text_config", "getattr", "hidden_states", "if", "in", "index", "init", "initializer_range", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_gated", "is_training", "isinstance", "kwargs", "layer_name", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "module", "nn", "normal_", "not", "npu", "output_attentions", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "reshape", "return", "sdpa", "self", "self_attn", "sequence_length", "shape", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_", "zeros_"], "mllama/modeling_mllama.py:MllamaVisionModel": ["BaseModelOutput", "Conv2d", "F", "False", "LayerNorm", "ModelPreTrainedModel", "ModelPrecomputedAspectRatioEmbedding", "ModelPrecomputedPositionEmbedding", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionModel", "None", "Parameter", "Tensor", "True", "_", "__init__", "_prepare_aspect_ratio_attention_mask", "all_intermediate_hidden_states", "apply_class_embedding", "aspect_ratio_ids", "aspect_ratio_mask", "attention_mask", "auto_docstring", "base_model_prefix", "batch_size", "bias", "cat", "check_model_inputs", "class", "class_embedding", "config", "constant", "def", "device", "dim", "dtype", "else", "expand", "flatten", "for", "forward", "gated_positional_embedding", "get_input_embeddings", "global_output", "global_transformer", "height", "hidden_size", "hidden_state", "hidden_states", "i", "if", "image_size", "in", "in_channels", "intermediate_hidden_states", "intermediate_layers_indices", "is_gated", "kernel_size", "kwargs", "last_hidden_state", "layernorm_post", "layernorm_pre", "max_num_tiles", "mode", "nn", "num_channels", "num_concurrent_media", "num_global_layers", "num_hidden_layers", "num_padding_patches", "num_patches", "num_tiles", "out_channels", "output", "pad", "padding", "patch_embedding", "patch_embeds", "patch_size", "pixel_values", "post_init", "post_tile_positional_embedding", "pre_tile_positional_embedding", "r", "randn", "reshape", "return", "scale", "self", "shape", "slice_index", "stack", "stride", "super", "target_device", "target_dtype", "target_length", "to", "torch", "transformer", "transpose", "valid", "value", "view", "vision_model", "weight", "width"], "mllama/modeling_mllama.py:MllamaTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelCrossAttentionDecoderLayer", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModelSelfAttentionDecoderLayer", "ModelTextConfig", "ModelTextModel", "ModelTextRMSNorm", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "_update_causal_mask", "and", "append", "arange", "attention_mask", "auto_docstring", "base_model_prefix", "bool", "cache_position", "can_return_tuple", "causal_mask", "check_model_inputs", "class", "config", "continue", "cross_attention_layers", "cross_attention_mask", "cross_attention_states", "decoder_layer", "def", "device", "else", "embed_tokens", "enumerate", "eps", "exactly", "for", "forward", "full_text_row_masked_out_mask", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "idx", "if", "in", "input_ids", "inputs_embeds", "is", "is_cross_attention_cache_empty", "is_cross_attention_layer", "kwargs", "language_model", "last_hidden_state", "layer_idx", "layers", "list", "model", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "tuple", "unsqueeze", "use_cache", "vocab_size"], "mllama/modeling_mllama.py:MllamaForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_from_config", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attention_mask", "cross_attention_states", "def", "else", "float", "forward", "full_text_row_masked_out_mask", "get_text_config", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "vocab_size", "weight"], "mllama/modeling_mllama.py:MllamaModel": ["BaseModelOutputWithPast", "Cache", "FlashAttentionKwargs", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionModel", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "_from_config", "_prepare_cross_attention_mask", "and", "aspect_ratio_ids", "aspect_ratio_mask", "attention_mask", "attentions", "auto_docstring", "be", "bias", "bool", "cache_position", "can_return_tuple", "cannot", "check_model_inputs", "class", "config", "cross_attention_mask", "cross_attention_states", "decoder", "def", "dtype", "else", "exactly", "forward", "full_text_row_masked_out_mask", "get_decoder", "get_input_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "language_model", "last_hidden_state", "max_num_tiles", "model", "multi_modal_projector", "must", "nn", "not", "num_patches", "num_vision_tokens", "of", "one", "or", "outputs", "pad_token_id", "past_key_values", "pixel_values", "position_ids", "post_init", "provided", "r", "raise", "reshape", "return", "self", "set_decoder", "set_input_embeddings", "shape", "simultaneously", "specify", "super", "text_config", "torch", "use_cache", "value", "vision_config", "vision_model", "vision_output_dim", "vision_outputs", "vocab_size"], "mllama/modeling_mllama.py:MllamaForConditionalGeneration": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "_update_model_kwargs_for_generation", "aspect_ratio_ids", "aspect_ratio_mask", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "cross_attention_mask", "cross_attention_mask_prev", "cross_attention_states", "decoder", "def", "dim", "else", "forward", "get", "get_decoder", "get_input_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "model_kwargs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "value", "vision_model", "vocab_size", "weight"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinModelOutputWithPooling": ["FloatTensor", "ModelOutput", "ModelSwinModelOutputWithPooling", "None", "Optional", "attentions", "class", "hidden_states", "hidden_states_spatial_dimensions", "int", "last_hidden_state", "pooler_output", "r", "torch", "tuple"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinBaseModelOutput": ["FloatTensor", "ModelOutput", "ModelSwinBaseModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "hidden_states_spatial_dimensions", "int", "last_hidden_state", "r", "torch", "tuple"], "maskformer/modeling_maskformer_swin.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "height", "input_feature", "num_channels", "permute", "return", "shape", "view", "width"], "maskformer/modeling_maskformer_swin.py:window_reverse": ["Model_reverse", "Model_size", "Models", "contiguous", "def", "height", "num_channels", "permute", "return", "shape", "view", "width"], "maskformer/modeling_maskformer_swin.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinEmbeddings": ["Dropout", "False", "LayerNorm", "ModelSwinEmbeddings", "ModelSwinPatchEmbeddings", "Module", "None", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "bicubic", "cat", "class", "class_pos_embed", "config", "def", "dim", "dropout", "else", "embed_dim", "embeddings", "forward", "functional", "grid_size", "height", "hidden_dropout_prob", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mode", "new_height", "new_width", "nn", "norm", "not", "num_channels", "num_patches", "num_positions", "output_dimensions", "patch_embeddings", "patch_grid", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "use_absolute_embeddings", "view", "width", "zeros"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinPatchEmbeddings": ["Conv2d", "FloatTensor", "Iterable", "ModelSwinPatchEmbeddings", "Module", "Optional", "Tensor", "_", "__init__", "abc", "class", "collections", "config", "def", "else", "embed_dim", "embeddings", "flatten", "forward", "functional", "grid_size", "height", "hidden_size", "if", "image_size", "int", "isinstance", "kernel_size", "maybe_pad", "nn", "num_channels", "num_patches", "output_dimensions", "pad", "pad_values", "patch_size", "pixel_values", "projection", "return", "self", "shape", "stride", "super", "torch", "transpose", "tuple", "width"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinPatchMerging": ["False", "LayerNorm", "Linear", "ModelSwinPatchMerging", "Module", "None", "Tensor", "__init__", "batch_size", "bias", "cat", "class", "def", "dim", "forward", "functional", "height", "if", "input_dimensions", "input_feature", "input_feature_0", "input_feature_1", "input_feature_2", "input_feature_3", "input_resolution", "int", "maybe_pad", "nn", "norm", "norm_layer", "num_channels", "or", "pad", "pad_values", "reduction", "return", "self", "shape", "should_pad", "super", "torch", "tuple", "view", "width"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinDropPath": ["ModelSwinDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinSelfAttention": ["Dropout", "False", "FloatTensor", "Iterable", "Linear", "ModelSwinSelfAttention", "Model_shape", "Module", "None", "Optional", "Parameter", "Tensor", "The", "ValueError", "__init__", "a", "abc", "all_head_size", "arange", "attention", "attention_Model", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "collections", "config", "context_layer", "contiguous", "coords", "coords_flatten", "coords_h", "coords_w", "def", "dim", "dropout", "else", "f", "flatten", "forward", "functional", "head_Model", "heads", "hidden", "hidden_shape", "hidden_states", "if", "ij", "indexing", "int", "is", "isinstance", "key", "key_layer", "math", "matmul", "meshgrid", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_channels", "num_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "register_buffer", "relative_coords", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "return", "self", "shape", "size", "softmax", "sqrt", "stack", "sum", "super", "the", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view", "window_size", "zeros"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinSelfOutput": ["Dropout", "Linear", "ModelSwinSelfOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinAttention": ["False", "FloatTensor", "ModelSwinAttention", "ModelSwinSelfAttention", "ModelSwinSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_Model", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_Model", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinIntermediate": ["ACT2FN", "Linear", "ModelSwinIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinOutput": ["Dropout", "Linear", "ModelSwinOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinLayer": ["False", "Identity", "LayerNorm", "ModelSwinAttention", "ModelSwinDropPath", "ModelSwinIntermediate", "ModelSwinLayer", "ModelSwinOutput", "Model_windows", "Modeled_fill", "Module", "None", "_", "__init__", "attention", "attention_output", "attention_windows", "attn_Model", "batch_size", "channels", "class", "config", "contiguous", "count", "def", "device", "dim", "dims", "drop_path", "drop_path_rate", "else", "eps", "for", "forward", "functional", "get_attn_Model", "head_Model", "height", "height_pad", "height_slice", "height_slices", "hidden_states", "hidden_states_windows", "if", "img_Model", "in", "input_dimensions", "input_resolution", "intermediate", "is", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "maybe_pad", "nn", "not", "num_heads", "or", "output", "output_attentions", "outputs", "pad", "pad_bottom", "pad_left", "pad_right", "pad_top", "pad_values", "return", "roll", "self", "self_attention_outputs", "shape", "shift_size", "shifted_hidden_states", "shifted_windows", "shifts", "shortcut", "size", "slice", "super", "to", "torch", "unsqueeze", "view", "was_padded", "width", "width_pad", "width_slice", "width_slices", "window_partition", "window_reverse", "window_size", "zeros"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinStage": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelSwinLayer", "ModelSwinStage", "ModuleList", "None", "__init__", "all_hidden_states", "block_hidden_states", "block_module", "blocks", "class", "config", "def", "depth", "dim", "downsample", "drop_path", "drop_path_rate", "else", "enumerate", "for", "forward", "head_Model", "height", "height_downsampled", "hidden_states", "i", "if", "in", "input_dimensions", "input_resolution", "is", "layer_head_Model", "nn", "norm_layer", "not", "num_heads", "output_attentions", "output_dimensions", "output_hidden_states", "pointing", "range", "return", "self", "shift_size", "super", "width", "width_downsampled", "window_size"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinEncoder": ["False", "ModelSwinBaseModelOutput", "ModelSwinEncoder", "ModelSwinPatchMerging", "ModelSwinStage", "Module", "ModuleList", "None", "True", "__init__", "all_hidden_states", "all_input_dimensions", "all_self_attentions", "attentions", "class", "config", "cpu", "def", "depth", "depths", "device", "dim", "downsample", "dpr", "drop_path", "drop_path_rate", "else", "embed_dim", "enumerate", "for", "forward", "gradient_checkpointing", "grid_size", "head_Model", "hidden_states", "hidden_states_spatial_dimensions", "i", "i_layer", "if", "in", "input_dimensions", "input_resolution", "int", "is", "item", "last_hidden_state", "layer_all_hidden_states", "layer_head_Model", "layer_hidden_states", "layer_module", "layers", "len", "linspace", "nn", "not", "num_heads", "num_layers", "output_attentions", "output_dimensions", "output_hidden_states", "range", "return", "return_dict", "self", "sum", "super", "torch", "tuple", "v", "x"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "ModelSwinConfig", "ModelSwinEmbeddings", "ModelSwinPreTrainedModel", "ModelSwinSelfAttention", "ModelSwinStage", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "relative_position_bias_table", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinModel": ["AdaptiveAvgPool1d", "False", "LayerNorm", "ModelSwinEmbeddings", "ModelSwinEncoder", "ModelSwinModel", "ModelSwinModelOutputWithPooling", "ModelSwinPreTrainedModel", "None", "True", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flatten", "for", "forward", "get_head_Model", "get_input_embeddings", "have", "head_Model", "heads", "heads_to_prune", "hidden_states", "hidden_states_spatial_dimensions", "if", "in", "input_dimensions", "int", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "len", "nn", "not", "num_features", "num_layers", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_grid", "pixel_values", "pooled_output", "pooler", "pooler_output", "prune_heads", "raise", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "use_return_dict"], "maskformer/modeling_maskformer_swin.py:MaskFormerSwinBackbone": ["BackboneMixin", "BackboneOutput", "LayerNorm", "ModelSwinBackbone", "ModelSwinConfig", "ModelSwinModel", "ModelSwinPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "This", "True", "ValueError", "_", "__init__", "_init_backbone", "attentions", "backbone", "batch_size", "bool", "class", "config", "contiguous", "def", "depths", "does", "else", "embed_dim", "enumerate", "feature_maps", "for", "forward", "height", "hidden_size", "hidden_state", "hidden_state_norm", "hidden_state_permuted", "hidden_state_unpolled", "hidden_states", "hidden_states_norms", "hidden_states_spatial_dimensions", "i", "if", "in", "int", "is", "len", "model", "nn", "norm", "not", "num_channels", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "outputs", "permute", "pixel_values", "post_init", "raise", "range", "return", "return_dict", "self", "shape", "spatial_dimensions", "stage", "stage_names", "stem", "super", "support", "the", "tuple", "use_return_dict", "view", "width", "zip"], "maskformer/modeling_maskformer.py:DetrDecoderOutput": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "ModelDecoderOutput", "None", "Optional", "class", "intermediate_hidden_states", "r", "torch"], "maskformer/modeling_maskformer.py:MaskFormerPixelLevelModuleOutput": ["FloatTensor", "ModelOutput", "ModelPixelLevelModuleOutput", "None", "Optional", "class", "decoder_hidden_states", "decoder_last_hidden_state", "encoder_hidden_states", "encoder_last_hidden_state", "r", "torch", "tuple"], "maskformer/modeling_maskformer.py:MaskFormerPixelDecoderOutput": ["FloatTensor", "ModelOutput", "ModelPixelDecoderOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "maskformer/modeling_maskformer.py:MaskFormerModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "encoder_hidden_states", "encoder_last_hidden_state", "hidden_states", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "r", "torch", "transformer_decoder_hidden_states", "transformer_decoder_last_hidden_state", "tuple"], "maskformer/modeling_maskformer.py:MaskFormerForInstanceSegmentationOutput": ["FloatTensor", "ModelForInstanceSegmentationOutput", "ModelOutput", "Models_queries_logits", "None", "Optional", "attentions", "auxiliary_logits", "class", "class_queries_logits", "encoder_hidden_states", "encoder_last_hidden_state", "hidden_states", "loss", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "r", "torch", "transformer_decoder_hidden_states", "transformer_decoder_last_hidden_state", "tuple"], "maskformer/modeling_maskformer.py:upsample_like": ["False", "Model_like", "Modeld", "Tensor", "_", "align_corners", "bilinear", "def", "functional", "height", "interpolate", "like", "mode", "nn", "pixel_values", "return", "shape", "size", "str", "width"], "maskformer/modeling_maskformer.py:dice_loss": ["Model_loss", "Tensor", "def", "denominator", "flatten", "inputs", "int", "labels", "loss", "num_masks", "numerator", "probs", "r", "return", "sigmoid", "sum"], "maskformer/modeling_maskformer.py:sigmoid_focal_loss": ["BCEWithLogitsLoss", "Model", "Model_focal_loss", "Tensor", "alpha", "alpha_t", "criterion", "cross_entropy_loss", "def", "float", "gamma", "if", "inputs", "int", "labels", "loss", "mean", "nn", "none", "num_masks", "p_t", "probs", "r", "reduction", "return", "sum"], "maskformer/modeling_maskformer.py:pair_wise_dice_loss": ["Model_wise_dice_loss", "None", "T", "Tensor", "def", "denominator", "flatten", "inputs", "labels", "loss", "matmul", "numerator", "return", "sigmoid", "sum", "torch"], "maskformer/modeling_maskformer.py:pair_wise_sigmoid_focal_loss": ["BCEWithLogitsLoss", "Model_wise_sigmoid_focal_loss", "T", "Tensor", "ValueError", "alpha", "be", "criterion", "cross_entropy_loss_neg", "cross_entropy_loss_pos", "def", "float", "focal_neg", "focal_pos", "gamma", "height_and_width", "if", "inputs", "labels", "loss", "matmul", "must", "nn", "none", "ones_like", "positive", "prob", "r", "raise", "reduction", "return", "shape", "sigmoid", "torch", "zeros_like"], "maskformer/modeling_maskformer.py:DetrAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "is_cross_attention", "k_proj", "key_states", "key_value_states", "key_value_states_original", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "object_queries", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "spatial_position_embeddings", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "maskformer/modeling_maskformer.py:DetrDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "key_value_states", "nn", "not", "num_heads", "object_queries", "output_attentions", "outputs", "p", "query_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_position_embeddings", "super", "torch", "training"], "maskformer/modeling_maskformer.py:DetrDecoder": ["False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "Module", "ModuleList", "None", "_", "__init__", "_prepare_4d_attention_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attention_mask", "attentions", "auxiliary_loss", "class", "config", "continue", "cross_attentions", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "input_shape", "inputs_embeds", "intermediate", "intermediate_hidden_states", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layernorm", "layers", "nn", "not", "object_queries", "output_attentions", "output_hidden_states", "query_position_embeddings", "r", "rand", "range", "return", "return_dict", "self", "size", "stack", "super", "tgt_len", "torch", "training", "tuple", "use_return_dict", "v"], "maskformer/modeling_maskformer.py:MaskFormerHungarianMatcher": ["All", "Matcher", "ModelHungarianMatcher", "Model_labels", "Models_queries_logits", "Module", "None", "Tensor", "ValueError", "__class__", "__init__", "__name__", "__repr__", "_repr_indent", "and", "append", "array", "as_tensor", "assigned_indices", "be", "body", "can", "class", "class_labels", "class_queries_logits", "cost_Model", "cost_class", "cost_dice", "cost_matrix", "costs", "cpu", "def", "dtype", "f", "flatten", "float", "for", "forward", "functional", "head", "i", "if", "in", "indices", "int64", "interpolate", "j", "join", "labels", "line", "linear_sum_assignment", "lines", "list", "matched_indices", "mode", "n", "nearest", "nn", "no_grad", "np", "pair_wise_dice_loss", "pair_wise_sigmoid_focal_loss", "pred_Model", "pred_Model_flat", "pred_probs", "preds_Models", "preds_probs", "raise", "return", "self", "shape", "size", "softmax", "super", "t", "target_Model", "target_Model_flat", "torch", "tuple", "zip"], "maskformer/modeling_maskformer.py:MaskFormerLoss": ["CrossEntropyLoss", "False", "ModelHungarianMatcher", "ModelLoss", "Model_labels", "Models_queries_logits", "Module", "None", "Number", "Optional", "PartialState", "Tensor", "_", "__init__", "_get_predictions_permutation_indices", "_get_targets_permutation_indices", "_max_by_axis", "_pad_images_to_max_in_batch", "_shared_state", "align_corners", "array", "as_tensor", "aux_outputs", "auxiliary_predictions", "b", "batch_indices", "batch_shape", "batch_size", "bilinear", "bool", "cat", "clamp", "class", "class_labels", "class_queries_logits", "classes", "copy_", "criterion", "def", "device", "dice_loss", "dict", "dtype", "empty_weight", "enumerate", "eos_coef", "f", "fill_value", "flatten", "float", "for", "forward", "full", "full_like", "functional", "get_num_Models", "h", "i", "idx", "if", "in", "index", "indices", "int", "int64", "interpolate", "is", "is_accelerate_available", "item", "items", "j", "key", "len", "list", "loss_Model", "loss_Models", "loss_ce", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_labels", "losses", "matcher", "max", "max_size", "maxes", "min", "mode", "nn", "not", "np", "num_Models", "num_labels", "num_processes", "num_queries", "ones", "padded_tensor", "padded_tensors", "padding_Model", "padding_Models", "pred_Models", "pred_logits", "pred_logits_transposed", "predictions_indices", "reduce", "register_buffer", "requires_backends", "return", "scipy", "self", "shape", "sigmoid_focal_loss", "size", "src", "src_idx", "str", "sublist", "sum", "super", "target", "target_Models", "target_classes", "target_classes_o", "target_indices", "tensor", "tensors", "tgt", "tgt_idx", "the_list", "torch", "transpose", "tuple", "update", "value", "w", "weight", "weight_dict", "world_size", "zeros", "zip"], "maskformer/modeling_maskformer.py:MaskFormerFPNConvLayer": ["Conv2d", "False", "GroupNorm", "ModelFPNConvLayer", "Module", "ReLU", "Tensor", "True", "__init__", "add_module", "bias", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_features", "inplace", "input", "int", "kernel_size", "layer", "layers", "nn", "out_features", "padding", "return", "self", "str", "super"], "maskformer/modeling_maskformer.py:MaskFormerFPNLayer": ["Conv2d", "False", "GroupNorm", "ModelFPNConvLayer", "ModelFPNLayer", "Module", "Sequential", "Tensor", "__init__", "bias", "block", "class", "def", "down", "forward", "functional", "in_features", "int", "interpolate", "kernel_size", "lateral_features", "left", "mode", "nearest", "nn", "padding", "proj", "return", "self", "shape", "size", "super"], "maskformer/modeling_maskformer.py:MaskFormerFPNModel": ["ModelFPNConvLayer", "ModelFPNLayer", "ModelFPNModel", "Module", "Sequential", "Tensor", "__init__", "append", "class", "def", "feature_size", "features", "for", "forward", "fpn_features", "in", "in_features", "int", "last_feature", "lateral_width", "lateral_widths", "layer", "layers", "left", "list", "nn", "other_features", "output", "return", "self", "stem", "super", "zip"], "maskformer/modeling_maskformer.py:MaskFormerPixelDecoder": ["Conv2d", "False", "ModelFPNModel", "ModelPixelDecoder", "ModelPixelDecoderOutput", "Model_feature_size", "Model_projection", "Module", "Tensor", "True", "__init__", "args", "bool", "class", "def", "else", "feature_size", "features", "forward", "fpn", "fpn_features", "hidden_states", "if", "int", "kernel_size", "kwargs", "last_feature_projected", "last_hidden_state", "list", "nn", "not", "output_hidden_states", "padding", "r", "return", "return_dict", "self", "super", "tuple"], "maskformer/modeling_maskformer.py:MaskFormerSinePositionEmbedding": ["False", "Model", "ModelSinePositionEmbedding", "Module", "None", "Optional", "Size", "Tensor", "True", "Union", "ValueError", "__init__", "and", "arange", "be", "bool", "cat", "class", "compile_compatible_method_lru_cache", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "else", "eps", "flatten", "float", "floor", "forward", "if", "int", "int64", "is", "math", "maxsize", "nn", "normalize", "not", "not_Model", "num_pos_feats", "passed", "permute", "pi", "pos", "pos_x", "pos_y", "raise", "return", "rounding_mode", "scale", "self", "shape", "should", "sin", "stack", "str", "super", "temperature", "to", "torch", "x_embed", "y_embed", "zeros"], "maskformer/modeling_maskformer.py:PredictionBlock": ["Linear", "ModelBlock", "Module", "None", "Tensor", "__init__", "activation", "add_module", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_dim", "input", "int", "layer", "layers", "nn", "out_dim", "return", "self", "str", "super"], "maskformer/modeling_maskformer.py:MaskformerMLPPredictionHead": ["Identity", "ModelMLPPredictionHead", "Module", "PredictionBlock", "ReLU", "Tensor", "__init__", "activation", "add_module", "append", "class", "def", "else", "enumerate", "for", "forward", "hidden_dim", "hidden_state", "i", "if", "in", "in_dim", "in_dims", "input", "input_dim", "int", "layer", "layers", "nn", "num_layers", "out_dim", "out_dims", "output_dim", "return", "self", "str", "super", "zip"], "maskformer/modeling_maskformer.py:MaskFormerPixelLevelModule": ["False", "ModelConfig", "ModelPixelDecoder", "ModelPixelLevelModule", "ModelPixelLevelModuleOutput", "ModelSwinConfig", "Model_feature_size", "Module", "None", "Tensor", "True", "__init__", "and", "backbone_config", "bool", "channels", "class", "config", "decoder", "decoder_hidden_states", "decoder_last_hidden_state", "decoder_output", "def", "else", "encoder", "encoder_hidden_states", "encoder_last_hidden_state", "feature_channels", "feature_maps", "feature_size", "features", "forward", "fpn_feature_size", "from_dict", "getattr", "hidden_states", "if", "in_features", "is", "last_hidden_state", "lateral_widths", "load_backbone", "model_type", "nn", "not", "out_features", "output_hidden_states", "outputs", "pixel_values", "return", "return_dict", "self", "stage1", "stage2", "stage3", "stage4", "super", "swin", "to_dict", "tuple"], "maskformer/modeling_maskformer.py:MaskFormerTransformerModule": ["Conv2d", "DetrDecoder", "DetrDecoderOutput", "Embedding", "False", "ModelConfig", "ModelSinePositionEmbedding", "ModelTransformerModule", "Module", "None", "Optional", "Tensor", "True", "__init__", "attention_Model", "batch_size", "bool", "class", "config", "decoder", "decoder_config", "decoder_output", "def", "device", "dtype", "else", "encoder_attention_Model", "encoder_hidden_states", "forward", "height", "hidden_size", "if", "image_features", "in_features", "input_projection", "inputs_embeds", "int", "is", "kernel_size", "nn", "normalize", "not", "num_channels", "num_pos_feats", "num_queries", "object_queries", "output_attentions", "output_hidden_states", "permute", "position_embedder", "queries_embedder", "queries_embeddings", "query_position_embeddings", "repeat", "requires_grad", "requires_grad_", "return", "return_dict", "self", "shape", "should_project", "super", "torch", "training", "unsqueeze", "view", "weight", "width", "zeros_like"], "maskformer/modeling_maskformer.py:MaskFormerPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelFPNConvLayer", "ModelFPNLayer", "ModelFPNModel", "ModelMLPPredictionHead", "ModelPreTrainedModel", "ModelTransformerModule", "Module", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "constant_", "data", "def", "elif", "fill_", "for", "gain", "get_submodule", "if", "in", "init", "init_std", "init_xavier_std", "input_projection", "is", "isinstance", "main_input_name", "mean", "model", "module", "modules", "nn", "normal_", "not", "padding_idx", "pixel_values", "proj", "self", "std", "stem", "submodule", "weight", "xavier_std", "xavier_uniform_", "zero_"], "maskformer/modeling_maskformer.py:MaskFormerModel": ["ModelConfig", "ModelModel", "ModelModelOutput", "ModelPixelLevelModule", "ModelPreTrainedModel", "ModelTransformerModule", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "attentions", "auto_docstring", "batch_size", "bool", "channels", "class", "config", "def", "device", "else", "encoder", "encoder_hidden_states", "encoder_last_hidden_state", "for", "forward", "have", "height", "hidden_states", "if", "image_features", "in", "in_features", "is", "last_hidden_state", "not", "ones", "output", "output_attentions", "output_hidden_states", "pixel_Model", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "pixel_embeddings", "pixel_level_module", "pixel_level_module_output", "pixel_values", "post_init", "queries", "r", "raise", "return", "return_dict", "self", "shape", "specify", "super", "to", "torch", "transformer_decoder_hidden_states", "transformer_decoder_last_hidden_state", "transformer_module", "transformer_module_output", "tuple", "use_return_dict", "v", "values", "width"], "maskformer/modeling_maskformer.py:MaskFormerForInstanceSegmentation": ["Linear", "ModelConfig", "ModelForInstanceSegmentation", "ModelForInstanceSegmentationOutput", "ModelHungarianMatcher", "ModelLoss", "ModelMLPPredictionHead", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "Model_embedder", "Model_embeddings", "Model_feature_size", "Model_labels", "Model_weight", "Models_queries_logits", "None", "Optional", "Tensor", "__init__", "and", "append", "attentions", "auto_docstring", "aux_binary_Models", "aux_classes", "auxiliary_logits", "bchw", "binaries_Models", "bool", "bqc", "bqhw", "class", "class_labels", "class_predictor", "class_queries_logits", "classes", "config", "cost_Model", "cost_class", "cost_dice", "criterion", "cross_entropy_weight", "decoder_config", "def", "dice_weight", "dict", "einsum", "else", "encoder_hidden_states", "encoder_last_hidden_state", "eos_coef", "float", "for", "forward", "get_logits", "get_loss", "get_loss_dict", "hidden_size", "hidden_states", "if", "in", "is", "items", "key", "lbqc", "lbqhw", "list", "loss", "loss_Model", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_key", "matcher", "model", "nn", "no_object_weight", "not", "num_labels", "or", "output", "output_attentions", "output_auxiliary_logits", "output_hidden_states", "outputs", "pixel_Model", "pixel_decoder_hidden_states", "pixel_decoder_last_hidden_state", "pixel_embeddings", "pixel_values", "post_init", "r", "raw_outputs", "return", "return_dict", "self", "stack", "stacked_transformer_decoder_outputs", "str", "sum", "super", "torch", "transformer_decoder_hidden_states", "transformer_decoder_last_hidden_state", "tuple", "use_auxiliary_loss", "use_return_dict", "v", "values", "weight", "weight_dict", "zip"], "blenderbot_small/modeling_blenderbot_small.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallLearnedPositionalEmbedding": ["Embedding", "ModelLearnedPositionalEmbedding", "None", "Optional", "Size", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "embedding_dim", "forward", "if", "input_ids_shape", "int", "is", "long", "nn", "num_embeddings", "past_key_values_length", "position_ids", "return", "self", "seq_len", "super", "torch", "weight"], "blenderbot_small/modeling_blenderbot_small.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "None", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "int", "isinf", "isnan", "layer_head_mask", "layer_idx", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "decoder_input_ids", "def", "device", "diagonal", "dim", "dtype", "dummy_inputs", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "ne", "nn", "normal_", "not", "npu", "ones", "pad_token", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "property", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tensor", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallEncoder": ["BaseModelOutput", "Embedding", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "The", "True", "ValueError", "You", "_", "__init__", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "both", "but", "cannot", "class", "config", "d_model", "def", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "The", "Transformers", "True", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "both", "but", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallModel": ["BaseModelOutput", "Cache", "Embedding", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "nn", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallForConditionalGeneration": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "provided", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "blenderbot_small/modeling_blenderbot_small.py:BlenderbotSmallForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "got_ocr2/modeling_got_ocr2.py:GotOcr2MLPBlock": ["ACT2FN", "Linear", "ModelMLPBlock", "Module", "Tensor", "__init__", "act", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_states", "lin1", "lin2", "mlp_dim", "nn", "return", "self", "super", "torch"], "got_ocr2/modeling_got_ocr2.py:GotOcr2VisionAttention": ["F", "Input", "Linear", "ModelVisionAttention", "Module", "None", "Parameter", "Tensor", "ValueError", "_", "__init__", "arange", "attention_dropout", "attn_output", "attn_probs", "attn_weights", "batch_size", "be", "bhwc", "bhwk", "bias", "class", "config", "decomposed_rel_pos", "def", "dim", "dropout", "dtype", "einsum", "else", "encoding", "float32", "forward", "functional", "get_decomposed_rel_pos", "get_rel_pos", "head_dim", "height", "hidden_size", "hidden_states", "hkc", "if", "image_size", "input_size", "int", "interpolate", "is", "k_coords", "k_size", "key", "key_height", "key_width", "linear", "long", "max", "max_rel_dist", "mode", "must", "nn", "num_attention_heads", "output_attentions", "p", "patch_size", "permute", "positional", "proj", "provided", "q_coords", "q_size", "qkv", "qkv_bias", "query", "query_height", "query_width", "raise", "rel_h", "rel_pos", "rel_pos_h", "rel_pos_resized", "rel_pos_w", "rel_w", "relative", "relative_coords", "relative_position_height", "relative_position_width", "reshape", "reshape_as", "reshaped_query", "return", "scale", "self", "shape", "size", "softmax", "super", "to", "torch", "training", "transpose", "tuple", "unbind", "use_rel_pos", "using", "value", "width", "window_size", "wkc", "zeros"], "got_ocr2/modeling_got_ocr2.py:GotOcr2VisionLayer": ["F", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelMLPBlock", "ModelVisionAttention", "ModelVisionLayer", "Tensor", "__init__", "attn", "attn_weights", "batch_size", "channel", "class", "config", "contiguous", "def", "eps", "forward", "height", "hidden_size", "hidden_states", "if", "int", "layer_norm1", "layer_norm2", "layer_norm_eps", "layernorm_output", "mlp", "nn", "original_shape", "pad", "pad_h", "pad_height", "pad_w", "pad_width", "padding_shape", "permute", "reshape", "residual", "return", "self", "shape", "super", "torch", "tuple", "width", "window_partition", "window_size", "window_unpartition", "windows"], "got_ocr2/modeling_got_ocr2.py:GotOcr2PreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionEncoder", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "data", "def", "elif", "if", "is", "isinstance", "module", "not", "past_key_values", "pos_embed", "rel_pos_h", "rel_pos_w", "self", "super", "supports_gradient_checkpointing", "use_rel_pos", "zero_"], "got_ocr2/modeling_got_ocr2.py:GotOcr2VisionEncoderOutput": ["FloatTensor", "ModelOutput", "ModelVisionEncoderOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "got_ocr2/modeling_got_ocr2.py:GotOcr2PatchEmbeddings": ["Conv2d", "Input", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "forward", "height", "hidden_size", "if", "image", "image_size", "in", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "of", "one", "or", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "values", "width", "with"], "got_ocr2/modeling_got_ocr2.py:GotOcr2LayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "got_ocr2/modeling_got_ocr2.py:GotOcr2VisionNeck": ["Conv2d", "False", "ModelLayerNorm", "ModelVisionConfig", "ModelVisionNeck", "Module", "__init__", "bias", "channels_first", "class", "config", "conv1", "conv2", "data_format", "def", "forward", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "nn", "output_channels", "padding", "permute", "return", "self", "super"], "got_ocr2/modeling_got_ocr2.py:GotOcr2VisionEncoder": ["False", "FloatTensor", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderOutput", "ModelVisionLayer", "ModelVisionNeck", "ModuleList", "None", "Optional", "Parameter", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "append", "attentions", "check_model_inputs", "class", "config", "def", "else", "for", "forward", "get_input_embeddings", "global_attn_indexes", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "i", "if", "image_size", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_module", "layers", "neck", "nn", "not", "num_hidden_layers", "patch_embed", "patch_size", "pixel_values", "pos_embed", "raise", "range", "return", "self", "specify", "super", "to", "torch", "use_abs_pos", "window_size", "zeros"], "got_ocr2/modeling_got_ocr2.py:GotOcr2MultiModalProjector": ["Conv2d", "False", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "Tensor", "__init__", "bias", "class", "config", "conv_upsampler1", "conv_upsampler2", "def", "flatten", "forward", "hidden_size", "hidden_state", "kernel_size", "language_hidden_size", "multimodal_projector", "nn", "output_channels", "padding", "permute", "return", "self", "stride", "super", "text_config", "torch", "vision_config", "vision_embeddings", "vision_output_channels"], "got_ocr2/modeling_got_ocr2.py:GotOcr2CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "got_ocr2/modeling_got_ocr2.py:GotOcr2ModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "got_ocr2/modeling_got_ocr2.py:GotOcr2Model": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "ModelVisionEncoder", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "device", "do", "dtype", "else", "exactly", "expand_as", "f", "features", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "input_ids", "inputs_embeds", "is", "kwargs", "language_model", "last_hidden_state", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "raise", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_config", "vision_tower"], "got_ocr2/modeling_got_ocr2.py:GotOcr2ForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "str", "super", "text_config", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "vjepa2/modeling_vjepa2.py:VJEPA2WithMaskedInputPredictorOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "masked_hidden_state", "r", "target_hidden_state", "torch", "tuple"], "vjepa2/modeling_vjepa2.py:VJEPA2WithMaskedInputModelOutput": ["FloatTensor", "Model", "ModelOutput", "ModelWithMaskedInputPredictorOutput", "None", "Optional", "attentions", "class", "def", "hidden_states", "if", "isinstance", "last_hidden_state", "list", "masked_hidden_state", "output", "predictor_output", "r", "return", "self", "super", "to_tuple", "torch", "tuple"], "vjepa2/modeling_vjepa2.py:VJEPA2PatchEmbeddings3D": ["Conv3d", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "crop_size", "def", "flatten", "forward", "frames_per_clip", "hidden_size", "in_channels", "in_chans", "int", "kernel_size", "nn", "num_patches", "out_channels", "patch_size", "pixel_values_videos", "proj", "return", "self", "staticmethod", "stride", "super", "torch", "transpose", "tubelet_size", "x"], "vjepa2/modeling_vjepa2.py:VJEPA2Embeddings": ["Model", "ModelConfig", "ModelPatchEmbeddings3D", "Module", "Tensor", "__init__", "class", "config", "def", "dtype", "embeddings", "forward", "hidden_size", "if", "int", "nn", "num_frames", "num_patches", "patch_embeddings", "patch_size", "permute", "pixel_values_videos", "proj", "repeat", "return", "self", "shape", "super", "target_dtype", "to", "torch", "tubelet_size", "weight"], "vjepa2/modeling_vjepa2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "vjepa2/modeling_vjepa2.py:rotate_queries_or_keys": ["B", "D", "Model_queries_or_keys", "N", "arange", "cos", "def", "device", "dim", "dtype", "emb_cos", "emb_sin", "flatten", "freq", "num_heads", "omega", "pos", "repeat", "return", "sin", "size", "squeeze", "stack", "torch", "unbind", "unflatten", "unsqueeze", "x", "y", "y1", "y2"], "vjepa2/modeling_vjepa2.py:VJEPA2RopeAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "The", "Union", "ValueError", "_", "__init__", "_attn_implementation", "_get_frame_pos", "_get_height_pos", "a", "all_head_size", "apply_rotary_embeddings", "arange", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "bool", "cat", "class", "config", "context_layer", "crop_size", "d_dim", "d_mask", "def", "device", "dim", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "f", "forward", "frame_ids", "frames_per_clip", "get_position_ids", "grid_depth", "grid_size", "h_dim", "h_mask", "head_mask", "heads", "height_ids", "hidden", "hidden_size", "hidden_states", "ids", "if", "int", "is", "is_causal", "key", "key_layer", "masks", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "patch_size", "pos", "pos_ids", "position_mask", "proj", "qk", "qkd", "qkh", "qkr", "qkv_bias", "qkw", "query", "query_layer", "raise", "repeat", "reshape", "return", "rotate_queries_or_keys", "s", "scaling", "self", "seq_length", "shape", "size", "super", "the", "token_size", "tokens_per_frame", "tokens_per_row", "torch", "training", "transpose", "tubelet_size", "tuple", "unsqueeze", "value", "value_layer", "view", "w_dim", "w_mask", "width_ids", "x"], "vjepa2/modeling_vjepa2.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "vjepa2/modeling_vjepa2.py:VJEPA2DropPath": ["Model", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "vjepa2/modeling_vjepa2.py:VJEPA2MLP": ["ACT2FN", "Linear", "Model", "ModelConfig", "Module", "Tensor", "True", "__init__", "activation", "bias", "class", "config", "def", "fc1", "fc2", "float", "forward", "hidden_act", "hidden_features", "hidden_size", "hidden_state", "in_features", "int", "mlp_ratio", "nn", "out_features", "return", "self", "super", "torch"], "vjepa2/modeling_vjepa2.py:VJEPA2Layer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "Model", "ModelConfig", "ModelDropPath", "ModelMLP", "ModelRopeAttention", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "bool", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "float", "forward", "head_mask", "hidden_size", "hidden_states", "if", "int", "layer_norm_eps", "mlp", "mlp_ratio", "nn", "norm1", "norm2", "num_attention_heads", "output_attentions", "outputs", "position_mask", "residual", "return", "self", "self_attention_outputs", "super", "torch", "tuple"], "vjepa2/modeling_vjepa2.py:VJEPA2Encoder": ["BaseModelOutput", "False", "LayerNorm", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "bool", "can_return_tuple", "class", "config", "def", "drop_path_rate", "drop_path_rates", "else", "embeddings", "enumerate", "eps", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_norm_eps", "layer_outputs", "layernorm", "mlp_ratio", "nn", "not", "num_attention_heads", "num_hidden_layers", "output_attentions", "output_hidden_states", "pixel_values_videos", "range", "return", "self", "super", "torch"], "vjepa2/modeling_vjepa2.py:apply_masks": ["Model_masks", "Tensor", "all_masked_tensors", "cat", "def", "device", "dim", "for", "gather", "in", "index", "list", "mask", "mask_keep", "masks", "repeat", "return", "size", "tensor", "to", "torch", "unsqueeze"], "vjepa2/modeling_vjepa2.py:VJEPA2PredictorEmbeddings": ["B", "Linear", "Model", "ModelConfig", "Module", "Parameter", "Tensor", "__init__", "apply_masks", "cat", "class", "cm", "config", "context", "context_mask", "crop_size", "def", "dim", "else", "embeddings", "forward", "frames_per_clip", "hidden_size", "hidden_states", "if", "int", "len", "list", "mask_index", "mask_tokens", "masks", "max", "max_patch_num", "nn", "num_mask_tokens", "num_patches", "patch_size", "pred_hidden_size", "pred_num_mask_tokens", "pred_zero_init_mask_tokens", "predictor_embeddings", "repeat", "return", "self", "size", "staticmethod", "super", "target", "target_mask", "tm", "torch", "tubelet_size", "tuple", "zero_init_mask_tokens", "zeros"], "vjepa2/modeling_vjepa2.py:VJEPA2Predictor": ["BaseModelOutput", "D", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "Module", "ModuleList", "N_ctxt", "None", "Optional", "Tensor", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "and", "apply_masks", "argsort", "argsort_4d", "argsort_5d", "attentions", "bias", "bool", "can_return_tuple", "class", "config", "context_mask", "def", "device", "dim", "drop_path_rate", "drop_path_rates", "else", "embeddings", "encoder_hidden_states", "enumerate", "eps", "expand", "for", "forward", "gather", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "hidden_states_argsort", "i", "if", "in", "index", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_norm_eps", "layer_outputs", "layernorm", "list", "mlp_ratio", "nn", "not", "num_attention_heads", "output_attentions", "output_hidden_states", "permute", "position_masks", "pred_hidden_size", "pred_mlp_ratio", "pred_num_attention_heads", "pred_num_hidden_layers", "proj", "range", "return", "reverse_argsort", "self", "shape", "size", "sort_tokens", "super", "target_mask", "to", "torch", "unsort_tokens", "unsqueeze"], "vjepa2/modeling_vjepa2.py:VJEPA2PoolerSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "vjepa2/modeling_vjepa2.py:VJEPA2PoolerCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "if", "is_causal", "k_proj", "keys", "kv_seq_length", "must", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "q_proj", "q_seq_length", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "vjepa2/modeling_vjepa2.py:VJEPA2PoolerSelfAttentionLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelConfig", "ModelMLP", "ModelPoolerSelfAttention", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "vjepa2/modeling_vjepa2.py:VJEPA2PoolerCrossAttentionLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelConfig", "ModelMLP", "ModelPoolerCrossAttention", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "cross_attn", "def", "eps", "forward", "hidden_size", "hidden_state", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "queries", "residual", "return", "self", "super", "torch", "tuple"], "vjepa2/modeling_vjepa2.py:VJEPA2AttentivePooler": ["Model", "ModelConfig", "ModelPoolerCrossAttentionLayer", "ModelPoolerSelfAttentionLayer", "Module", "ModuleList", "None", "Parameter", "Tensor", "_", "__init__", "attention_mask", "class", "config", "cross_attention_layer", "def", "for", "forward", "hidden_size", "hidden_state", "in", "layer", "nn", "num_pooler_layers", "queries", "query_tokens", "range", "repeat", "return", "self", "self_attention_layers", "shape", "squeeze", "super", "torch", "zeros"], "vjepa2/modeling_vjepa2.py:VJEPA2PreTrainedModel": ["Conv2d", "Conv3d", "LayerNorm", "Linear", "Model", "ModelAttentivePooler", "ModelConfig", "ModelLayer", "ModelPoolerCrossAttentionLayer", "ModelPoolerSelfAttentionLayer", "ModelPredictorEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "cross_attention_layer", "data", "data_float_32", "data_init", "def", "dtype", "elif", "else", "enumerate", "fc2", "fill_", "float32", "for", "i", "if", "in", "init", "init_std", "initializer_range", "is", "isinstance", "layer", "len", "main_input_name", "mask_tokens", "mean", "mlp", "module", "nn", "not", "out_proj", "pixel_values_videos", "query_tokens", "self", "self_attention_layers", "self_attn", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "trunc_normal_f32_", "weight", "zero_", "zero_init_mask_tokens"], "vjepa2/modeling_vjepa2.py:_convert_head_mask_to_5d": ["None", "_convert_head_mask_to_5d", "def", "else", "expand", "head_mask", "if", "is", "not", "num_hidden_layers", "return", "unsqueeze"], "vjepa2/modeling_vjepa2.py:VJEPA2Model": ["B", "BaseModelOutput", "False", "Model", "ModelConfig", "ModelEncoder", "ModelPatchEmbeddings3D", "ModelPreTrainedModel", "ModelPredictor", "ModelWithMaskedInputModelOutput", "ModelWithMaskedInputPredictorOutput", "N", "None", "Optional", "Tensor", "True", "ValueError", "You", "__init__", "_convert_head_mask_to_5d", "and", "apply_masks", "arange", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "context_head_mask", "context_mask", "def", "device", "else", "embeddings", "encoder", "encoder_hidden_states", "encoder_output", "encoder_outputs", "forward", "get_input_embeddings", "get_vision_features", "have", "head_mask", "hidden_states", "if", "is", "kwargs", "last_hidden_state", "list", "masked_hidden_state", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_embeddings", "pixel_values_videos", "post_init", "pred_num_hidden_layers", "predictor", "predictor_output", "predictor_outputs", "r", "raise", "repeat", "return", "self", "sequence_output", "size", "skip_predictor", "specify", "super", "target_head_mask", "target_hidden_state", "target_mask", "to", "torch", "unsqueeze"], "vjepa2/modeling_vjepa2.py:VJEPA2ForVideoClassification": ["ImageClassifierOutput", "Linear", "Model", "ModelAttentivePooler", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "classifier", "config", "def", "forward", "hidden_size", "hidden_states", "if", "is", "labels", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pixel_values_videos", "pooled_logits", "pooler", "pooler_output", "post_init", "r", "return", "self", "skip_predictor", "super", "torch", "tuple"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1RMSNorm": ["ModelYuanMoEV1RMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1MLP": ["ACT2FN", "False", "Linear", "ModelYuanMoEV1Config", "ModelYuanMoEV1MLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "is_shared_mlp", "layer_idx", "nn", "return", "self", "super", "up_proj", "x"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelYuanMoEV1Attention", "ModelYuanMoEV1Config", "ModelYuanMoEV1RMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_layernorm", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_layernorm", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1Gate": ["False", "Linear", "ModelYuanMoEV1Config", "ModelYuanMoEV1Gate", "Module", "None", "Optional", "__init__", "bias", "bsz", "class", "config", "def", "dtype", "else", "float", "float32", "forward", "hidden_size", "hidden_states", "if", "int", "isinstance", "layer_idx", "logits", "nn", "num_experts", "reshape", "return", "self", "seq_len", "shape", "super", "torch", "weight", "wg"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1Moe": ["F", "False", "ModelYuanMoEV1Config", "ModelYuanMoEV1Gate", "ModelYuanMoEV1MLP", "ModelYuanMoEV1Moe", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "_", "__init__", "batch_size", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "else", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_states", "hidden_states_mlp", "idx", "if", "in", "index_add_", "int", "is_shared_mlp", "isinstance", "keepdim", "layer_idx", "moe_topk", "nn", "nonzero", "num_classes", "num_experts", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "shared_mlp", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelYuanMoEV1Attention", "ModelYuanMoEV1Config", "ModelYuanMoEV1DecoderLayer", "ModelYuanMoEV1Moe", "ModelYuanMoEV1RMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1PreTrainedModel": ["Embedding", "False", "Linear", "ModelYuanMoEV1Attention", "ModelYuanMoEV1Config", "ModelYuanMoEV1DecoderLayer", "ModelYuanMoEV1PreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1RotaryEmbedding": ["False", "ModelYuanMoEV1Config", "ModelYuanMoEV1RotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "alpha", "and", "arange", "attention_scaling", "autocast", "base", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "head_dim", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_theta", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelYuanMoEV1Config", "ModelYuanMoEV1DecoderLayer", "ModelYuanMoEV1Model", "ModelYuanMoEV1PreTrainedModel", "ModelYuanMoEV1RMSNorm", "ModelYuanMoEV1RotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelYuanMoEV1ForCausalLM", "ModelYuanMoEV1Model", "ModelYuanMoEV1PreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "hunyuan_v1_moe/modeling_hunyuan_v1_moe.py:HunYuanMoEV1ForSequenceClassification": ["GenericForSequenceClassification", "ModelYuanMoEV1ForSequenceClassification", "ModelYuanMoEV1PreTrainedModel", "class", "pass"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextRMSNorm": ["ModelTextRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextRouter": ["False", "Linear", "ModelTextRouter", "True", "__init__", "bias", "class", "config", "def", "dim", "dtype", "float", "forward", "functional", "hidden_size", "hidden_states", "keepdim", "nn", "num_experts", "num_experts_per_tok", "reshape", "return", "router_indices", "router_logits", "router_weights", "routing_weights", "scatter_", "self", "softmax", "sum", "super", "to", "top_k", "topk", "torch", "zeros_like"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextExperts": ["ACT2FN", "ModelTextExperts", "Module", "None", "Parameter", "Tensor", "_", "__init__", "act_fn", "batch_size", "bmm", "chunk", "class", "config", "current_state", "def", "device", "dim", "down_proj", "dtype", "else", "empty", "expert_dim", "expert_hit", "expert_idx", "expert_mask", "for", "forward", "functional", "gate", "gate_up", "gate_up_proj", "gated_output", "greater", "hidden_act", "hidden_size", "hidden_states", "if", "in", "index_add_", "intermediate_size", "moe_intermediate_size", "next_states", "nn", "no_grad", "nonzero", "num_classes", "num_experts", "one_hot", "out", "permute", "repeat", "reshape", "return", "router_indices", "routing_weights", "self", "shape", "sum", "super", "to", "token_idx", "torch", "training", "transpose", "up", "view", "weighted_output", "where", "with", "zeros_like"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextSparseMoeBlock": ["ModelTextExperts", "ModelTextRouter", "ModelTextSparseMoeBlock", "Module", "Tensor", "__init__", "class", "config", "def", "experts", "forward", "gate", "hidden_size", "hidden_states", "nn", "num_experts", "return", "routed_out", "router_indices", "router_logits", "router_weights", "self", "super", "torch"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextMLP": ["ACT2FN", "False", "Linear", "ModelTextMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "not", "return", "self", "super", "up_proj", "x"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextDecoderLayer": ["Cache", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextMLP", "ModelTextRMSNorm", "ModelTextSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "and", "attention_mask", "cache_position", "class", "config", "decoder_sparse_step", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "int", "intermediate_size", "isinstance", "kwargs", "layer_idx", "mlp", "mlp_only_layers", "new_name", "not", "num_experts", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoePreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextDecoderLayer", "ModelTextExperts", "ModelTextSparseMoeBlock", "ModelVisionBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "down_proj", "else", "gate_up_proj", "get_text_config", "getattr", "hasattr", "hidden_states", "if", "index", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "router_logits", "self", "std", "super", "supports_gradient_checkpointing"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "True", "__init__", "act_fn", "bias", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "linear_fc1", "linear_fc2", "nn", "return", "self", "super"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionPatchEmbed": ["Conv3d", "ModelVisionPatchEmbed", "Module", "None", "Tensor", "True", "__init__", "bias", "class", "config", "def", "dtype", "embed_dim", "forward", "hidden_size", "hidden_states", "in_channels", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionRotaryEmbedding": ["False", "ModelVisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionPatchMerger": ["False", "GELU", "LayerNorm", "Linear", "ModelVisionConfig", "ModelVisionPatchMerger", "Module", "None", "Tensor", "__init__", "act_fn", "class", "config", "def", "else", "eps", "forward", "hidden_size", "if", "linear_fc1", "linear_fc2", "nn", "norm", "out_hidden_size", "return", "self", "spatial_merge_size", "super", "torch", "use_postshuffle_norm", "view", "x"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionBlock": ["GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionBlock", "ModelVisionMLP", "None", "Optional", "Tensor", "__init__", "attn", "attn_implementation", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "nn", "norm1", "norm2", "position_embeddings", "return", "rotary_pos_emb", "sdpa", "self", "str", "super", "torch", "tuple"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeVisionModel": ["Embedding", "F", "False", "ModelPreTrainedModel", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionModel", "ModelVisionPatchEmbed", "ModelVisionPatchMerger", "ModelVisionRotaryEmbedding", "ModuleList", "None", "T", "Tensor", "True", "_", "__init__", "_no_split_modules", "append", "arange", "base_h", "base_h_ceil", "blk", "block_cols", "block_rows", "blocks", "cat", "class", "clip", "col_idx", "config", "coords", "cos", "cu_seqlens", "cumsum", "deepstack_feature", "deepstack_feature_lists", "deepstack_merger_list", "deepstack_visual_indexes", "def", "depth", "device", "dh", "dim", "dtype", "dw", "else", "emb", "embeddings", "empty", "enumerate", "expand", "extend", "fast_pos_embed_interpolate", "flatten", "for", "forward", "freq_table", "gradient_checkpointing", "grid_hs", "grid_thw", "grid_ts", "grid_ws", "h", "h_idxs", "h_idxs_ceil", "h_idxs_floor", "head_dim", "height", "hidden_size", "hidden_states", "i", "idx_list", "idx_tensor", "if", "in", "index", "indices", "inputs", "int", "int32", "intra_col", "intra_row", "is_tracing", "item", "jit", "kwargs", "layer_num", "len", "linspace", "long", "max", "max_hw", "merge_size", "merged_h", "merged_w", "merger", "nn", "num_frames", "num_grid_per_side", "num_heads", "num_position_embeddings", "num_tokens", "offset", "pad", "patch_embed", "patch_pos_embeds", "patch_pos_embeds_permute", "patch_size", "permute", "pos_embed", "pos_embeds", "pos_ids", "position_embeddings", "prod", "range", "repeat", "repeat_interleave", "reshape", "return", "rot_pos_emb", "rotary_pos_emb", "row_idx", "self", "seq_len", "shape", "sin", "size", "spatial_merge_size", "spatial_merge_unit", "split", "stack", "sum", "super", "t", "tensor", "tolist", "torch", "total_tokens", "use_postshuffle_norm", "value", "view", "w", "w_idxs", "w_idxs_ceil", "w_idxs_floor", "weight", "weight_list", "weight_tensor", "weights", "width", "zip"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "apply_interleaved_mrope", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "enumerate", "expand", "float", "for", "forward", "freqs", "freqs_t", "get", "hasattr", "idx", "if", "in", "inv_freq", "inv_freq_expanded", "is", "isinstance", "length", "max_position_embeddings", "max_seq_len_cached", "mps", "mrope_section", "ndim", "nn", "no_grad", "not", "offset", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "slice", "start", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextRMSNorm", "ModelTextRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "_deepstack_process", "_no_split_modules", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "check_model_inputs", "class", "clone", "config", "create_causal_mask", "decoder_layer", "deepstack_visual_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "enumerate", "eps", "exactly", "expand", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "len", "list", "local_this", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "text_position_ids", "to", "torch", "tuple", "use_cache", "view", "visual_embeds", "visual_pos_masks", "vocab_size"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeModelOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeModel": ["Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelTextDecoderLayer", "ModelTextModel", "ModelVisionBlock", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "_from_config", "_no_split_modules", "accepts_loss_kwargs", "add", "all", "and", "append", "arange", "argwhere", "attention_mask", "attention_mask_tensor", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "cache_position", "can_return_tuple", "cat", "class", "config", "cumsum", "decoder", "deepstack_image_embeds", "deepstack_video_embeds", "deepstack_visual_embeds", "def", "delta", "device", "diagonal", "dict", "dim", "dim1", "dim2", "do", "dtype", "ed", "ed_image", "ed_video", "elif", "else", "embed_joint", "enumerate", "exactly", "expand", "expand_as", "f", "features", "finfo", "flatten", "for", "forward", "full_attention", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_seq_length", "get_video_features", "grid_thw", "h", "h_index", "hidden_states", "i", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_index", "image_mask", "image_mask_joint", "image_nums", "image_token_id", "img_embed", "in", "index", "input_ids", "input_tokens", "inputs_embeds", "int", "is", "is_floating_point", "is_torchdynamo_compiling", "isinstance", "item", "keepdim", "kwargs", "language_model", "last_hidden_state", "len", "list", "llm_grid_h", "llm_grid_t", "llm_grid_w", "llm_pos_ids_list", "llm_positions", "long", "masked_fill_", "masked_scatter", "match", "max", "max_position_ids", "min", "mrope_position_deltas", "must", "n_image_tokens", "n_video_tokens", "ndim", "new_zeros", "not", "numel", "of", "one", "ones", "ones_like", "or", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prefill_compiled_stage", "prefill_noncompiled_stage", "prod", "r", "raise", "range", "remain_images", "remain_videos", "repeat_interleave", "reshape", "return", "rope_deltas", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "special_video_mask", "specify", "split", "split_sizes", "squeeze", "st", "st_idx", "stack", "sum", "super", "t", "t_index", "tensor", "text_config", "text_len", "to", "tokens", "tolist", "torch", "total_input_ids", "tuple", "type", "unsqueeze", "value", "vid_embed", "video", "video_embeds", "video_features", "video_grid_thw", "video_index", "video_mask", "video_mask_joint", "video_nums", "video_token_id", "view", "vision_config", "vision_start_indices", "vision_start_token_id", "vision_tokens", "visual", "visual_pos_masks", "w", "w_index", "zeros", "zip"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen3_vl_moe/modeling_qwen3_vl_moe.py:Qwen3VLMoeForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "If", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_expand_dict_for_generation", "_expand_dict_for_generation_visual", "_expand_inputs_for_generation", "_get_image_nums_and_video_nums", "_repeat_interleave_samples", "_tied_weights_keys", "accepts_loss_kwargs", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "defined", "device", "dict", "dict_to_expand", "dim", "dims", "dtype", "elif", "else", "encoder_outputs", "expand_size", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "get_video_features", "hidden_size", "hidden_states", "if", "image_grid_thw", "image_mask", "image_nums", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "isinstance", "key", "kwargs", "labels", "language_model", "lengths", "list", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "make", "model", "model_inputs", "model_kwargs", "nn", "not", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "prod", "property", "r", "raise", "repeat", "repeat_args", "repeat_interleave", "repeat_times", "result", "return", "roll", "rope_deltas", "sample", "samples", "second_per_grid_ts", "self", "set_decoder", "set_input_embeddings", "shifts", "slice", "slice_indices", "split", "str", "sum", "super", "sure", "tensor", "text_config", "that", "torch", "tuple", "use_cache", "value", "video_grid_thw", "video_mask", "video_nums", "video_token_id", "vision_first_mask", "vision_start_mask", "vision_start_token_id", "visual", "visual_keys", "vocab_size", "weight", "x"], "evolla/modeling_evolla.py:create_position_ids_from_input_ids": ["Model_position_ids_from_input_ids", "cumsum", "def", "dim", "incremental_indices", "input_ids", "int", "long", "mask", "ne", "padding_idx", "return", "torch", "type_as"], "evolla/modeling_evolla.py:EvollaSaProtEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelSaProtEmbeddings", "Module", "None", "__init__", "absolute", "and", "arange", "attention_mask", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "def", "device", "dropout", "dtype", "else", "emb_layer_norm_before", "embeddings", "eps", "expand", "float", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm", "layer_norm_eps", "long", "mask_ratio_observed", "mask_ratio_train", "mask_token_id", "masked_fill", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "sequence_length", "shape", "size", "src_lengths", "sum", "super", "to", "token_dropout", "torch", "unsqueeze", "vocab_size", "word_embeddings"], "evolla/modeling_evolla.py:rotate_half_esm": ["Model_half_esm", "cat", "chunk", "def", "dim", "return", "torch", "x", "x1", "x2"], "evolla/modeling_evolla.py:apply_rotary_pos_emb_esm": ["Model_rotary_pos_emb_esm", "cos", "def", "return", "rotate_half_esm", "shape", "sin", "x"], "evolla/modeling_evolla.py:EvollaSaProtRotaryEmbedding": ["ModelSaProtRotaryEmbedding", "Module", "None", "Tensor", "__init__", "_cos_cached", "_seq_len_cached", "_sin_cached", "_update_cos_sin_tables", "apply_rotary_pos_emb_esm", "arange", "cat", "class", "cos", "def", "device", "dim", "dtype", "emb", "float", "forward", "freqs", "if", "int", "int64", "inv_freq", "k", "nn", "or", "outer", "q", "register_buffer", "return", "self", "seq_dimension", "seq_len", "shape", "sin", "super", "t", "to", "torch", "tuple", "type_as", "x"], "evolla/modeling_evolla.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "causal_mask", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "float", "float32", "functional", "hasattr", "head_mask", "if", "in", "is", "key", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "nn", "not", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "seq_length", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "evolla/modeling_evolla.py:EvollaSaProtSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "ESM", "Embedding", "False", "FloatTensor", "Linear", "ModelSaProtRotaryEmbedding", "ModelSaProtSelfAttention", "Module", "None", "Optional", "Set", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "batch_size", "class", "config", "contiguous", "current_states", "def", "dim", "distance_embedding", "does", "dropout", "eager", "eager_attention_forward", "elif", "else", "embedding_size", "embeddings", "encoder_attention_mask", "encoder_hidden_states", "explicitly", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "in", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "key", "key_layer", "kwargs", "layer_idx", "max_position_embeddings", "model", "multiple", "nn", "not", "num_attention_heads", "number", "of", "or", "position_embedding_type", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "rotary", "rotary_embeddings", "scaling", "self", "seq_length", "set_attn_implementation", "shape", "size", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "value", "value_layer", "view", "with"], "evolla/modeling_evolla.py:EvollaSaProtSelfOutput": ["Dropout", "Linear", "ModelSaProtSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super"], "evolla/modeling_evolla.py:EvollaSaProtAttention": ["False", "LayerNorm", "ModelSaProtAttention", "ModelSaProtSelfAttention", "ModelSaProtSelfOutput", "Module", "None", "TransformersKwargs", "Unpack", "_", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attn_output", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "eps", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "hidden_states_ln", "if", "index", "is_cross_attention", "key", "kwargs", "layer_idx", "layer_norm_eps", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "union", "value"], "evolla/modeling_evolla.py:gelu": ["Model", "def", "erf", "math", "return", "sqrt", "torch", "x"], "evolla/modeling_evolla.py:EvollaSaProtIntermediate": ["Linear", "ModelSaProtIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "forward", "gelu", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "evolla/modeling_evolla.py:EvollaSaProtOutput": ["Dropout", "Linear", "ModelSaProtOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super"], "evolla/modeling_evolla.py:EvollaSaProtLayer": ["AttributeError", "GradientCheckpointingLayer", "If", "LayerNorm", "ModelSaProtAttention", "ModelSaProtIntermediate", "ModelSaProtLayer", "ModelSaProtOutput", "None", "RuntimeError", "TransformersKwargs", "True", "Unpack", "__init__", "a", "add_cross_attention", "added", "and", "are", "as", "attention", "attention_mask", "attention_output", "attention_output_ln", "be", "by", "chunk_size_feed_forward", "class", "config", "cross", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_cross_attention", "is_decoder", "kwargs", "layer_norm_eps", "layer_output", "layers", "model", "nn", "not", "output", "passed", "raise", "return", "self", "seq_len_dim", "setting", "should", "super", "to", "used", "with"], "evolla/modeling_evolla.py:EvollaSaProtEncoder": ["BaseModelOutputWithCrossAttentions", "False", "LayerNorm", "ModelSaProtEncoder", "ModelSaProtLayer", "Module", "ModuleList", "None", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "else", "emb_layer_norm_after", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eps", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_norm_eps", "nn", "not", "num_hidden_layers", "range", "return", "self", "super"], "evolla/modeling_evolla.py:EvollaSaProtPooler": ["Linear", "ModelSaProtPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "evolla/modeling_evolla.py:EvollaSaProtPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelSaProtLayer", "ModelSaProtPreTrainedModel", "ModelSaProtSelfAttention", "None", "OutputRecorder", "PreTrainedModel", "SaProtConfig", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "attention", "attentions", "bias", "class", "config", "cross_attentions", "crossattention", "data", "def", "elif", "fill_", "hidden_states", "if", "index", "initializer_range", "is", "isinstance", "layer_name", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "evolla/modeling_evolla.py:EvollaSaProtProteinEncoder": ["BaseModelOutputWithPoolingAndCrossAttentions", "FutureWarning", "ModelSaProtEmbeddings", "ModelSaProtEncoder", "ModelSaProtPreTrainedModel", "ModelSaProtProteinEncoder", "ModuleUtilsMixin", "None", "Optional", "SaProtConfig", "Tensor", "The", "Transformers", "Union", "ValueError", "Wrong", "__init__", "_prune_heads", "and", "argument", "attention", "attention_mask", "attentions", "batch_size", "be", "check_model_inputs", "class", "config", "create_extended_attention_mask_for_decoder", "cross_attentions", "def", "deprecated", "device", "dim", "dtype", "elif", "else", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "f", "finfo", "for", "forward", "get_extended_attention_mask", "get_input_embeddings", "get_parameter_dtype", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "int", "is", "is_decoder", "items", "last_hidden_state", "layer", "min", "not", "of", "ones", "or", "prune_heads", "raise", "removed", "return", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "super", "to", "torch", "tuple", "v5", "value", "warn", "warnings", "will", "word_embeddings"], "evolla/modeling_evolla.py:EvollaSequenceCompressorAttention": ["False", "LayerNorm", "Linear", "ModelSequenceCompressorAttention", "Module", "None", "True", "__init__", "amax", "attn", "bias", "bool", "bs", "cat", "chunk", "class", "def", "detach", "device", "dim", "dim_head", "forward", "h", "heads", "inner_dim", "k", "keepdim", "kv_input", "latents", "mask", "mask_exp", "masked_fill", "matmul", "nh", "nn", "norm_latents", "norm_media", "okd", "ones", "ones_exp", "out", "permute", "q", "reshape", "return", "scale", "self", "shape", "sim", "size", "skd", "softmax", "super", "to", "to_kv", "to_out", "to_q", "torch", "transpose", "v", "view", "x"], "evolla/modeling_evolla.py:EvollaFeedForward": ["False", "GELU", "LayerNorm", "Linear", "ModelFeedForward", "Module", "__init__", "activation", "bias", "class", "def", "dim", "fc1", "fc2", "forward", "inner_dim", "int", "mult", "nn", "norm", "return", "self", "super", "x"], "evolla/modeling_evolla.py:EvollaSequenceCompressorResampler": ["LayerNorm", "Linear", "ModelConfig", "ModelFeedForward", "ModelSequenceCompressorAttention", "ModelSequenceCompressorResampler", "Module", "ModuleList", "None", "Parameter", "True", "_", "__init__", "append", "attn", "b", "bs", "cat", "class", "config", "def", "device", "dim", "dim_head", "dtype", "embeds", "ff", "for", "forward", "heads", "hidden_size", "in", "latent_mask", "latents", "layers", "mask", "mult", "nn", "norm", "num_latents", "ones", "protein_encoder_config", "protein_projector", "protein_repr_dim", "randn", "range", "requires_grad", "resampler_depth", "resampler_dim_head", "resampler_ff_mult", "resampler_heads", "resampler_num_latents", "return", "self", "shape", "super", "to", "torch", "transformed_feature", "view"], "evolla/modeling_evolla.py:EvollaProteinEncoderModelOutput": ["FloatTensor", "ModelOutput", "ModelProteinEncoderModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "sequence_compressor_output", "torch", "tuple"], "evolla/modeling_evolla.py:EvollaProteinEncoder": ["FloatTensor", "LongTensor", "ModelConfig", "ModelProteinEncoder", "ModelProteinEncoderModelOutput", "ModelSaProtProteinEncoder", "ModelSequenceCompressorResampler", "Module", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "forward", "input_ids", "kwargs", "last_hidden_state", "model", "nn", "protein_embeds", "protein_encoder_config", "protein_output", "return", "self", "sequence_compressor_output", "sequence_compressor_resampler", "sequence_repr", "super", "torch"], "evolla/modeling_evolla.py:EvollaSequenceAlignerCrossAttention": ["At", "Dropout", "Linear", "ModelFeedForward", "ModelRMSNorm", "ModelSequenceAlignerCrossAttention", "Module", "None", "Optional", "Parameter", "Softmax", "T", "True", "ValueError", "_", "__init__", "aligner_attention_probs_dropout_prob", "aligner_enable_bias", "aligner_ffn_mult", "all_head_size", "amax", "and", "any", "attention", "attention_head_size", "attention_mask", "attention_norm", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "attn_weights", "be", "bias", "bool", "bs", "cat", "class", "config", "context_layer", "contiguous", "cross", "cross_attention", "def", "deprecate_kwarg", "detach", "device", "dim", "dropout", "dtype", "else", "enable_bias", "expand", "ff", "ffn_mult", "finfo", "for", "forward", "gate_attention", "gate_ffw", "hidden_size", "hidden_states", "if", "in", "int", "is", "keepdim", "key_layer", "key_layer_msa", "key_layer_protein", "key_layer_structure", "key_msa", "key_protein", "key_structure", "kv_attn_mask", "least", "masked_fill", "matmul", "min", "modality", "msa_batch_mask", "msa_encoder_dim", "msa_key_value_states", "msa_kv_attn_mask", "msa_kv_seq_len", "msa_kv_states", "new_context_layer_shape", "new_key_layer_shape", "new_name", "new_query_layer_shape", "new_value_layer_shape", "nn", "not", "num_attention_heads", "one", "ones", "or", "out_proj", "past_key_value", "past_key_values", "permute", "protein_batch_mask", "protein_encoder_dim", "protein_key_value_states", "protein_kv_attn_mask", "protein_kv_seq_len", "protein_kv_states", "provided", "query", "query_attn_mask", "query_layer", "query_states", "raise", "residual", "return", "scale", "self", "shape", "should", "size", "structure_batch_mask", "structure_encoder_dim", "structure_key_value_states", "structure_kv_attn_mask", "structure_kv_seq_len", "structure_kv_states", "super", "tanh", "tensor", "to", "torch", "transpose", "value_layer", "value_layer_msa", "value_layer_protein", "value_layer_structure", "value_msa", "value_protein", "value_structure", "version", "view"], "evolla/modeling_evolla.py:EvollaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "evolla/modeling_evolla.py:EvollaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "evolla/modeling_evolla.py:EvollaMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "evolla/modeling_evolla.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "evolla/modeling_evolla.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "evolla/modeling_evolla.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "evolla/modeling_evolla.py:EvollaAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "evolla/modeling_evolla.py:EvollaDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSequenceAlignerCrossAttention", "None", "Optional", "Tensor", "_", "__init__", "adapter", "aligner_num_add_layers", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hasattr", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "max", "mlp", "msa_batch_mask", "msa_kv_states", "new_name", "num_hidden_layers", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "protein_batch_mask", "protein_encoder_dim", "protein_kv_states", "query_attn_mask", "query_states", "residual", "return", "rms_norm_eps", "self", "self_attn", "structure_batch_mask", "structure_kv_states", "super", "torch", "tuple", "use_cache", "version"], "evolla/modeling_evolla.py:EvollaPreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSequenceAlignerCrossAttention", "ModelSequenceCompressorResampler", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention_norm", "attentions", "base_model_prefix", "class", "config", "data", "def", "elif", "fill_", "gate_attention", "gate_ffw", "hidden_states", "if", "initializer_range", "isinstance", "latents", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight", "zero_"], "evolla/modeling_evolla.py:EvollaModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelProteinEncoder", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_input_embeddings", "get_seq_length", "getattr", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "msa_batch_mask", "msa_feats", "msa_kv_states", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "protein_attention_mask", "protein_batch_mask", "protein_encoder", "protein_feats", "protein_input_ids", "protein_kv_states", "protein_outputs", "query_attn_mask", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "sequence_compressor_output", "set_input_embeddings", "shape", "specify", "structure_batch_mask", "structure_feats", "structure_kv_states", "super", "tensor", "torch", "tuple", "unsqueeze", "use_cache", "value", "vocab_size"], "evolla/modeling_evolla.py:EvollaForProteinText2Text": ["CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForProteinText2Text", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "__init__", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "def", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_outputs", "logits", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "post_init", "protein_attention_mask", "protein_input_ids", "r", "return", "self", "set_input_embeddings", "super", "torch", "use_cache", "value", "vocab_size"], "sam2/modeling_sam2.py:Sam2VisionEncoderOutput": ["FloatTensor", "ModelOutput", "ModelVisionEncoderOutput", "None", "Optional", "attentions", "class", "fpn_hidden_states", "fpn_position_encoding", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "sam2/modeling_sam2.py:Sam2ImageSegmentationOutput": ["FloatTensor", "ModelImageSegmentationOutput", "ModelOutput", "None", "Optional", "class", "image_embeddings", "iou_scores", "mask_decoder_attentions", "object_score_logits", "pred_masks", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states"], "sam2/modeling_sam2.py:Sam2PatchEmbeddings": ["Conv2d", "ModelHieraDetConfig", "ModelPatchEmbeddings", "Module", "_", "__init__", "class", "config", "def", "embeddings", "forward", "height", "hidden_size", "kernel_size", "nn", "num_channels", "padding", "patch_kernel_size", "patch_padding", "patch_stride", "permute", "pixel_values", "projection", "r", "return", "self", "shape", "stride", "super", "width"], "sam2/modeling_sam2.py:Sam2SinePositionEmbedding": ["False", "ModelSinePositionEmbedding", "Module", "None", "Optional", "Size", "Tensor", "True", "Union", "ValueError", "__init__", "and", "arange", "be", "bool", "cat", "class", "compile_compatible_method_lru_cache", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "else", "eps", "flatten", "float", "floor", "forward", "if", "int", "int64", "is", "mask", "math", "maxsize", "nn", "normalize", "not", "not_mask", "num_pos_feats", "passed", "permute", "pi", "pos", "pos_x", "pos_y", "raise", "return", "rounding_mode", "scale", "self", "shape", "should", "sin", "stack", "str", "super", "temperature", "to", "torch", "x_embed", "y_embed", "zeros"], "sam2/modeling_sam2.py:Sam2VisionNeck": ["Conv2d", "F", "False", "ModelSinePositionEmbedding", "ModelVisionConfig", "ModelVisionNeck", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "align_corners", "antialias", "append", "backbone_channel_list", "class", "config", "convs", "def", "device", "dtype", "else", "float32", "for", "forward", "fpn_hidden_size", "fpn_hidden_states", "fpn_kernel_size", "fpn_padding", "fpn_position_encoding", "fpn_stride", "fpn_top_down_levels", "hidden_states", "i", "if", "in", "in_channels", "interpolate", "kernel_size", "lateral_features", "len", "mode", "n", "nearest", "nn", "normalize", "not", "num_pos_feats", "or", "out_channels", "padding", "permute", "position_encoding", "prev_features", "prev_position_encoding", "range", "return", "scale_factor", "self", "shape", "stride", "super", "to", "top_down_features", "torch", "tuple"], "sam2/modeling_sam2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "sam2/modeling_sam2.py:do_pool": ["False", "None", "Optional", "Tensor", "ceil_mode", "def", "do_pool", "functional", "if", "int", "is", "kernel_size", "max_pool2d", "nn", "permute", "query_stride", "return", "stride", "torch", "x"], "sam2/modeling_sam2.py:Sam2MultiScaleAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelHieraDetConfig", "ModelMultiScaleAttention", "Module", "None", "Optional", "Tensor", "_", "__init__", "_attn_implementation", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "class", "config", "def", "dim", "dim_out", "do_pool", "dtype", "eager", "eager_attention_forward", "float32", "forward", "functional", "head_dim", "height", "hidden_states", "if", "int", "is_causal", "key", "kwargs", "nn", "num_attention_heads", "proj", "qkv", "query", "query_stride", "reshape", "return", "scale", "scaling", "self", "shape", "softmax", "super", "to", "torch", "transpose", "tuple", "unbind", "value", "width"], "sam2/modeling_sam2.py:Sam2FeedForward": ["ACT2FN", "F", "False", "Linear", "ModelFeedForward", "Module", "ModuleList", "_", "__init__", "activation", "bool", "class", "def", "for", "forward", "hidden_dim", "hidden_states", "if", "in", "input_dim", "int", "layer", "layers", "nn", "num_layers", "output_dim", "proj_in", "proj_out", "range", "relu", "return", "self", "sigmoid", "sigmoid_output", "str", "super"], "sam2/modeling_sam2.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "functional", "height", "hidden_state", "nn", "num_channels", "pad", "pad_height", "pad_width", "padded_height", "padded_width", "permute", "return", "shape", "view", "width"], "sam2/modeling_sam2.py:window_unpartition": ["Model_size", "Model_unpartition", "Models", "batch_size", "contiguous", "def", "height", "height_width", "hidden_state", "pad_height_width", "padded_height", "padded_width", "permute", "return", "shape", "view", "width"], "sam2/modeling_sam2.py:Sam2MultiScaleBlock": ["FloatTensor", "GradientCheckpointingLayer", "H", "LayerNorm", "Linear", "ModelFeedForward", "ModelHieraDetConfig", "ModelMultiScaleAttention", "ModelMultiScaleBlock", "None", "Tensor", "TransformersKwargs", "Unpack", "W", "__init__", "activation", "and", "attn", "attn_output", "block_idx", "class", "config", "def", "dim", "dim_out", "do_pool", "else", "embed_dim_per_stage", "eps", "forward", "global_attention_blocks", "hidden_act", "hidden_states", "if", "in", "int", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "layernorm_output", "mlp", "mlp_ratio", "nn", "num_attention_heads", "num_attention_heads_per_stage", "num_layers", "num_query_pool_stages", "pad_h", "pad_hw", "pad_w", "proj", "query_stride", "residual", "return", "self", "shape", "stage_idx", "super", "torch", "total_block_idx", "window_partition", "window_size", "window_size_per_stage", "window_unpartition"], "sam2/modeling_sam2.py:Sam2HieraDetModelOutput": ["FloatTensor", "ModelHieraDetModelOutput", "ModelOutput", "None", "Optional", "class", "intermediate_hidden_states", "last_hidden_state", "r", "torch", "tuple"], "sam2/modeling_sam2.py:Sam2PreTrainedModel": ["Conv2d", "ConvTranspose2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelHieraDetModel", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_supports_attention_backend", "_supports_flash_attn_2", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "config_class", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "no_memory_embedding", "normal_", "not", "padding_idx", "pixel_values", "pos_embed", "pos_embed_window", "self", "std", "weight", "zero_"], "sam2/modeling_sam2.py:Sam2HieraDetModel": ["F", "FloatTensor", "ModelHieraDetConfig", "ModelHieraDetModel", "ModelHieraDetModelOutput", "ModelMultiScaleAttention", "ModelMultiScaleBlock", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "_get_pos_embed", "append", "attentions", "bicubic", "block", "block_idx", "block_module", "blocks", "blocks_per_stage", "check_model_inputs", "class", "config", "config_class", "cumsum", "def", "enumerate", "for", "forward", "get_input_embeddings", "h", "have", "hidden_size", "hidden_states", "hw", "i", "if", "in", "int", "intermediate_hidden_states", "interpolate", "is", "kwargs", "last_hidden_state", "main_input_name", "mode", "nn", "np", "patch_embed", "permute", "pixel_values", "pos_embed", "pos_embed_window", "raise", "range", "return", "self", "shape", "size", "specify", "stage_ends", "stage_idx", "super", "tile", "to", "tolist", "torch", "total_block_idx", "tuple", "w", "window_embed", "window_positional_embedding_background_size", "window_size_per_stage", "x", "y", "zeros", "zip"], "sam2/modeling_sam2.py:Sam2VisionModel": ["AutoModel", "FloatTensor", "ModelMultiScaleAttention", "ModelMultiScaleBlock", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEncoderOutput", "ModelVisionModel", "ModelVisionNeck", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "attentions", "backbone", "backbone_config", "backbone_output", "check_model_inputs", "class", "config", "config_class", "def", "forward", "fpn_hidden_states", "fpn_position_encoding", "from_config", "get_input_embeddings", "have", "hidden_states", "if", "intermediate_hidden_states", "is", "kwargs", "last_hidden_state", "main_input_name", "neck", "num_feature_levels", "pixel_values", "post_init", "raise", "return", "self", "specify", "super", "to", "torch", "tuple"], "sam2/modeling_sam2.py:Sam2PositionalEmbedding": ["ModelPositionalEmbedding", "ModelPromptEncoderConfig", "Module", "None", "__init__", "cat", "class", "clone", "config", "coordinates", "cos", "def", "dim", "dtype", "float32", "forward", "hidden_size", "if", "input_coords", "input_shape", "is", "nn", "not", "np", "pi", "positional_embedding", "randn", "register_buffer", "return", "scale", "self", "sin", "super", "to", "torch"], "sam2/modeling_sam2.py:Sam2MaskEmbedding": ["ACT2FN", "Conv2d", "ModelLayerNorm", "ModelMaskEmbedding", "ModelPromptEncoderConfig", "Module", "__init__", "activation", "channels_first", "class", "config", "conv1", "conv2", "conv3", "data_format", "def", "dense_embeddings", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "layer_norm_eps", "mask_input_channels", "masks", "nn", "return", "self", "stride", "super"], "sam2/modeling_sam2.py:Sam2PromptEncoder": ["Embedding", "If", "ModelMaskEmbedding", "ModelPositionalEmbedding", "ModelPromptEncoder", "ModelPromptEncoderConfig", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_embed_boxes", "_embed_points", "also", "are", "batch_size", "be", "bool", "box_embeddings", "boxes", "cat", "clamp", "class", "config", "constant", "coords", "corner_embedding", "def", "dense_embeddings", "dim", "else", "expand", "expand_as", "forward", "functional", "hidden_size", "if", "image_embedding_size", "image_size", "input_boxes", "input_image_size", "input_labels", "input_masks", "input_points", "input_shape", "is", "labels", "mask_embed", "mask_input_size", "min", "mode", "must", "nn", "no_mask_embed", "not", "not_a_point_embed", "num_point_embeddings", "pad", "patch_size", "point_embed", "point_embedding", "point_embeddings", "points", "provided", "raise", "reshape", "return", "self", "shape", "shared_embedding", "sparse_embeddings", "super", "torch", "tuple", "unsqueeze", "value", "view", "weight", "where", "zeros_like"], "sam2/modeling_sam2.py:Sam2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "attention_downModelple_rate", "attention_interface", "attention_mask", "attention_similarity", "attn_output", "attn_weights", "batch_size", "class", "config", "contiguous", "def", "downModelple_rate", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_size", "if", "internal_dim", "is", "is_causal", "k_proj", "key", "kwargs", "new_shape", "nn", "num_attention_heads", "o_proj", "point_batch_size", "q_proj", "query", "reshape", "return", "scaling", "self", "shape", "super", "torch", "transpose", "tuple", "v_proj", "value", "view"], "sam2/modeling_sam2.py:Sam2TwoWayAttentionBlock": ["False", "LayerNorm", "ModelAttention", "ModelFeedForward", "ModelMaskDecoderConfig", "ModelTwoWayAttentionBlock", "Module", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_similarity", "attn_out", "bool", "class", "config", "cross_attn_image_to_token", "cross_attn_token_to_image", "def", "downModelple_rate", "else", "forward", "hidden_size", "if", "key", "key_point_embedding", "keys", "kwargs", "layer_norm1", "layer_norm2", "layer_norm3", "layer_norm4", "mlp", "mlp_dim", "mlp_out", "nn", "num_hidden_layers", "num_layers", "queries", "query", "query_point_embedding", "return", "self", "self_attn", "skip_first_layer_pe", "super", "value"], "sam2/modeling_sam2.py:Sam2TwoWayTransformer": ["BaseModelOutput", "LayerNorm", "ModelAttention", "ModelMaskDecoderConfig", "ModelTwoWayAttentionBlock", "ModelTwoWayTransformer", "Module", "ModuleList", "None", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "_", "__init__", "an", "append", "attention_similarity", "attn_out", "class", "config", "def", "final_attn_token_to_image", "flatten", "for", "forward", "have", "hidden_size", "i", "if", "image_embedding", "image_embeddings", "image_positional_embeddings", "in", "is", "key", "key_point_embedding", "keys", "kwargs", "layer", "layer_norm_final_attn", "layers", "nn", "not", "num_hidden_layers", "permute", "point_embeddings", "queries", "query", "query_point_embedding", "raise", "range", "return", "self", "skip_first_layer_pe", "specify", "super", "target_embedding", "to", "tuple", "unsqueeze", "value"], "sam2/modeling_sam2.py:Sam2LayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "sam2/modeling_sam2.py:Sam2MaskDecoder": ["Conv2d", "ConvTranspose2d", "Embedding", "GELU", "ModelFeedForward", "ModelLayerNorm", "ModelMaskDecoder", "ModelMaskDecoderConfig", "ModelTwoWayTransformer", "Model_tokens_out", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "_", "__init__", "_dynamic_multimask_via_stability", "_get_stability_scores", "activation", "all_iou_scores", "all_mask_logits", "and", "area_i", "area_u", "argmax", "attention_similarity", "batch_size", "best_multimask_iou_scores", "best_multimask_logits", "best_scores_inds", "best_scores_inds_expanded", "bool", "cat", "channels_first", "class", "config", "conv_s0", "conv_s1", "current_mlp", "data_format", "def", "dense_prompt_embeddings", "dim", "dtype", "dynamic_multimask_stability_delta", "dynamic_multimask_stability_thresh", "dynamic_multimask_via_stability", "elif", "else", "expand", "expand_as", "feat_s0", "feat_s1", "flatten", "float", "for", "forward", "gather", "height", "hidden_size", "high_resolution_features", "hyper_in", "hyper_in_list", "i", "if", "image_embeddings", "image_positional_embeddings", "in", "iou_head_depth", "iou_head_hidden_dim", "iou_pred", "iou_prediction_head", "iou_scores_out", "iou_token", "iou_token_out", "is_stable", "kernel_size", "kwargs", "list", "mask_logits", "mask_logits_out", "mask_slice", "mask_tokens", "mask_tokens_out", "masks", "mlps_list", "multimask_iou_scores", "multimask_logits", "multimask_output", "nn", "not", "num_channels", "num_mask_tokens", "num_multimask_outputs", "obj_score_token", "object_score_logits", "output_hypernetworks_mlps", "output_tokens", "point_batch_size", "point_embeddings", "pred_obj_score_head", "range", "repeat", "repeat_interleave", "return", "self", "shape", "sigmoid_output", "singlemask_iou_scores", "singlemask_logits", "size", "slice", "sparse_prompt_embeddings", "stability_delta", "stability_scores", "stack", "stride", "sum", "super", "target_embedding", "to", "tokens", "torch", "training", "transformer", "transpose", "tuple", "unsqueeze", "upscale_conv1", "upscale_conv2", "upscale_layer_norm", "upscaled_embedding", "view", "weight", "where", "width"], "sam2/modeling_sam2.py:Sam2Model": ["AutoModel", "Exactly", "F", "False", "FloatTensor", "Got", "LongTensor", "ModelConfig", "ModelImageSegmentationOutput", "ModelMaskDecoder", "ModelModel", "ModelPositionalEmbedding", "ModelPreTrainedModel", "ModelPromptEncoder", "ModelTwoWayAttentionBlock", "ModelVisionEncoderOutput", "None", "Optional", "OutputRecorder", "Parameter", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "_", "__init__", "_attn_implementation", "_can_record_outputs", "_keys_to_ignore_on_load_missing", "_keys_to_ignore_on_load_unexpected", "_tie_weights", "_tied_weights_keys", "align_corners", "and", "antialias", "as", "attention_similarity", "attentions", "auto_docstring", "backbone_feature_sizes", "batch_size", "be", "bilinear", "bool", "bounding", "box", "boxes", "check_model_inputs", "class", "config", "conv_s0", "conv_s1", "cumsum", "data", "def", "dense_embeddings", "dense_prompt_embeddings", "device", "dim", "dtype", "else", "f", "feat", "feat_size", "feature_map", "feature_map_position_embedding", "feature_maps", "feature_maps_position_embeddings", "flatten", "float", "for", "forward", "fpn_hidden_size", "fpn_hidden_states", "fpn_position_encoding", "from_config", "get_image_embeddings", "get_image_features", "get_image_wide_positional_embeddings", "get_input_embeddings", "get_prompt_embeddings", "grid", "hidden_dim", "hidden_states", "high_resolution_features", "if", "image_embedding_size", "image_embeddings", "image_positional_embeddings", "in", "index", "input", "input_boxes", "input_labels", "input_masks", "input_points", "int", "int32", "interpolate", "iou_scores", "is", "kwargs", "list", "low_res_multimasks", "many", "mask_decoder", "mask_decoder_attentions", "mask_decoder_config", "mask_downModelple", "mask_input_size", "memory_", "mode", "multimask_output", "must", "nn", "no_grad", "no_memory_embedding", "no_memory_positional_encoding", "no_object_pointer", "not", "num_feature_levels", "object_pointer_proj", "object_score_logits", "occlusion_spatial_embedding_parameter", "of", "one", "ones", "ones_like", "or", "per", "permute", "pixel_values", "points", "positional_embedding", "post_init", "pred_masks", "prompt_encoder", "prompt_encoder_config", "prompt_output", "provide", "provided", "r", "raise", "repeat", "return", "self", "shape", "shared_embedding", "shared_image_embedding", "should", "size", "sparse_embeddings", "sparse_prompt_embeddings", "stack", "super", "target_device", "target_dtype", "target_embedding", "temporal_positional_encoding_projection_layer", "to", "torch", "tuple", "unsqueeze", "view", "vision_attentions", "vision_config", "vision_encoder", "vision_hidden_states", "vision_outputs", "x_embed", "y_embed", "zeros", "zip"], "pixtral/modeling_pixtral.py:position_ids_in_meshgrid": ["Model_ids_in_meshgrid", "Models", "append", "arange", "cat", "chunk", "def", "dim", "for", "h_grid", "height", "ids", "ij", "in", "indexing", "max_width", "mesh", "meshgrid", "patch", "patch_embeds_list", "reshape", "return", "shape", "stack", "torch", "v_grid", "width"], "pixtral/modeling_pixtral.py:PixtralRotaryEmbedding": ["False", "ModelRotaryEmbedding", "Module", "None", "Tensor", "__init__", "and", "arange", "autocast", "base", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "float", "forward", "freqs", "freqs_h", "freqs_w", "h", "head_dim", "if", "image_size", "inv_freq", "isinstance", "max_patches_per_side", "mps", "nn", "no_grad", "outer", "patch_size", "persistent", "position_ids", "register_buffer", "repeat", "reshape", "return", "rope_theta", "rope_type", "self", "sin", "str", "super", "to", "torch", "type", "w", "with", "x"], "pixtral/modeling_pixtral.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "pixtral/modeling_pixtral.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "pixtral/modeling_pixtral.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "pixtral/modeling_pixtral.py:PixtralAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "Unpack", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "bool", "class", "config", "contiguous", "cos", "def", "device", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "flash_attention_2", "forward", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "key_states", "kwargs", "nn", "non_blocking", "not", "num_attention_heads", "num_heads", "o_proj", "output_attentions", "patches", "position_embeddings", "position_ids", "q_proj", "query_states", "reshape", "return", "scaling", "self", "sin", "size", "super", "to", "torch", "training", "transpose", "tuple", "unsqueeze_dim", "v_proj", "value_states", "view"], "pixtral/modeling_pixtral.py:PixtralMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "pixtral/modeling_pixtral.py:PixtralRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "pixtral/modeling_pixtral.py:PixtralAttentionLayer": ["FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelAttentionLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "__init__", "attention", "attention_mask", "attention_norm", "attn_weights", "bool", "class", "config", "def", "eps", "feed_forward", "ffn_norm", "forward", "hidden_size", "hidden_states", "if", "kwargs", "output_attentions", "outputs", "position_embeddings", "residual", "return", "self", "super", "torch", "tuple"], "pixtral/modeling_pixtral.py:PixtralTransformer": ["BaseModelOutput", "False", "FlashAttentionKwargs", "ModelAttentionLayer", "ModelTransformer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "_", "__init__", "all_attentions", "append", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "encoder_layer", "encoder_states", "for", "forward", "gradient_checkpointing", "hidden_states", "if", "in", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "position_embeddings", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "pixtral/modeling_pixtral.py:PixtralPreTrainedModel": ["Conv2d", "Linear", "ModelAttentionLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionConfig", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "pixtral/modeling_pixtral.py:generate_block_attention_mask": ["Model_block_attention_mask", "None", "block_end_idx", "block_start_idx", "causal_mask", "cumsum", "d_min", "def", "device", "dtype", "end", "expand", "fill_value", "finfo", "for", "full", "in", "min", "patch_embeds_list", "return", "seq_len", "shape", "start", "tensor", "torch", "zip"], "pixtral/modeling_pixtral.py:PixtralVisionModel": ["BaseModelOutput", "Conv2d", "False", "FlashAttentionKwargs", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTransformer", "ModelVisionModel", "None", "Optional", "T", "Tensor", "True", "Union", "Unpack", "_", "__init__", "_attn_implementation", "args", "attention_mask", "auto_docstring", "base_model_prefix", "batch_size", "bias", "bool", "can_return_tuple", "cat", "class", "config", "def", "dim", "else", "embed", "eps", "flash_attention_2", "flatten", "for", "forward", "generate_block_attention_mask", "get_input_embeddings", "height", "hidden_size", "if", "image_size", "image_sizes", "in", "in_channels", "is", "kernel_size", "kwargs", "ln_pre", "max_width", "nn", "num_channels", "out_channels", "output_attentions", "output_hidden_states", "p", "patch_conv", "patch_embeds", "patch_embeds_list", "patch_positional_embedding", "patch_size", "pixel_values", "position_embeddings", "position_ids", "position_ids_in_meshgrid", "post_init", "return", "return_dict", "self", "shape", "size", "stride", "super", "torch", "transformer", "tuple", "unsqueeze", "vision_encoder", "width", "zip"], "vit_mae/modeling_vit_mae.py:ViTMAEModelOutput": ["FloatTensor", "LongTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "ids_restore", "last_hidden_state", "mask", "r", "torch", "tuple"], "vit_mae/modeling_vit_mae.py:ViTMAEDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "r", "torch", "tuple"], "vit_mae/modeling_vit_mae.py:ViTMAEForPreTrainingOutput": ["FloatTensor", "LongTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "ids_restore", "logits", "loss", "mask", "r", "torch", "tuple"], "vit_mae/modeling_vit_mae.py:get_2d_sincos_pos_embed": ["False", "Model_2d_sincos_pos_embed", "Model_2d_sincos_pos_embed_from_grid", "add_cls_token", "arange", "axis", "concatenate", "def", "dtype", "embed_dim", "float32", "grid", "grid_h", "grid_size", "grid_w", "if", "meshgrid", "np", "pos_embed", "reshape", "return", "stack", "zeros"], "vit_mae/modeling_vit_mae.py:get_2d_sincos_pos_embed_from_grid": ["Model_1d_sincos_pos_embed_from_grid", "Model_2d_sincos_pos_embed_from_grid", "ValueError", "axis", "be", "concatenate", "def", "emb", "emb_h", "emb_w", "embed_dim", "even", "grid", "if", "must", "np", "raise", "return"], "vit_mae/modeling_vit_mae.py:get_1d_sincos_pos_embed_from_grid": ["Model_1d_sincos_pos_embed_from_grid", "ValueError", "arange", "axis", "be", "concatenate", "cos", "d", "def", "dtype", "einsum", "emb", "emb_cos", "emb_sin", "embed_dim", "even", "float", "if", "m", "md", "must", "np", "omega", "out", "pos", "raise", "reshape", "return", "sin"], "vit_mae/modeling_vit_mae.py:ViTMAEEmbeddings": ["False", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Parameter", "Tensor", "True", "__init__", "add_cls_token", "align_corners", "and", "argsort", "batch_size", "bicubic", "bool", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "copy_", "data", "def", "device", "dim", "else", "embeddings", "expand", "float", "forward", "from_numpy", "functional", "gather", "get_2d_sincos_pos_embed", "height", "hidden_size", "ids_keep", "ids_restore", "ids_shuffle", "if", "index", "init", "initialize_weights", "initializer_range", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "len_keep", "mask", "mask_ratio", "mode", "new_height", "new_width", "nn", "noise", "normal_", "not", "num_channels", "num_patches", "num_positions", "ones", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "pos_embed", "position_embeddings", "projection", "rand", "random_masking", "repeat", "requires_grad", "reshape", "return", "self", "seq_length", "sequence", "sequence_unmasked", "shape", "size", "sqrt_num_positions", "std", "super", "to", "torch", "torch_int", "unsqueeze", "view", "w", "weight", "width", "xavier_uniform_", "zeros"], "vit_mae/modeling_vit_mae.py:ViTMAEPatchEmbeddings": ["Conv2d", "False", "Input", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "and", "batch_size", "bool", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "f", "flatten", "forward", "height", "hidden_size", "if", "image", "image_size", "in", "interpolate_pos_encoding", "isinstance", "kernel_size", "match", "model", "nn", "not", "num_channels", "num_patches", "of", "one", "or", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "transpose", "values", "width", "with", "x"], "vit_mae/modeling_vit_mae.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "vit_mae/modeling_vit_mae.py:ViTMAESelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "vit_mae/modeling_vit_mae.py:ViTMAESelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "vit_mae/modeling_vit_mae.py:ViTMAEAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "vit_mae/modeling_vit_mae.py:ViTMAEIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "vit_mae/modeling_vit_mae.py:ViTMAEOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "vit_mae/modeling_vit_mae.py:ViTMAELayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "vit_mae/modeling_vit_mae.py:ViTMAEEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "vit_mae/modeling_vit_mae.py:ViTMAEPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "ModelConfig", "ModelDecoder", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "decoder_pos_embed", "def", "elif", "fill_", "hidden_states", "if", "initialize_weights", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "vit", "weight", "zero_"], "vit_mae/modeling_vit_mae.py:ViTMAEModel": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "ids_restore", "if", "in", "interpolate_pos_encoding", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "mask", "nn", "noise", "num_hidden_layers", "patch_embeddings", "pixel_values", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch"], "vit_mae/modeling_vit_mae.py:ViTMAEDecoder": ["False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelLayer", "Module", "ModuleList", "None", "Parameter", "Tensor", "True", "_", "__init__", "add_cls_token", "align_corners", "bias", "bicubic", "bool", "cat", "class", "class_pos_embed", "config", "copy_", "data", "decoder_config", "decoder_embed", "decoder_hidden_size", "decoder_intermediate_size", "decoder_layers", "decoder_norm", "decoder_num_attention_heads", "decoder_num_hidden_layers", "decoder_pos_embed", "decoder_pred", "deepcopy", "def", "device", "dim", "else", "embeddings", "embeddings_positions", "eps", "float", "for", "forward", "from_numpy", "functional", "gather", "get_2d_sincos_pos_embed", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "ids_restore", "if", "in", "index", "init", "initialize_weights", "initializer_range", "int", "intermediate_size", "interpolate", "interpolate_pos_encoding", "layer_module", "layer_norm_eps", "logits", "mask_token", "mask_tokens", "mode", "nn", "normal_", "num_attention_heads", "num_channels", "num_hidden_layers", "num_patches", "patch_pos_embed", "patch_size", "permute", "range", "repeat", "requires_grad", "reshape", "return", "self", "shape", "size", "std", "super", "to", "torch", "unsqueeze", "view", "x", "x_", "zeros"], "vit_mae/modeling_vit_mae.py:ViTMAEForPreTraining": ["False", "FloatTensor", "Make", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "__init__", "_prune_heads", "a", "and", "attention", "attentions", "auto_docstring", "batch_size", "bool", "by", "can_return_tuple", "channels", "class", "config", "configuration", "decoder", "decoder_outputs", "def", "dim", "divisible", "does", "einsum", "else", "embeddings", "encoder", "equal", "f", "for", "forward", "forward_loss", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "ids_restore", "if", "image", "image_size", "in", "int", "interpolate_pos_encoding", "is", "items", "keepdim", "kwargs", "last_hidden_state", "latent", "layer", "logits", "loss", "mask", "match", "mean", "nchpwq", "nhwpqc", "noise", "norm_pix_loss", "not", "num_channels", "num_patches", "num_patches_h", "num_patches_w", "number", "of", "on", "one", "or", "original", "original_height", "original_image_size", "original_width", "outputs", "patch", "patch_embeddings", "patch_size", "patches", "patchified", "patchified_pixel_values", "patchify", "pixel", "pixel_values", "post_init", "pred", "prune_heads", "r", "raise", "reshape", "return", "self", "set", "shape", "size", "squared", "sum", "super", "sure", "target", "that", "the", "to", "torch", "tuple", "unpatchify", "values", "var", "vit"], "gemma3n/modeling_gemma3n.py:Gemma3nModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "audio_hidden_states", "class", "image_hidden_states", "r", "torch"], "gemma3n/modeling_gemma3n.py:Gemma3nCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "audio_hidden_states", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "gemma3n/modeling_gemma3n.py:Gemma3nRMSNorm": ["False", "ModelRMSNorm", "Module", "Parameter", "Tensor", "True", "__init__", "_norm", "bool", "class", "def", "dim", "else", "eps", "extra_repr", "f", "float", "forward", "if", "int", "keepdim", "mean", "nn", "ones", "output", "persistent", "pow", "register_buffer", "return", "self", "shape", "sqrt", "super", "tensor", "torch", "tuple", "type_as", "weight", "with_scale", "x"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioRelativePositionEmbedding": ["False", "Linear", "ModelAudioConfig", "ModelAudioRelativePositionEmbedding", "Module", "Tensor", "_", "__init__", "_get_timing_signal_1d_pos", "_relative_shift", "arange", "batch_size", "bias", "cat", "channels", "class", "conf_attention_context_left", "conf_attention_context_right", "conf_num_attention_heads", "config", "cos", "def", "device", "dim", "dtype", "exp", "float", "float32", "forward", "functional", "head_dim", "hidden_size", "int", "inv_timescales", "key_context_size", "keys", "keys_p_t", "log", "log_timescale_increment", "math", "matmul", "max", "max_backward", "max_forward", "max_span_plus_1", "max_timescale", "min_timescale", "nn", "num_heads", "num_query_blocks", "num_timescales", "pad", "pad_amount_last_dim", "padding_tuple", "permute", "persistent", "pos_indices", "pos_proj", "position", "projected_sin_emb", "q_permuted", "q_reshaped", "queries", "queries_p", "query_block_size", "register_buffer", "reshape", "return", "s_permuted", "scaled_time", "self", "shape", "sin", "sin_emb", "sin_emb_timing_signal", "squeeze", "super", "term_ac", "term_bd_before_shift", "term_bd_padded", "term_bd_reshaped", "term_bd_shifted", "term_bd_sliced", "term_bd_unshifed", "term_bd_unshifed_matmul", "timing_signal", "to", "torch", "type", "unsqueeze"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioAttention": ["BoolTensor", "False", "Linear", "ModelAudioAttention", "ModelAudioConfig", "ModelAudioRelativePositionEmbedding", "Module", "Parameter", "Shape", "T", "Tensor", "ValueError", "_", "__init__", "_convert_to_block", "_extract_block_context", "_pad_dim1", "after", "and", "attention_logits_soft_cap", "b", "b_dim", "batch", "batch_size", "bias", "bmm", "bool", "broadcast_shape", "c_dim", "cat", "chunk_size", "class", "clone", "condition_from_causality", "condition_from_input_validity", "conf_attention_chunk_size", "conf_attention_context_left", "conf_attention_context_right", "conf_attention_logit_cap", "conf_num_attention_heads", "config", "context_size", "context_vectors", "contiguous", "def", "destination", "detach", "device", "diagonal", "dim", "dimension", "dtype", "extracted_valid_mask_blocks", "f", "final_condition_for_where", "finfo", "float", "float32", "forward", "frame_len", "frame_step", "functional", "h_dim", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_blocks", "key_states", "left", "local_causal_valid_mask", "logical_and", "logits", "lower_causal_mask", "mask", "max", "max_future_horizon", "max_past_horizon", "min", "movedim", "n_dim", "ndim", "new_zeros", "nn", "not", "num_blocks", "num_heads", "num_query_blocks", "of", "ones", "original_valid_mask", "pad_left", "pad_right", "padding_len", "per_dim_scale", "per_dim_scale_sp", "per_dim_scale_sp_broadcast", "permute", "permute_dims", "persistent", "potential", "prob_bun", "probabilities", "q_proj", "q_scale", "q_time", "qkv_shape", "query_blocks", "query_states", "r_softplus_0", "raise", "register_buffer", "relative_position_embedding", "reshape", "result_bmm", "return", "right", "self", "shape", "size", "softcap", "softcap_val", "softmax", "softplus", "source", "step", "super", "t", "tail_shape", "tanh", "tensor", "to", "torch", "tril", "u_dim", "unfold", "unsqueeze", "upper_causal_mask", "v_bun", "v_proj", "value_blocks", "value_states", "view", "w_dim", "where", "x", "x_unfolded", "zeros"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioCumulativeGroupNorm": ["Input", "ModelAudioCumulativeGroupNorm", "Module", "Parameter", "Sequence", "Tensor", "True", "ValueError", "__init__", "calc_dtype", "clamp", "class", "cum_count_elements", "cum_mean", "cum_sum_sq_diff", "cum_sum_values", "cum_variance", "cumsum", "def", "dim", "does", "dtype", "elements_in_group_at_t", "eps", "expected", "expected_input_suffix", "f", "feature_dims", "final_output", "float", "float32", "forward", "hidden_states", "if", "input_dtype", "int", "keepdim", "len", "mask_calc", "match", "min", "nn", "normalized_x", "not", "num_channels", "ones", "ones_like", "pow", "raise", "range", "reduction_axes", "return", "rsqrt", "safe_cum_count_elements", "scale", "scale_view_shape", "self", "shape", "squared_diff_from_mean", "suffix", "sum", "sum_sq_diff_at_t", "sum_values_at_t", "super", "tensor", "to", "torch", "tuple", "view", "weight", "x_calc"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioSSCPConvBlock": ["Conv2d", "F", "False", "ModelAudioConfig", "ModelAudioCumulativeGroupNorm", "ModelAudioSSCPConvBlock", "Module", "ReLU", "Tensor", "__init__", "activation", "audio_encodings", "audio_encodings_conv", "audio_encodings_normed", "audio_encodings_padded", "bias", "class", "config", "constant", "contiguous", "conv", "def", "dtype", "else", "eps", "f_in_padded", "f_out_conv", "feature_dims", "forward", "idx", "if", "in_channels", "input_freq_dim", "int", "kernel_h", "kernel_size", "kernel_w", "manual_padding", "mode", "nn", "norm", "num_channels", "out_channels", "pad", "padding", "permute", "return", "self", "sscp_conv_channel_size", "sscp_conv_group_norm_eps", "sscp_conv_kernel_size", "sscp_conv_stride_size", "stride", "stride_h", "stride_w", "super", "to", "torch", "tuple", "value", "weight", "x_for_norm", "x_normed"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioSubSampleConvProjection": ["False", "Linear", "ModelAudioConfig", "ModelAudioSSCPConvBlock", "ModelAudioSubSampleConvProjection", "Module", "Tensor", "__init__", "append", "audio_encodings", "audio_encodings_reshaped", "b", "bias", "c_out", "calculated_block_padding", "calculated_f_out_dims", "class", "config", "contiguous", "conv_0", "conv_1", "current_f_for_block_input", "def", "f_in_padded", "f_out", "f_out_after_conv", "final_c_out", "final_f_out", "for", "forward", "hidden_size", "i", "idx", "in", "input_feat_size", "input_freq_dim", "input_proj_in_features", "input_proj_linear", "kernel_h", "kernel_w", "manual_padding", "manual_padding_tuple", "nn", "output", "output_flattened", "pad_f_left", "pad_f_right", "pad_t_bottom", "pad_t_top", "permute", "range", "return", "self", "shape", "sscp_conv_channel_size", "sscp_conv_kernel_size", "sscp_conv_stride_size", "stride_h", "stride_w", "super", "t_out", "torch", "unsqueeze", "view", "x", "x_permuted"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioConformerAttention": ["BoolTensor", "False", "Linear", "ModelAudioAttention", "ModelAudioConfig", "ModelAudioConformerAttention", "ModelRMSNorm", "Module", "Tensor", "__init__", "attn", "audio_encodings", "audio_encodings_attn_out", "audio_encodings_input_to_attn", "audio_encodings_norm", "audio_encodings_reshaped", "audio_mel_mask", "b", "bias", "clamp", "class", "config", "def", "forward", "gradient_clipping", "head_dim", "hidden_size", "nn", "num_heads", "persistent", "post", "post_in_features", "post_norm", "pre_attn_norm", "register_buffer", "reshape", "return", "self", "shape", "super", "t", "tensor", "torch"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioConformerFeedForward": ["False", "Linear", "ModelAudioConfig", "ModelAudioConformerFeedForward", "ModelRMSNorm", "Module", "Tensor", "__init__", "audio_encodings", "bias", "clamp", "class", "conf_residual_weight", "config", "def", "ffw_layer_1", "ffw_layer_2", "forward", "functional", "gradient_clipping", "hidden_size", "nn", "persistent", "post_layer_norm", "post_layer_scale", "pre_layer_norm", "register_buffer", "residual", "return", "self", "silu", "super", "tensor", "torch"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioConformerLightConv1d": ["Conv1d", "F", "False", "Linear", "ModelAudioConfig", "ModelAudioConformerLightConv1d", "ModelRMSNorm", "Module", "Tensor", "__init__", "audio_encodings", "audio_encodings_permuted", "audio_encodings_permuted_padded", "audio_encodings_residual", "bias", "causal_padding", "clamp", "class", "conf_conv_kernel_size", "config", "conv_norm", "def", "depthwise_conv1d", "dim", "eps", "forward", "functional", "glu", "gradient_clipping", "groups", "hidden_size", "in_channels", "kernel_size", "linear_end", "linear_start", "nn", "out_channels", "output", "pad", "padding", "permute", "persistent", "pre_layer_norm", "register_buffer", "return", "rms_norm_eps", "self", "silu", "stride", "super", "tensor", "torch"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioConformerBlock": ["BoolTensor", "False", "ModelAudioConfig", "ModelAudioConformerAttention", "ModelAudioConformerBlock", "ModelAudioConformerFeedForward", "ModelAudioConformerLightConv1d", "ModelRMSNorm", "Module", "Tensor", "__init__", "attention", "audio_encodings", "audio_encodings_for_lconv_input", "audio_mel_mask", "clamp", "class", "config", "def", "dtype", "ffw_layer_end", "ffw_layer_start", "forward", "gradient_clipping", "hidden_size", "lconv1d", "nn", "norm", "output", "persistent", "register_buffer", "return", "self", "super", "tensor", "to", "torch", "unsqueeze", "validity_mask_for_lconv"], "gemma3n/modeling_gemma3n.py:Gemma3nAudioEncoder": ["BoolTensor", "ModelAudioConfig", "ModelAudioConformerBlock", "ModelAudioEncoder", "ModelAudioSubSampleConvProjection", "ModuleList", "PreTrainedModel", "Tensor", "_", "__init__", "and", "arange", "audio_encodings", "audio_mel", "audio_mel_mask", "block", "clamp", "class", "conf_num_hidden_layers", "conf_reduction_factor", "config", "conformer", "current_mask", "def", "device", "elif", "expand", "for", "forward", "gather", "if", "in", "indices", "len", "main_input_name", "masked_fill", "max", "ndim", "nn", "range", "return", "self", "shape", "sscp_conv_stride_size", "stride_pair_idx", "subsample_conv_projection", "super", "t_sub", "time_stride_product", "torch", "tuple", "unsqueeze"], "gemma3n/modeling_gemma3n.py:Gemma3nTextScaledWordEmbedding": ["Embedding", "False", "ModelTextScaledWordEmbedding", "Tensor", "__init__", "class", "def", "dtype", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "persistent", "register_buffer", "return", "self", "super", "tensor", "to", "torch", "weight"], "gemma3n/modeling_gemma3n.py:Gemma3nTextLaurelBlock": ["False", "Linear", "ModelRMSNorm", "ModelTextConfig", "ModelTextLaurelBlock", "Module", "Tensor", "__init__", "bias", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "laurel_hidden_states", "laurel_rank", "linear_left", "linear_right", "nn", "normed_laurel_hidden_states", "post_laurel_norm", "return", "rms_norm_eps", "self", "super", "torch"], "gemma3n/modeling_gemma3n.py:Gemma3nTextMLP": ["ACT2FN", "False", "Linear", "ModelTextConfig", "ModelTextMLP", "Module", "Normal", "Tensor", "True", "__init__", "_gaussian_topk", "act_fn", "activation_sparsity", "activation_sparsity_pattern", "activations", "bias", "class", "config", "cutoff_x", "def", "device", "dim", "distributions", "down_proj", "dtype", "float32", "forward", "functional", "gate_proj", "hidden_activation", "hidden_size", "hidden_states", "icdf", "if", "inputs", "inputs_mean", "inputs_std", "int", "intermediate_size", "keepdim", "layer_idx", "mean", "nn", "normal", "normal_dist", "relu", "return", "self", "std", "std_multiplier", "super", "target_sparsity_tensor", "tensor", "torch", "type", "unbiased", "up_proj"], "gemma3n/modeling_gemma3n.py:Gemma3nTextAltUp": ["False", "Linear", "ModelRMSNorm", "ModelTextAltUp", "ModelTextConfig", "Module", "None", "Parameter", "Tensor", "__init__", "activated", "all_coefs", "altup_active_idx", "altup_coef_clip", "altup_num_inputs", "and", "bias", "clamp_", "class", "compute_router_modalities", "config", "contiguous", "correct", "correct_output_scale", "corrected", "correction_coefs", "data", "def", "eps", "float", "forward", "hidden_size", "hidden_states", "if", "innovation", "is", "matmul", "modalities", "modality_router", "mul", "nn", "not", "permute", "persistent", "predict", "prediction_coefs", "predictions", "register_buffer", "repeat", "reshape", "return", "rms_norm_eps", "routed", "router_input_scale", "router_inputs", "router_norm", "scale_corrected_output", "self", "shape", "super", "tanh", "tensor", "torch", "training", "type_as", "unsqueeze", "weight", "x", "zeros"], "gemma3n/modeling_gemma3n.py:Gemma3nTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gemma3n/modeling_gemma3n.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "gemma3n/modeling_gemma3n.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "gemma3n/modeling_gemma3n.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_dim", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softcap", "softmax", "tanh", "to", "torch", "training", "transpose", "tuple", "value", "value_states"], "gemma3n/modeling_gemma3n.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "Optional", "Tensor", "cos", "def", "int", "position_ids", "return", "rotate_half", "sin", "torch", "unsqueeze", "unsqueeze_dim", "x"], "gemma3n/modeling_gemma3n.py:Gemma3nTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelRMSNorm", "ModelTextAttention", "ModelTextConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "device", "dim", "dropout", "eager", "eager_attention_forward", "else", "eps", "first_kv_shared_layer_idx", "forward", "getattr", "hasattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "index", "input_shape", "int", "is", "is_causal", "is_kv_shared_layer", "is_sliding", "k_norm", "k_proj", "key_states", "kv_shared_layer_index", "kwargs", "layer_idx", "layer_types", "len", "new_name", "nn", "not", "num_attention_heads", "num_hidden_layers", "num_key_value_groups", "num_key_value_heads", "num_kv_shared_layers", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "prev_layers", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "shared_layers", "sin", "sliding_attention", "sliding_window", "store_full_length_kv", "super", "to", "torch", "training", "transpose", "tuple", "unsqueeze_dim", "update", "v_norm", "v_proj", "value_states", "version", "view", "with_scale"], "gemma3n/modeling_gemma3n.py:Gemma3nTextDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "Linear", "LongTensor", "ModelRMSNorm", "ModelTextAltUp", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextLaurelBlock", "ModelTextMLP", "None", "Optional", "Tensor", "__init__", "act_fn", "active_prediction", "active_prediction_normed", "altup", "altup_active_idx", "altup_correct_scale", "attention_mask", "attention_type", "attn", "attn_ffw", "attn_ffw_laurel_gated", "attn_ffw_norm", "attn_gated", "attn_laurel", "attn_norm", "bias", "bool", "cache_position", "class", "clone", "config", "correct", "corrected_predictions", "def", "deprecate_kwarg", "else", "eps", "first_prediction", "forward", "hidden_activation", "hidden_size", "hidden_size_per_layer_input", "hidden_states", "if", "input_layernorm", "int", "is_sliding", "kwargs", "laurel", "laurel_output", "layer_idx", "layer_types", "math", "mlp", "multiply", "new_name", "nn", "output_attentions", "outputs", "past_key_value", "past_key_values", "per_layer_input", "per_layer_input_gate", "per_layer_projection", "position_embeddings", "position_embeddings_global", "position_embeddings_local", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "post_per_layer_input_norm", "pre_feedforward_layernorm", "predict", "predictions", "return", "rms_norm_eps", "scale_corrected_output", "self", "self_attn", "self_attn_weights", "sqrt", "super", "torch", "tuple", "use_cache", "version"], "gemma3n/modeling_gemma3n.py:Gemma3nPreTrainedModel": ["ModelAudioAttention", "ModelAudioCumulativeGroupNorm", "ModelConfig", "ModelPreTrainedModel", "ModelTextAltUp", "ModelTextAttention", "ModelTextDecoderLayer", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "correct_output_scale", "data", "def", "elif", "fill_", "hidden_states", "if", "isinstance", "module", "past_key_values", "per_dim_scale", "self", "super", "supports_gradient_checkpointing", "weight", "zero_"], "gemma3n/modeling_gemma3n.py:Gemma3nTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "False", "FloatTensor", "Linear", "LongTensor", "ModelPreTrainedModel", "ModelRMSNorm", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextRotaryEmbedding", "ModelTextScaledWordEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "_", "__init__", "all_hidden_states", "all_self_attns", "altup_num_inputs", "altup_proj", "altup_projections", "altup_unemb_proj", "altup_unembed_projections", "and", "append", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "causal_mask", "causal_mask_mapping", "checkpointing", "class", "config", "copy", "create_causal_mask", "create_sliding_window_causal_mask", "current_hidden_state", "decoder_layer", "deepcopy", "def", "default", "device", "dict", "dim", "dtype", "else", "embed_scale", "embed_tokens", "embed_tokens_per_layer", "eps", "epsilon_tensor", "exactly", "for", "forward", "full_attention", "get_per_layer_inputs", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_size_per_layer_input", "hidden_states", "hidden_states_0", "i", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "keepdim", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "logger", "mask_kwargs", "maximum", "mean", "must", "new_magnitude", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "per_layer_input", "per_layer_input_scale", "per_layer_inputs", "per_layer_model_projection", "per_layer_projection", "per_layer_projection_norm", "per_layer_projection_scale", "persistent", "position_embeddings_global", "position_embeddings_local", "position_ids", "post_init", "project_per_layer_inputs", "r", "raise", "range", "register_buffer", "reshape", "return", "rms_norm_eps", "rope_local_base_freq", "rope_scaling", "rope_theta", "rope_type", "rotary_emb", "rotary_emb_local", "rsqrt", "self", "shape", "sliding_attention", "specify", "sqrt", "stack", "super", "target_magnitude", "temp_hidden_states", "tensor", "to", "torch", "training", "unsqueeze", "use_cache", "vocab_size", "vocab_size_per_layer_input", "warning_once", "with"], "gemma3n/modeling_gemma3n.py:Gemma3nForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "None", "Optional", "Tensor", "Union", "__init__", "_checkpoint_conversion_mapping", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "final_logit_softcapping", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "tanh", "torch", "use_cache", "vocab_size", "weight"], "gemma3n/modeling_gemma3n.py:Gemma3nMultimodalEmbedder": ["Embedding", "False", "Linear", "LongTensor", "ModelAudioConfig", "ModelMultimodalEmbedder", "ModelRMSNorm", "ModelTextConfig", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "bias", "class", "def", "else", "emb_norm", "emb_norm_proj", "embedding", "embedding_post_projection_norm", "embedding_projection", "eps", "exactly", "forward", "hard_emb", "hard_embedding_norm", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "multimodal_config", "multimodal_hidden_size", "must", "nn", "not", "of", "one", "or", "raise", "return", "rms_norm_eps", "self", "soft_embedding_norm", "specify", "super", "text_config", "text_hidden_size", "torch", "vocab_offset", "vocab_size", "with_scale"], "gemma3n/modeling_gemma3n.py:Gemma3nModel": ["Audio", "AutoModel", "Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultimodalEmbedder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "ValueError", "You", "_", "__init__", "_checkpoint_conversion_mapping", "accepts_loss_kwargs", "all", "and", "attention_mask", "attentions", "audio_batch_size", "audio_config", "audio_embed_dim", "audio_embeds", "audio_features", "audio_hidden_states", "audio_input_ids", "audio_mask", "audio_outputs", "audio_padding_embs", "audio_padding_toks", "audio_seq_len", "audio_soft_tokens_per_image", "audio_token_id", "audio_tower", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "device", "dim", "do", "do_pooling", "dtype", "dummy_audio_token_id", "dummy_vision_token_id", "else", "embed_audio", "embed_vision", "exactly", "expand", "expand_as", "expanded_audio_mask", "expanded_vision_mask", "extra_padding_features", "extra_padding_tokens", "f", "features", "forward", "from_config", "get_audio_features", "get_decoder", "get_image_features", "get_input_embeddings", "get_per_layer_inputs", "get_placeholder_mask", "hidden_size", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_token_id", "input_features", "input_features_mask", "input_ids", "inputs_embeds", "is", "labels", "language_model", "last_hidden_state", "lm_kwargs", "logical_and", "long", "masked_scatter", "match", "must", "n_audio_tokens", "n_image_tokens", "not", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "per_layer_inputs", "per_layer_inputs_mask", "per_layer_inputs_tokens", "permute", "pixel_values", "position_ids", "post_init", "r", "raise", "reshape", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "special_audio_mask", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "token_type_ids", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "value", "vision_config", "vision_embeds", "vision_input_ids", "vision_mask", "vision_outputs", "vision_soft_tokens_per_image", "vision_tower", "vocab_offset", "vocab_size", "vocab_size_per_layer_input", "where", "zeros_like"], "gemma3n/modeling_gemma3n.py:Gemma3nForConditionalGeneration": ["AttributeError", "Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "Use", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "audio_hidden_states", "audio_tower", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "contiguous", "decoder", "def", "device", "else", "embed_vision", "final_logit_softcapping", "flat_labels", "flat_logits", "float", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_text_config", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_features", "input_features_mask", "input_ids", "inputs_embeds", "instead", "int", "is", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "lm_head", "lm_kwargs", "logits", "logits_to_keep", "loss", "loss_fct", "model", "model_inputs", "multi_modal_projector", "nn", "not", "of", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "raise", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "shift_attention_mask", "shift_labels", "shift_logits", "slice", "slice_indices", "super", "tanh", "text_config", "to", "token_type_ids", "torch", "use_cache", "value", "view", "vision_tower", "vocab_size", "weight"], "persimmon/modeling_persimmon.py:PersimmonRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "persimmon/modeling_persimmon.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "persimmon/modeling_persimmon.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "persimmon/modeling_persimmon.py:PersimmonMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act", "class", "config", "def", "dense_4h_to_h", "dense_h_to_4h", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super"], "persimmon/modeling_persimmon.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value"], "persimmon/modeling_persimmon.py:PersimmonAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "False", "FlashAttentionKwargs", "Instantiating", "LayerNorm", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "_split_heads", "a", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "cat", "class", "config", "cos", "creating", "def", "dense", "deprecate_kwarg", "dim", "divisible", "dropout", "during", "eager", "eager_attention_forward", "elementwise_affine", "else", "eps", "errors", "f", "forward", "fused_qkv", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_layernorm", "key_pass", "key_rot", "key_states", "kwargs", "layer_idx", "layer_norm_eps", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "partial_rotary_factor", "partial_rotation_size", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_layernorm", "q_len", "qk_layernorm", "query_key_value", "query_pass", "query_rot", "query_states", "raise", "recommended", "reshape", "return", "rope_theta", "rotary_emb", "rotary_ndims", "scaling", "self", "seq_length", "shape", "sin", "size", "super", "sure", "the", "this", "three_times_hidden_size", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "value_states", "version", "view", "warning_once", "when", "will", "without"], "persimmon/modeling_persimmon.py:PersimmonDecoderLayer": ["Cache", "Dropout", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "eps", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_norm_eps", "mlp", "new_name", "nn", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "persimmon/modeling_persimmon.py:PersimmonPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "persimmon/modeling_persimmon.py:PersimmonModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "final_layernorm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_eps", "layer_outputs", "layers", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "nn", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "rotary_emb", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "type", "unsqueeze", "use_cache", "using_compilable_cache", "vocab_size", "warning_once", "with", "xpu"], "persimmon/modeling_persimmon.py:PersimmonForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "persimmon/modeling_persimmon.py:PersimmonForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "persimmon/modeling_persimmon.py:PersimmonForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class"], "xlm/modeling_xlm.py:create_sinusoidal_embeddings": ["False", "FloatTensor", "Model_sinusoidal_embeddings", "array", "cos", "def", "detach_", "dim", "for", "in", "j", "n_pos", "np", "out", "pos", "position_enc", "power", "range", "requires_grad", "sin", "torch"], "xlm/modeling_xlm.py:get_masks": ["False", "Model_masks", "None", "alen", "arange", "assert", "attn_mask", "bs", "causal", "def", "device", "dtype", "else", "if", "is", "item", "lengths", "long", "mask", "max", "not", "or", "padding_mask", "repeat", "return", "size", "slen", "torch"], "xlm/modeling_xlm.py:XLMSquadHeadOutput": ["FloatTensor", "LongTensor", "Model", "ModelOutput", "None", "Optional", "class", "cls_logits", "end_top_index", "end_top_log_probs", "loss", "r", "start_top_index", "start_top_log_probs", "torch"], "xlm/modeling_xlm.py:XLMPoolerStartLogits": ["FloatTensor", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "__init__", "class", "config", "def", "dense", "dtype", "else", "float16", "forward", "hidden_size", "hidden_states", "if", "is", "nn", "not", "p_mask", "return", "self", "squeeze", "super", "torch", "x"], "xlm/modeling_xlm.py:XLMPoolerEndLogits": ["FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "One", "Optional", "Tanh", "__init__", "activation", "assert", "be", "cat", "class", "config", "def", "dense_0", "dense_1", "dim", "dtype", "else", "eps", "expand", "float16", "forward", "gather", "hidden_size", "hidden_states", "hsz", "if", "is", "layer_norm_eps", "nn", "not", "of", "or", "p_mask", "return", "self", "shape", "should", "slen", "squeeze", "start_positions", "start_states", "super", "torch", "x"], "xlm/modeling_xlm.py:XLMPoolerAnswerClass": ["False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "One", "Optional", "Tanh", "__init__", "activation", "assert", "be", "bias", "cat", "class", "cls_index", "cls_token_state", "config", "def", "dense_0", "dense_1", "dim", "else", "expand", "forward", "gather", "hidden_size", "hidden_states", "hsz", "if", "is", "nn", "not", "of", "or", "return", "self", "shape", "should", "squeeze", "start_positions", "start_states", "super", "torch", "x"], "xlm/modeling_xlm.py:XLMSQuADHead": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelOutput", "ModelPoolerAnswerClass", "ModelPoolerEndLogits", "ModelPoolerStartLogits", "Module", "None", "Optional", "Union", "__init__", "and", "answer_class", "auto_docstring", "bh", "bl", "blh", "bool", "bsz", "class", "cls_index", "cls_logits", "cls_loss", "config", "def", "dim", "einsum", "else", "end_log_probs", "end_logits", "end_loss", "end_n_top", "end_positions", "end_top_index", "end_top_log_probs", "expand", "expand_as", "for", "forward", "functional", "gather", "hidden_states", "hidden_states_expanded", "hsz", "if", "in", "is", "is_impossible", "loss", "loss_fct", "loss_fct_cls", "nn", "not", "p_mask", "r", "return", "return_dict", "self", "size", "slen", "softmax", "squeeze_", "start_log_probs", "start_logits", "start_loss", "start_n_top", "start_positions", "start_states", "start_top_index", "start_top_index_exp", "start_top_log_probs", "super", "topk", "torch", "total_loss", "tuple", "unsqueeze", "view", "x"], "xlm/modeling_xlm.py:XLMSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "xlm/modeling_xlm.py:MultiHeadAttention": ["EncoderDecoderCache", "False", "Linear", "ModelHeadAttention", "Module", "None", "True", "__init__", "and", "assert", "attention_dropout", "attention_head_size", "bs", "cache", "cache_position", "class", "config", "context", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "dim", "dropout", "dtype", "else", "expand_as", "find_pruneable_heads_and_indices", "finfo", "float", "forward", "functional", "get", "head_dim", "head_mask", "heads", "if", "index", "input", "int", "is", "is_cross_attention", "is_updated", "isinstance", "k", "k_lin", "key_cache", "kv", "layer_id", "layer_idx", "len", "mask", "mask_reshape", "masked_fill_", "math", "matmul", "min", "n_heads", "nn", "not", "out_lin", "output_attentions", "outputs", "p", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "q_lin", "qlen", "return", "scores", "self", "self_attention_cache", "set", "size", "softmax", "sqrt", "super", "torch", "training", "transpose", "type_as", "union", "update", "v", "v_lin", "value_cache", "view", "weights"], "xlm/modeling_xlm.py:TransformerFFN": ["Linear", "ModelFFN", "Module", "__init__", "act", "apply_chunking_to_forward", "chunk_size_feed_forward", "class", "config", "def", "dim_hidden", "dropout", "else", "ff_chunk", "forward", "functional", "gelu", "gelu_activation", "if", "in_dim", "input", "lin1", "lin2", "nn", "out_dim", "p", "relu", "return", "self", "seq_len_dim", "super", "training", "x"], "xlm/modeling_xlm.py:XLMPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "None", "PreTrainedModel", "__init__", "_init_weights", "and", "attention_mask", "attns_list", "base_model_prefix", "bias", "class", "config", "constant_", "create_sinusoidal_embeddings", "data", "def", "dummy_inputs", "else", "emb_dim", "embed_init_std", "fill_", "if", "init", "init_std", "input_ids", "inputs", "inputs_list", "is", "isinstance", "kwargs", "langs", "langs_list", "max_position_embeddings", "mean", "module", "n_langs", "nn", "normal_", "not", "out", "padding_idx", "position_embeddings", "property", "return", "self", "sinusoidal_embeddings", "std", "super", "tensor", "torch", "transformer", "use_lang_emb", "weight", "zero_"], "xlm/modeling_xlm.py:XLMForQuestionAnsweringOutput": ["FloatTensor", "LongTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cls_logits", "end_top_index", "end_top_log_probs", "hidden_states", "loss", "r", "start_top_index", "start_top_log_probs", "torch", "tuple"], "xlm/modeling_xlm.py:XLMModel": ["BaseModelOutput", "Currently", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "Model", "ModelPreTrainedModel", "ModuleList", "MultiHeadAttention", "None", "NotImplementedError", "Optional", "Tensor", "TransformerFFN", "Union", "__init__", "_prune_heads", "_slen", "a", "an", "and", "append", "arange", "as", "assert", "attention_dropout", "attention_mask", "attentions", "attn", "attn_mask", "attn_outputs", "auto_docstring", "be", "bool", "bs", "cache", "cache_position", "can", "causal", "class", "config", "copy", "def", "device", "dict", "dim", "dropout", "dtype", "else", "emb_dim", "embeddings", "encoder", "eos_index", "eps", "expand", "expand_as", "ffns", "for", "forward", "from_legacy_cache", "functional", "get_head_mask", "get_input_embeddings", "get_masks", "get_seq_length", "hasattr", "head_mask", "heads", "heads_to_prune", "hidden_dim", "hidden_states", "i", "if", "in", "input_ids", "inputs_embeds", "int", "is", "is_decoder", "is_encoder", "isinstance", "item", "items", "kwargs", "lang_embeddings", "langs", "last_hidden_state", "layer", "layer_idx", "layer_norm1", "layer_norm2", "layer_norm_emb", "layer_norm_eps", "lengths", "list", "long", "map", "mask", "max", "max_position_embeddings", "multiple", "must", "n_heads", "n_langs", "n_layers", "n_words", "new_embeddings", "nn", "not", "of", "only", "output_attentions", "output_hidden_states", "p", "pad_index", "padding_idx", "padding_mask", "persistent", "position_embeddings", "position_ids", "post_init", "prune_heads", "pruned_heads", "r", "raise", "range", "register_buffer", "return", "return_dict", "self", "set_input_embeddings", "size", "slen", "str", "sum", "super", "tensor", "to", "token_type_ids", "torch", "training", "transformer", "tuple", "unsqueeze", "use_lang_emb", "use_return_dict", "used", "v"], "xlm/modeling_xlm.py:XLMPredLayer": ["AdaptiveLogSoftmaxWithLoss", "False", "Linear", "Model", "Module", "None", "True", "_", "__init__", "asm", "asm_cutoffs", "asm_div_value", "bias", "class", "config", "cross_entropy", "cutoffs", "def", "dim", "div_value", "else", "emb_dim", "forward", "functional", "head_bias", "if", "in_features", "is", "log_prob", "loss", "mean", "n_classes", "n_words", "nn", "not", "outputs", "pad_index", "proj", "reduction", "return", "scores", "self", "super", "view", "x", "y"], "xlm/modeling_xlm.py:XLMWithLMHeadModel": ["GenerationMixin", "MaskedLMOutput", "Model", "ModelModel", "ModelPreTrainedModel", "ModelPredLayer", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "cache_position", "cat", "class", "config", "def", "device", "dict", "dim", "dtype", "effective_batch_size", "else", "for", "forward", "full", "full_like", "get_output_embeddings", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "lang_id", "langs", "lengths", "logits", "long", "loss", "mask_token", "mask_token_id", "model_inputs", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pop", "position_ids", "post_init", "pred_layer", "prepare_inputs_for_generation", "proj", "r", "return", "return_dict", "self", "set_output_embeddings", "shape", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict", "value", "weight"], "xlm/modeling_xlm.py:XLMForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "class", "config", "def", "dict", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "langs", "lengths", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_summary", "single_label_classification", "squeeze", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict", "view"], "xlm/modeling_xlm.py:XLMForQuestionAnsweringSimple": ["CrossEntropyLoss", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "clamp", "class", "config", "contiguous", "def", "dict", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "langs", "len", "lengths", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "str", "super", "token_type_ids", "torch", "total_loss", "transformer", "transformer_outputs", "tuple", "use_return_dict"], "xlm/modeling_xlm.py:XLMForQuestionAnswering": ["Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelSQuADHead", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "class", "cls_index", "cls_logits", "config", "def", "dict", "else", "end_positions", "end_top_index", "end_top_log_probs", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_impossible", "langs", "lengths", "loss", "not", "output", "output_attentions", "output_hidden_states", "outputs", "p_mask", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "start_positions", "start_top_index", "start_top_log_probs", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict"], "xlm/modeling_xlm.py:XLMForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "class", "classifier", "config", "def", "dict", "dropout", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "langs", "lengths", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "str", "super", "token_type_ids", "torch", "transformer", "tuple", "use_return_dict", "view"], "xlm/modeling_xlm.py:XLMForMultipleChoice": ["CrossEntropyLoss", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "MultipleChoiceModelOutput", "None", "Optional", "Please", "Tensor", "The", "Union", "__init__", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache", "cannot", "choice", "class", "config", "def", "dict", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs", "inputs_embeds", "instead", "is", "kwargs", "labels", "langs", "lengths", "logger", "logits", "logits_proj", "loss", "loss_fct", "mask", "models", "multiple", "nn", "not", "num_choices", "num_labels", "output", "output_attentions", "output_hidden_states", "parameter", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_summary", "shape", "size", "str", "super", "the", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use", "use_return_dict", "used", "view", "warning", "with"], "xmod/modeling_xmod.py:XmodEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "xmod/modeling_xmod.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "xmod/modeling_xmod.py:XmodSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "xmod/modeling_xmod.py:XmodCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "xmod/modeling_xmod.py:XmodSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "xmod/modeling_xmod.py:XmodAttention": ["False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "pre_norm", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "residual", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "xmod/modeling_xmod.py:XmodIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "xmod/modeling_xmod.py:XmodAdapter": ["ACT2FN", "Linear", "ModelAdapter", "Module", "Tensor", "__init__", "adapter_act_fn", "adapter_reduction_factor", "bottleneck_size", "class", "config", "def", "dense1", "dense2", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "nn", "return", "self", "str", "super", "torch"], "xmod/modeling_xmod.py:XmodOutput": ["Dropout", "LayerNorm", "Linear", "ModelAdapter", "ModelOutput", "Module", "ModuleDict", "None", "Tensor", "True", "__init__", "adapter_layer_norm", "adapter_modules", "adapter_reuse_layer_norm", "append", "cat", "class", "config", "def", "dense", "dropout", "elif", "else", "enumerate", "eps", "for", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "i", "if", "in", "input_tensor", "int", "intermediate_size", "is", "item", "keys", "lang", "lang_adapter", "lang_id", "lang_ids", "lang_lengths", "lang_wise_outputs", "language", "languages", "layer_norm_eps", "list", "ln_before_adapter", "nn", "not", "residual", "return", "return_counts", "self", "split", "split_hidden_state", "split_hidden_states", "str", "super", "tolist", "torch", "unique_consecutive", "zip"], "xmod/modeling_xmod.py:XmodLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "If", "LayerNorm", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "lang_ids", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "pre_norm", "raise", "residual", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "xmod/modeling_xmod.py:XmodEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eps", "for", "forward", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "is", "is_pre_norm", "kwargs", "lang_ids", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_eps", "nn", "not", "num_hidden_layers", "past_key_values", "pre_norm", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "xmod/modeling_xmod.py:XmodPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "xmod/modeling_xmod.py:XmodPreTrainedModel": ["Embedding", "False", "Freezing", "LayerNorm", "Linear", "ModelConfig", "ModelCrossAttention", "ModelEmbeddings", "ModelLMHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "Supported", "True", "ValueError", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "adapter", "adapter_layer_norm", "adapter_modules", "adapters", "an", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "default_language", "does", "elif", "embeddings", "encoder", "f", "fill_", "for", "freeze_embeddings_and_language_adapters", "have", "hidden_states", "if", "in", "info", "initializer_range", "is", "isinstance", "language", "languages", "layer", "list", "logger", "mean", "module", "nn", "no_split_modules", "normal_", "not", "output", "padding_idx", "parameter", "parameters", "raise", "requires_grad", "roberta", "self", "set_default_language", "std", "str", "supports_gradient_checkpointing", "weight", "zero_"], "xmod/modeling_xmod.py:XmodModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "Input", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Please", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "adapter_languages", "adapter_modules", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "batch_size", "be", "bool", "cache_position", "call", "check_model_inputs", "class", "config", "create_causal_mask", "def", "default_lang_id", "default_language", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "index", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "keys", "kwargs", "lang_ids", "language", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "ones", "or", "output", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_default_language", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "unknown", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "xmod/modeling_xmod.py:XmodForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "ModelForCausalLM", "ModelLMHead", "ModelLMHeadModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lang_ids", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "xmod/modeling_xmod.py:XmodForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "ModelForMaskedLM", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lang_ids", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "xmod/modeling_xmod.py:XmodLMHead": ["LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "xmod/modeling_xmod.py:XmodForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "lang_ids", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "roberta", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "xmod/modeling_xmod.py:XmodForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_lang_ids", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lang_ids", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "repeat", "reshaped_logits", "return", "return_dict", "roberta", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "xmod/modeling_xmod.py:XmodForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lang_ids", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "roberta", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "view"], "xmod/modeling_xmod.py:XmodClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "xmod/modeling_xmod.py:XmodForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "lang_ids", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "roberta", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "roberta/modeling_roberta.py:RobertaEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "roberta/modeling_roberta.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "roberta/modeling_roberta.py:RobertaSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "roberta/modeling_roberta.py:RobertaCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "roberta/modeling_roberta.py:RobertaSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "roberta/modeling_roberta.py:RobertaAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "roberta/modeling_roberta.py:RobertaIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "roberta/modeling_roberta.py:RobertaOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "roberta/modeling_roberta.py:RobertaLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "roberta/modeling_roberta.py:RobertaPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLMHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "roberta/modeling_roberta.py:RobertaEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "roberta/modeling_roberta.py:RobertaPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "roberta/modeling_roberta.py:RobertaModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "roberta/modeling_roberta.py:RobertaForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "Model", "ModelForCausalLM", "ModelLMHead", "ModelLMHeadModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "device", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "roberta/modeling_roberta.py:RobertaForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "device", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "roberta/modeling_roberta.py:RobertaLMHead": ["LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "roberta/modeling_roberta.py:RobertaForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "view"], "roberta/modeling_roberta.py:RobertaForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "to", "token_type_ids", "torch", "tuple", "view"], "roberta/modeling_roberta.py:RobertaForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "view"], "roberta/modeling_roberta.py:RobertaClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "roberta/modeling_roberta.py:RobertaForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "csm/modeling_csm.py:CsmOutputWithPast": ["Cache", "FloatTensor", "ModelOutput", "ModelOutputWithPast", "None", "Optional", "attentions", "backbone_loss", "class", "depth_decoder_attentions", "depth_decoder_hidden_states", "depth_decoder_logits", "depth_decoder_loss", "depth_decoder_past_key_values", "hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "csm/modeling_csm.py:CsmRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "csm/modeling_csm.py:CsmRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "csm/modeling_csm.py:CsmMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "csm/modeling_csm.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "csm/modeling_csm.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "csm/modeling_csm.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "csm/modeling_csm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "csm/modeling_csm.py:CsmAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "csm/modeling_csm.py:CsmDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "csm/modeling_csm.py:CsmPreTrainedModel": ["ModelAttention", "ModelCodebooksHead", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "for", "hidden_states", "i", "if", "in", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "num_codebooks", "past_key_values", "range", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "csm/modeling_csm.py:CsmDepthDecoderModel": ["BaseModelOutputWithPast", "Cache", "Custom", "DynamicCache", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelDecoderLayer", "ModelDepthDecoderConfig", "ModelDepthDecoderModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "When", "You", "__init__", "across", "also", "and", "arange", "as", "attention_mask", "auto_docstring", "automatically", "backbone_hidden_size", "backbone_last_hidden_state", "batch", "be", "bias", "bool", "but", "cache_position", "causal_mask", "check_model_inputs", "clamp", "class", "codebook", "codebook_idxs", "compiler", "config", "correct", "create_causal_mask", "decoder", "decoder_layer", "def", "depth", "determines", "device", "else", "embed_tokens", "eps", "exactly", "first", "for", "forward", "from", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "identical", "if", "ignored", "in", "inference", "input_embeds", "input_ids", "input_ids_are_first_codebook", "inputs_embeds", "inputs_embeds_projector", "inputs_seq_length", "is", "is_compiling", "it", "kwargs", "last_hidden_state", "layer_idx", "layers", "logger", "min", "must", "nn", "norm", "not", "num_codebooks", "num_hidden_layers", "of", "offset", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "provided", "r", "raise", "range", "requires", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "should", "specify", "super", "the", "them", "to", "token", "torch", "tuple", "unsqueeze", "use_cache", "vocab_size", "warning", "warning_once", "were", "will"], "csm/modeling_csm.py:CsmCodebooksHead": ["ModelCodebooksHead", "Module", "None", "Parameter", "T", "__init__", "arange", "cache_position", "class", "codebook_idx", "codebook_idxs", "codebook_weight", "def", "dim", "else", "empty", "for", "forward", "functional", "hidden_size", "hidden_states", "if", "in", "is", "linear", "nn", "num_codebooks", "range", "return", "self", "seq_length", "shape", "stack", "super", "torch", "vocab_size", "weight"], "csm/modeling_csm.py:CsmDepthDecoderForCausalLM": ["Cache", "CausalLMOutputWithPast", "FloatTensor", "GenerationMixin", "LongTensor", "ModelCodebooksHead", "ModelDepthDecoderForCausalLM", "ModelDepthDecoderModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "backbone_last_hidden_state", "bool", "cache_position", "can_return_tuple", "class", "codebooks_head", "config", "contiguous", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "is_first_generation_step", "isinstance", "kwargs", "labels", "list", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "not", "num_codebooks", "outputs", "past_key_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "self", "shift_labels", "slice", "slice_indices", "super", "torch", "tuple", "use_cache", "vocab_size"], "csm/modeling_csm.py:CsmBackboneModelEmbeddings": ["Embedding", "False", "ModelBackboneModelEmbeddings", "Module", "__init__", "arange", "audio_tokens_offsets", "class", "config", "def", "dim", "embed_audio_tokens", "forward", "hidden_size", "input_embeds", "input_ids", "nn", "num_codebooks", "persistent", "register_buffer", "return", "self", "sum", "super", "torch", "vocab_size"], "csm/modeling_csm.py:CsmBackboneModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "False", "FloatTensor", "LongTensor", "ModelBackboneModel", "ModelBackboneModelEmbeddings", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "csm/modeling_csm.py:CsmForConditionalGeneration": ["AutoModel", "Cache", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "ModelBackboneModel", "ModelDepthDecoderForCausalLM", "ModelForConditionalGeneration", "ModelGenerationMixin", "ModelOutputWithPast", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_from_config", "_from_model_config", "_merge_input_ids_with_input_values", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "all", "and", "append", "arange", "args", "as_tuple", "attention_mask", "attentions", "attr", "audio_batch", "audio_codes", "audio_codes_mask", "audio_embeds", "audio_eos_embeds", "audio_eos_frame_ids", "audio_eos_token_id", "audio_eos_token_mask", "audio_lengths", "audio_token_id", "audio_token_mask", "audio_tokens_list", "auto_docstring", "backbone_hidden_states", "backbone_labels", "backbone_last_hidden_state", "backbone_last_hidden_states", "backbone_logits", "backbone_loss", "backbone_model", "backbone_outputs", "batch_input_values", "batch_input_values_cutoffs", "batched_audio_token_ids", "bias", "bool", "cache_position", "can_return_tuple", "class", "classmethod", "cls", "codebook_eos_token_id", "codebook_ids", "codec_config", "codec_model", "codec_outputs", "config", "def", "delattr", "depth_decoder", "depth_decoder_", "depth_decoder_attentions", "depth_decoder_attrs", "depth_decoder_config", "depth_decoder_hidden_states", "depth_decoder_ignore_frames_idxs", "depth_decoder_input_ids", "depth_decoder_labels", "depth_decoder_logits", "depth_decoder_loss", "depth_decoder_outputs", "depth_decoder_past_key_values", "device", "diff", "dim", "dtype", "el", "else", "embed_audio_tokens", "embed_text_tokens", "embed_tokens", "encode", "end_idx", "expand", "for", "forward", "from_config", "from_pretrained", "functional", "generation_config", "get", "get_audio_codes_mask", "get_input_embeddings", "hidden_size", "hidden_states", "i", "if", "in", "input_ids", "input_values", "input_values_cutoffs", "input_values_mask", "inputs_embeds", "int", "is", "isinstance", "items", "kwargs", "labels", "labels_expanded", "len", "list", "lm_head", "loading_info", "logits", "logits_to_keep", "long", "loss", "loss_function", "max", "max_audio_frames", "merged_inputs", "model", "model_inputs", "ndim", "nn", "no_grad", "nonzero", "not", "num_codebooks", "ones", "output_loading_info", "pad", "past_key_values", "pop", "position_ids", "post_init", "prefix", "prefix_len", "prepare_inputs_for_generation", "r", "range", "repeat", "return", "return_dict", "save_pretrained", "self", "set_input_embeddings", "setattr", "shape", "slice", "slice_indices", "squeeze", "stack", "start_idx", "startswith", "sum", "super", "text_vocab_size", "tie_codebooks_embeddings", "to_diff_dict", "torch", "train_idxs", "train_mask", "transformers_version", "transpose", "tuple", "unsqueeze", "update", "use_cache", "value", "vars", "vocab_size", "weight", "with", "zip"], "mra/modeling_mra.py:load_cuda_kernels": ["Model", "Model_cuda_kernel", "Model_cuda_kernels", "Path", "True", "__file__", "append_root", "cpp", "cu", "cuda_kernel", "cuda_launch", "def", "file", "files", "for", "global", "in", "kernels", "parent", "resolve", "return", "src_files", "src_folder", "torch_extension", "verbose"], "mra/modeling_mra.py:sparse_max": ["Model_cuda_kernel", "Model_max", "Model_qk_prod", "None", "The", "ValueError", "a", "be", "contiguous", "def", "dim", "dimension", "dimensional", "if", "index_max", "index_vals", "indices", "int", "key_num_block", "len", "max", "max_vals", "max_vals_scatter", "must", "of", "query_num_block", "raise", "return", "second", "size", "tensor", "the", "third", "transpose", "values"], "mra/modeling_mra.py:sparse_mask": ["Model_mask", "None", "ValueError", "a", "and", "arange", "batch_idx", "batch_size", "be", "block_size", "def", "device", "dimension", "dimensional", "dtype", "have", "if", "in", "indices", "len", "long", "mask", "must", "num_block", "raise", "reshape", "return", "same", "seq_len", "shape", "size", "tensor", "th", "the", "torch", "zero"], "mra/modeling_mra.py:mm_to_sparse": ["Model_cuda_kernel", "The", "ValueError", "_", "a", "batch_size", "be", "block_size", "by", "contiguous", "def", "dense_key", "dense_query", "dim", "dimension", "dimensional", "divisible", "first", "if", "indices", "int", "key_size", "len", "mm_to_sparse", "must", "of", "query_size", "raise", "reshape", "return", "size", "tensor", "third", "transpose"], "mra/modeling_mra.py:sparse_dense_mm": ["Model_cuda_kernel", "Model_dense_mm", "Model_query", "The", "ValueError", "a", "batch_size", "be", "block_size", "by", "contiguous", "def", "dense_key", "dense_qk_prod", "dim", "dimension", "dimensional", "divisible", "equal", "first", "if", "indices", "int", "key_size", "len", "must", "of", "query_num_block", "raise", "reshape", "return", "second", "size", "tensor", "the", "third", "to", "transpose"], "mra/modeling_mra.py:transpose_indices": ["Model_indices", "def", "dim_1_block", "dim_2_block", "div", "floor", "indices", "long", "return", "rounding_mode", "torch"], "mra/modeling_mra.py:MraSampledDenseMatMul": ["Function", "ModelSampledDenseMatMul", "None", "apply", "autograd", "backward", "block_size", "class", "ctx", "def", "dense_key", "dense_query", "forward", "grad", "grad_key", "grad_query", "indices", "indices_T", "key_num_block", "mm_to_sparse", "operator_call", "query_num_block", "return", "save_for_backward", "saved_tensors", "size", "sparse_dense_mm", "sparse_qk_prod", "staticmethod", "torch", "transpose", "transpose_indices"], "mra/modeling_mra.py:MraSparseDenseMatMul": ["Function", "ModelSparseDenseMatMul", "None", "apply", "autograd", "backward", "class", "ctx", "def", "dense_key", "forward", "grad", "grad_key", "grad_query", "indices", "indices_T", "key_num_block", "mm_to_sparse", "operator_call", "query_num_block", "return", "save_for_backward", "saved_tensors", "size", "sparse_dense_mm", "sparse_qk_prod", "sparse_query", "staticmethod", "torch", "transpose", "transpose_indices"], "mra/modeling_mra.py:MraReduceSum": ["ModelReduceSum", "None", "ValueError", "_", "a", "arange", "batch_idx", "batch_size", "be", "block_size", "class", "def", "device", "dim", "dimensional", "div", "dtype", "floor", "global_idxes", "if", "index_add", "indices", "key_num_block", "len", "long", "must", "num_block", "operator_call", "output", "query_num_block", "raise", "reshape", "return", "rounding_mode", "size", "sparse_query", "staticmethod", "sum", "temp", "tensor", "torch", "zeros"], "mra/modeling_mra.py:get_low_resolution_logit": ["Model_low_resolution_logit", "None", "True", "batch_size", "block_size", "def", "device", "dim", "dtype", "else", "float", "head_dim", "if", "is", "keepdims", "key", "key_hat", "low_resolution_logit", "low_resolution_logit_row_max", "mask", "math", "matmul", "max", "mean", "not", "num_block_per_row", "ones", "query", "query_hat", "reshape", "return", "seq_len", "size", "sqrt", "sum", "token_count", "torch", "transpose", "value", "value_hat", "values"], "mra/modeling_mra.py:get_block_idxes": ["False", "Model_block_idxes", "None", "True", "ValueError", "_", "a", "approx_mode", "approx_model", "batch_size", "def", "device", "diagonal", "diagonal_mask", "dim", "elif", "else", "f", "float", "full", "high_resolution_mask", "if", "indices", "initial_prior_diagonal_n_blocks", "initial_prior_first_n_blocks", "is", "largest", "low_resolution_logit", "min", "not", "num_blocks", "offset", "ones", "raise", "reshape", "return", "shape", "sorted", "sparse", "temp_mask", "threshold", "top_k_vals", "topk", "torch", "total_blocks_per_row", "tril", "triu", "valid", "value", "values"], "mra/modeling_mra.py:mra2_attention": ["Exception", "ModelReduceSum", "ModelSampledDenseMatMul", "ModelSparseDenseMatMul", "Model_attention", "Model_cuda_kernel", "None", "ValueError", "_", "approx_mode", "batch_size", "be", "block_size", "by", "config", "context_layer", "def", "dim", "divisible", "elif", "else", "exp", "float", "full", "get_block_idxes", "get_low_resolution_logit", "head_dim", "high_resolution_attn", "high_resolution_attn_out", "high_resolution_corr", "high_resolution_logit", "high_resolution_mask", "high_resolution_normalizer", "if", "indices", "initial_prior_diagonal_n_blocks", "initial_prior_first_n_blocks", "is", "key", "length", "log_correction", "low_resolution_attn", "low_resolution_attn_out", "low_resolution_corr", "low_resolution_logit", "low_resolution_logit_normalized", "low_resolution_logit_row_max", "low_resolution_normalizer", "mask", "math", "matmul", "max_vals", "max_vals_scatter", "meta_batch", "must", "no_grad", "not", "num_block_per_row", "num_blocks", "num_head", "operator_call", "or", "query", "raise", "repeat", "requires_grad_", "reshape", "return", "seq_len", "sequence", "size", "sparse", "sparse_mask", "sparse_max", "sqrt", "sum", "the", "token_count", "torch", "value", "value_hat", "with", "zeros_like"], "mra/modeling_mra.py:MraEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "mra/modeling_mra.py:MraSelfAttention": ["Could", "Dropout", "Exception", "Linear", "Model2_attention", "ModelSelfAttention", "Model_cuda_kernel", "Module", "None", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "approx_mode", "as", "attention", "attention_head_size", "attention_mask", "attention_probs_dropout_prob", "batch_size", "block_per_row", "cat", "class", "config", "context_layer", "contiguous", "custom", "def", "deformable", "device", "dim", "dropout", "e", "else", "embedding_size", "except", "f", "float", "for", "forward", "gpu_warp_size", "hasattr", "heads", "hidden", "hidden_size", "hidden_states", "if", "initial_prior_diagonal_n_blocks", "initial_prior_first_n_blocks", "int", "is", "is_cuda_platform", "is_ninja_available", "is_torch_cuda_available", "kernel", "kernel_loaded", "key", "key_layer", "load", "load_cuda_kernels", "logger", "max_position_embeddings", "min", "multi", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_block", "number", "of", "outputs", "pad_size", "permute", "position_embedding_type", "query", "query_layer", "raise", "repeat", "reshape", "return", "scale", "self", "seq_len", "shape", "size", "squeeze", "super", "the", "torch", "transpose", "try", "value", "value_layer", "view", "warning", "zeros"], "mra/modeling_mra.py:MraSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "mra/modeling_mra.py:MraAttention": ["ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "outputs", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "mra/modeling_mra.py:MraIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "mra/modeling_mra.py:MraOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "mra/modeling_mra.py:MraLayer": ["GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "add_cross_attention", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "mra/modeling_mra.py:MraEncoder": ["BaseModelOutputWithCrossAttentions", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "attention_mask", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "mra/modeling_mra.py:MraPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "mra/modeling_mra.py:MraLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "mra/modeling_mra.py:MraOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "mra/modeling_mra.py:MraPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "mra/modeling_mra.py:MraModel": ["BaseModelOutputWithCrossAttentions", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_hidden_states", "position_ids", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "mra/modeling_mra.py:MraForMaskedLM": ["CrossEntropyLoss", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "mra/modeling_mra.py:MraClassificationHead": ["ACT2FN", "Dropout", "Linear", "ModelClassificationHead", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "x"], "mra/modeling_mra.py:MraForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "mra/modeling_mra.py:MraForMultipleChoice": ["CrossEntropyLoss", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "ReLU", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_state", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "pre_classifier", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "mra/modeling_mra.py:MraForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "active_labels", "active_logits", "active_loss", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "ignore_index", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "tensor", "token_type_ids", "torch", "tuple", "type_as", "use_return_dict", "view", "where"], "mra/modeling_mra.py:MraForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTEmbeddings": ["ASTConfig", "ASTPatchEmbeddings", "Dropout", "Model", "Module", "None", "Parameter", "Tensor", "__init__", "batch_size", "cat", "class", "cls_token", "cls_tokens", "config", "def", "dim", "distillation_token", "distillation_tokens", "dropout", "embeddings", "expand", "forward", "frequency_out_dimension", "frequency_stride", "get_shape", "hidden_dropout_prob", "hidden_size", "input_values", "max_length", "nn", "num_mel_bins", "num_patches", "patch_embeddings", "patch_size", "position_embeddings", "return", "self", "shape", "super", "time_out_dimension", "time_stride", "torch", "zeros"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTPatchEmbeddings": ["ASTConfig", "Conv2d", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "embeddings", "flatten", "forward", "frequency_stride", "hidden_size", "input_values", "kernel_size", "nn", "patch_size", "projection", "return", "self", "stride", "super", "time_stride", "torch", "transpose", "unsqueeze"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "ASTConfig", "Callable", "False", "Linear", "Model", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTSelfOutput": ["ASTConfig", "Dropout", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTAttention": ["ASTConfig", "ASTSelfAttention", "ASTSelfOutput", "Model", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTIntermediate": ["ACT2FN", "ASTConfig", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTOutput": ["ASTConfig", "Dropout", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTLayer": ["ASTAttention", "ASTConfig", "ASTIntermediate", "ASTOutput", "GradientCheckpointingLayer", "LayerNorm", "Model", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTEncoder": ["ASTConfig", "ASTLayer", "BaseModelOutput", "False", "Model", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTPreTrainedModel": ["ASTConfig", "ASTEmbeddings", "ASTLayer", "ASTSelfAttention", "Conv2d", "LayerNorm", "Linear", "Model", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "distillation_token", "dtype", "elif", "fill_", "float32", "hidden_states", "if", "init", "initializer_range", "input_values", "is", "isinstance", "main_input_name", "mean", "module", "nn", "not", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTModel": ["ASTConfig", "ASTEmbeddings", "ASTEncoder", "ASTPatchEmbeddings", "ASTPreTrainedModel", "BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "Model", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "check_model_inputs", "class", "config", "def", "dict", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "input_values", "int", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "num_hidden_layers", "patch_embeddings", "pooled_output", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTMLPHead": ["ASTConfig", "Identity", "LayerNorm", "Linear", "Model", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_size", "hidden_state", "if", "layer_norm_eps", "layernorm", "nn", "num_labels", "return", "self", "super"], "audio_spectrogram_transformer/modeling_audio_spectrogram_transformer.py:ASTForAudioClassification": ["ASTConfig", "ASTMLPHead", "ASTModel", "ASTPreTrainedModel", "BaseModelOutputWithPooling", "Model", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "forward", "head_mask", "hidden_states", "if", "input_values", "is", "kwargs", "labels", "logits", "loss", "loss_function", "not", "num_labels", "outputs", "pooled_output", "pooler_output", "post_init", "r", "return", "self", "super", "torch"], "owlv2/modeling_owlv2.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "owlv2/modeling_owlv2.py:owlv2_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "owlv2/modeling_owlv2.py:Owlv2Output": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "owlv2/modeling_owlv2.py:_upcast": ["Tensor", "_upcast", "def", "dtype", "else", "float", "float32", "float64", "if", "in", "int", "int32", "int64", "is_floating_point", "return", "t", "torch"], "owlv2/modeling_owlv2.py:box_area": ["Model_area", "Modeles", "Tensor", "_upcast", "def", "return"], "owlv2/modeling_owlv2.py:box_iou": ["Model_area", "Model_iou", "Modeles1", "Modeles2", "None", "area1", "area2", "clamp", "def", "inter", "iou", "left_top", "max", "min", "return", "right_bottom", "torch", "union", "width_height"], "owlv2/modeling_owlv2.py:generalized_box_iou": ["Model_box_iou", "None", "ValueError", "all", "area", "be", "bottom_right", "box_iou", "boxes1", "boxes2", "but", "clamp", "corner", "def", "f", "format", "got", "if", "in", "iou", "max", "min", "must", "not", "raise", "return", "top_left", "torch", "union", "width_height", "x0", "x1", "y0", "y1"], "owlv2/modeling_owlv2.py:Owlv2ObjectDetectionOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "class", "class_embeds", "def", "dict", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits", "loss", "loss_dict", "not", "objectness_logits", "pred_boxes", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "owlv2/modeling_owlv2.py:Owlv2ImageGuidedObjectDetectionOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelImageGuidedObjectDetectionOutput", "ModelOutput", "None", "Optional", "class", "class_embeds", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits", "not", "query_image_embeds", "query_pred_boxes", "r", "return", "self", "target_pred_boxes", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "owlv2/modeling_owlv2.py:Owlv2VisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "bool", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "owlv2/modeling_owlv2.py:Owlv2TextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "arange", "class", "config", "def", "else", "embeddings", "expand", "forward", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "max_position_embeddings", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "shape", "super", "token_embedding", "torch", "vocab_size"], "owlv2/modeling_owlv2.py:Owlv2Attention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_shape", "and", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bmm", "bool", "bsz", "but", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_states", "mask", "must", "nn", "not", "num_attention_heads", "num_heads", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scale", "self", "seq_len", "should", "size", "softmax", "src_len", "super", "tensor", "tgt_len", "to", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights"], "owlv2/modeling_owlv2.py:Owlv2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "owlv2/modeling_owlv2.py:Owlv2EncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "owlv2/modeling_owlv2.py:Owlv2PreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "ModelModel", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "logit_scale", "logit_scale_init_value", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_embedding", "v_proj", "vision_embed_dim", "visual_projection", "weight", "zero_"], "owlv2/modeling_owlv2.py:Owlv2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "for", "forward", "gradient_checkpointing", "hidden_states", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "owlv2/modeling_owlv2.py:Owlv2TextTransformer": ["BaseModelOutputWithPooling", "LayerNorm", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "final_layer_norm", "forward", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "r", "return", "return_dict", "self", "shape", "size", "super", "to", "torch", "tuple", "use_return_dict", "view"], "owlv2/modeling_owlv2.py:Owlv2TextModel": ["BaseModelOutputWithPooling", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "tuple", "value"], "owlv2/modeling_owlv2.py:Owlv2VisionTransformer": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "dtype", "else", "embeddings", "encoder", "encoder_outputs", "eps", "expected_input_dtype", "forward", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layernorm", "return", "return_dict", "self", "super", "to", "torch", "tuple", "use_return_dict", "weight"], "owlv2/modeling_owlv2.py:Owlv2VisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "vision_model"], "owlv2/modeling_owlv2.py:Owlv2Model": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextTransformer", "ModelVisionConfig", "ModelVisionTransformer", "Model_loss", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "Union", "__init__", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "class", "config", "def", "device", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "linalg", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "ord", "output", "output_attentions", "output_hidden_states", "pixel_values", "pooler_output", "post_init", "projection_dim", "r", "raise", "return", "return_base_image_embeds", "return_dict", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_embeds_norm", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "use_return_dict", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "owlv2/modeling_owlv2.py:Owlv2BoxPredictionHead": ["FloatTensor", "GELU", "Linear", "ModelBoxPredictionHead", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense0", "dense1", "dense2", "forward", "gelu", "hidden_size", "image_features", "int", "nn", "out_dim", "output", "return", "self", "super", "torch", "vision_config", "width"], "owlv2/modeling_owlv2.py:Owlv2ClassPredictionHead": ["ELU", "FloatTensor", "Linear", "ModelClassPredictionHead", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "__init__", "batch_size", "class", "config", "def", "dense0", "device", "dim", "dtype", "einsum", "elu", "finfo", "float32", "forward", "hidden_size", "if", "image_class_embeds", "image_embeds", "is", "keepdim", "linalg", "logit_scale", "logit_shift", "min", "ndim", "nn", "norm", "not", "num_patches", "out_dim", "pd", "pq", "pred_logits", "qd", "query_dim", "query_embeds", "query_mask", "return", "self", "shape", "super", "text_config", "to", "torch", "tuple", "unsqueeze", "vision_config", "where", "zeros"], "owlv2/modeling_owlv2.py:Owlv2ForObjectDetection": ["False", "FloatTensor", "LayerNorm", "Model", "ModelBoxPredictionHead", "ModelClassPredictionHead", "ModelConfig", "ModelForObjectDetection", "ModelImageGuidedObjectDetectionOutput", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "Please", "Sigmoid", "Tensor", "True", "ValueError", "_", "__init__", "all", "an", "append", "arange", "argmin", "as", "attention_mask", "auto_docstring", "axis", "batch_size", "been", "best_box_ind", "best_box_indices", "best_class_embeds", "bool", "box_bias", "box_coord_bias", "box_coordinates", "box_head", "box_indices", "box_iou", "box_predictor", "box_size", "box_size_bias", "broadcast_to", "cat", "center_to_corners_format", "class", "class_embeds", "class_head", "class_predictor", "class_token_out", "clip", "compute_box_bias", "config", "d", "def", "deprecated", "detach", "device", "dim", "dtype", "each_query_box", "each_query_pred_boxes", "einsum", "else", "embed_image_query", "eps", "feature_map", "float32", "for", "forward", "full_like", "generalized_box_iou", "has", "height", "hidden_dim", "hidden_size", "i", "id", "if", "image_class_embeds", "image_embedder", "image_embeds", "image_feats", "image_features", "image_guided_detection", "image_size", "image_text_embedder", "in", "indexing", "input", "input_ids", "instead", "int", "interpolate_pos_encoding", "iou_threshold", "ious", "is", "last_hidden_state", "layer_norm", "layer_norm_eps", "log", "log1p", "logits", "lru_cache", "max", "max_text_queries", "maxsize", "mean", "mean_embeds", "mean_sim", "meshgrid", "new_size", "nn", "nonzero", "normalize_grid_corner_coordinates", "not", "num_patches", "num_patches_height", "num_patches_width", "numel", "objectness_head", "objectness_logits", "objectness_predictor", "out_dim", "output", "output_attentions", "output_hidden_states", "outputs", "pass", "patch_size", "pixel_values", "post_init", "post_layernorm", "pred_boxes", "pred_boxes_as_corners", "pred_boxes_device", "pred_logits", "query_embeds", "query_feature_map", "query_image_embeds", "query_image_feats", "query_image_features", "query_mask", "query_pixel_values", "query_pred_boxes", "r", "raise", "range", "reshape", "return", "return_dict", "selected_embeddings", "selected_inds", "self", "shape", "sigmoid", "squeeze", "stack", "staticmethod", "super", "target_pred_boxes", "tensor", "text_embeds", "text_model_output", "text_outputs", "to", "to_tuple", "torch", "tuple", "view", "vision_config", "vision_model", "vision_model_output", "vision_outputs", "width", "x", "x_coordinates", "xx", "xy", "y_coordinates", "yy"], "decision_transformer/modeling_decision_transformer.py:eager_attention_forward": ["Model_attention_forward", "None", "attention_mask", "attn_dropout", "attn_output", "attn_weights", "bias", "causal_mask", "def", "device", "dim", "dtype", "finfo", "float", "full", "functional", "head_mask", "if", "is", "is_cross_attention", "key", "key_length", "kwargs", "layer_idx", "mask_value", "matmul", "min", "module", "nn", "not", "query", "query_length", "return", "scale_attn_by_inverse_layer_idx", "scale_attn_weights", "shape", "size", "softmax", "to", "torch", "transpose", "type", "value", "where"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerGPT2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Conv1D", "Dropout", "EncoderDecoderCache", "Error", "False", "FloatTensor", "If", "LongTensor", "ModelGPT2Attention", "Module", "None", "Optional", "Please", "RuntimeError", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_attn_implementation", "_upcast_and_reordered_attn", "alpha", "and", "as", "attention", "attention_interface", "attention_mask", "attn_dropout", "attn_output", "attn_pdrop", "attn_weights", "autocast", "baddbmm", "be", "beta", "bias", "bool", "bsz", "by", "c_attn", "c_proj", "cache_position", "cat", "causal_mask", "class", "config", "contiguous", "cross", "cross_attention_cache", "curr_past_key_value", "def", "defined", "deprecate_kwarg", "device", "dim", "divisible", "dk", "does", "dropout", "dtype", "eager", "eager_attention_forward", "else", "embed_dim", "empty", "enabled", "encoder_attention_mask", "encoder_hidden_states", "f", "find_pruneable_heads_and_indices", "finfo", "float", "float32", "forward", "functional", "get", "got", "hasattr", "have", "head_dim", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "index_attn", "instantiate", "is", "is_causal", "is_cross_attention", "is_updated", "isinstance", "k", "k_seq_len", "key", "key_length", "key_states", "keys", "kwargs", "layer_idx", "layers", "len", "make", "mask_value", "masked_bias", "matmul", "max_position_embeddings", "max_positions", "min", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "ones", "or", "output_attentions", "p", "past_key_value", "past_key_values", "persistent", "prune_conv1d_layer", "prune_heads", "pruned_heads", "q", "q_attn", "q_seq_len", "query", "query_length", "query_states", "raise", "register_buffer", "reorder_and_upcast_attn", "reshape", "resid_dropout", "resid_pdrop", "return", "scale_attn_by_inverse_layer_idx", "scale_attn_weights", "scale_factor", "self", "self_attention_cache", "set", "shape", "shape_kv", "shape_q", "size", "softmax", "split", "split_size", "super", "sure", "tensor", "the", "to", "torch", "training", "transpose", "tril", "tuple", "type", "union", "upcasting", "update", "used", "using_eager", "value", "value_states", "values", "version", "view", "weights", "where", "with"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerGPT2MLP": ["ACT2FN", "Conv1D", "Dropout", "FloatTensor", "ModelGPT2MLP", "Module", "Optional", "__init__", "act", "activation_function", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "resid_pdrop", "return", "self", "super", "torch", "tuple"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerGPT2Block": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "LayerNorm", "LongTensor", "ModelGPT2Attention", "ModelGPT2Block", "ModelGPT2MLP", "None", "Optional", "Tensor", "True", "Union", "ValueError", "__init__", "add_cross_attention", "are", "attention", "attention_mask", "attn", "attn_output", "be", "bool", "by", "cache_position", "class", "config", "cross", "cross_attn_output", "cross_attn_weights", "crossattention", "def", "deprecate_kwarg", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_hidden_states", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "inner_dim", "instantiated", "is", "is_cross_attention", "kwargs", "layer_idx", "layer_norm_epsilon", "layers", "ln_1", "ln_2", "ln_cross_attn", "mlp", "n_inner", "new_name", "nn", "not", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "raise", "residual", "return", "self", "self_attn_weights", "setting", "super", "to", "torch", "tuple", "use_cache", "version", "with"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerGPT2PreTrainedModel": ["Conv1D", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelGPT2PreTrainedModel", "None", "PreTrainedModel", "True", "__init__", "_can_compile_fullgraph", "_init_weights", "and", "base_model_prefix", "bias", "c_proj", "class", "config", "data", "def", "elif", "fill_", "for", "if", "in", "initializer_range", "inputs", "is", "is_parallelizable", "isinstance", "kwargs", "math", "mean", "module", "n_layer", "name", "named_parameters", "nn", "normal_", "not", "p", "padding_idx", "self", "sqrt", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerGPT2Model": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelGPT2Block", "ModelGPT2Model", "ModelGPT2PreTrainedModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "arange", "at", "attention_mask", "attentions", "batch_size", "be", "block", "bool", "both", "cache_position", "cannot", "checkpointing", "class", "config", "cross_attentions", "cuda", "def", "defined", "device", "device_map", "drop", "dtype", "either", "elif", "else", "embd_pdrop", "embed_dim", "encoder_attention_mask", "encoder_batch_size", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "finfo", "for", "forward", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "h", "has", "have", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "isinstance", "items", "k", "last_device", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "ln_f", "logger", "max_position_embeddings", "min", "model_parallel", "n_layer", "new_embeddings", "nn", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "past_key_values", "past_seen_tokens", "position_embeds", "position_ids", "post_init", "raise", "range", "return", "return_dict", "same", "self", "set_device", "set_input_embeddings", "shape", "size", "specify", "str", "super", "the", "time", "to", "token_type_embeds", "token_type_ids", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "warning_once", "with", "wpe", "wte"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerOutput": ["FloatTensor", "ModelOutput", "None", "Optional", "action_preds", "attentions", "class", "hidden_states", "last_hidden_state", "r", "return_preds", "state_preds", "torch"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "states", "std", "supports_gradient_checkpointing", "weight", "zero_"], "decision_transformer/modeling_decision_transformer.py:DecisionTransformerModel": ["Embedding", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "ModelGPT2Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Sequential", "Tanh", "Union", "__init__", "act_dim", "action_embeddings", "action_preds", "action_tanh", "actions", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "dim", "dtype", "else", "embed_action", "embed_ln", "embed_return", "embed_state", "embed_timestep", "encoder", "encoder_outputs", "forward", "hidden_size", "hidden_states", "if", "inputs_embeds", "is", "last_hidden_state", "long", "max_ep_len", "nn", "not", "ones", "output_attentions", "output_hidden_states", "permute", "position_ids", "post_init", "predict_action", "predict_return", "predict_state", "r", "reshape", "return", "return_dict", "return_preds", "returns_embeddings", "returns_to_go", "rewards", "self", "seq_length", "shape", "stack", "stacked_attention_mask", "stacked_inputs", "state_dim", "state_embeddings", "state_preds", "states", "super", "time_embeddings", "timesteps", "torch", "tuple", "use_return_dict", "x", "zeros"], "mpt/modeling_mpt.py:build_mpt_alibi_tensor": ["Model_Model_alibi_tensor", "None", "alibi", "alibi_bias_max", "arange", "base", "ceil", "concat", "def", "device", "dim", "dtype", "float", "if", "int32", "int64", "log2", "math", "num_heads", "num_heads_power_of_2", "pow", "r", "return", "sequence_length", "slopes", "squeeze", "torch", "view"], "mpt/modeling_mpt.py:MptAttention": ["Cache", "Expecting", "False", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "ValueError", "Wqkv", "__init__", "attention_mask", "attention_scores", "attn_config", "attn_dropout_p", "attn_output", "attn_pdrop", "attn_weights", "batch_size", "be", "bias", "cache_kwargs", "cache_position", "chunk", "clamp", "class", "clip_qkv", "config", "context_states", "contiguous", "def", "deprecate_kwarg", "dim", "dimensions", "dropout", "dtype", "else", "f", "finfo", "float", "forward", "functional", "get_seq_length", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "key_length", "key_states", "layer_idx", "len", "masked_fill", "math", "matmul", "max", "max_seq_len", "max_seq_length", "min", "mixed_qkv", "n_heads", "new_name", "nn", "not", "out_proj", "p", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_key_index", "position_bias_query_index", "query_length", "query_states", "raise", "reshape", "return", "self", "seq_length", "shape", "size", "softmax", "softmax_scale", "sqrt", "super", "to", "torch", "training", "transpose", "update", "value_states", "version", "view"], "mpt/modeling_mpt.py:MptMLP": ["F", "False", "GELU", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "__init__", "act", "approximate", "attn_config", "attn_pdrop", "bias", "class", "config", "def", "down_proj", "dropout", "forward", "hidden_dropout", "hidden_size", "hidden_states", "intermediate_output", "nn", "none", "output", "p", "residual", "return", "self", "super", "torch", "training", "up_proj"], "mpt/modeling_mpt.py:MptBlock": ["Cache", "Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelBlock", "ModelConfig", "ModelMLP", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn", "attn_config", "attn_outputs", "attn_pdrop", "attn_weights", "bias", "bool", "cache_position", "class", "config", "def", "dropout_rate", "eps", "ffn", "forward", "hidden_size", "hidden_states", "int", "layer_idx", "layer_norm_epsilon", "layer_past", "layernorm_output", "n_heads", "nn", "norm_1", "norm_2", "num_heads", "output", "output_attentions", "past_key_values", "position_bias", "resid_attn_dropout", "residual", "return", "self", "super", "torch", "use_cache"], "mpt/modeling_mpt.py:MptPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelBlock", "ModelConfig", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "Tensor", "True", "__init__", "_convert_to_Model_cache", "_init_weights", "_keys_to_ignore_on_load_missing", "_no_split_modules", "base_model_prefix", "batch_size", "batch_size_times_num_heads", "bias", "class", "config", "data", "def", "deprecate_kwarg", "elif", "fill_", "for", "head_dim", "if", "in", "initializer_range", "inputs", "is", "isinstance", "kwargs", "layer_past", "lm_head", "mean", "module", "new_name", "nn", "normal_", "not", "num_heads", "padding_idx", "past_key_value", "past_key_values", "r", "reshape", "return", "self", "seq_length", "shape", "staticmethod", "std", "super", "supports_gradient_checkpointing", "torch", "transformer", "tuple", "version", "weight", "zero_"], "mpt/modeling_mpt.py:MptModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "False", "LayerNorm", "LongTensor", "ModelBlock", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_prepare_4d_causal_attention_mask", "a", "alibi", "alibi_bias_max", "all_hidden_states", "all_self_attentions", "an", "and", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bias", "block", "blocks", "bool", "both", "build_Model_alibi_tensor", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "def", "deprecated", "device", "e", "either", "elif", "else", "eps", "for", "forward", "from_legacy_cache", "g", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "inputs_embeds", "instance", "instead", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layer_past", "logger", "max_seq_len", "n_heads", "n_layers", "new_embeddings", "nn", "norm_f", "not", "num_heads", "of", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "pass", "past_key_values", "past_key_values_length", "position_bias", "post_init", "r", "raise", "range", "removed", "return", "return_dict", "same", "self", "seq_length", "seq_length_with_past", "sequence_length", "set_input_embeddings", "shape", "should", "specify", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "vocab_size", "warning_once", "will", "with", "wte"], "mpt/modeling_mpt.py:MptForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_function", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "to", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "mpt/modeling_mpt.py:MptForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "Linear", "LongTensor", "MSELoss", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "warning_once", "will", "with"], "mpt/modeling_mpt.py:MptForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "Linear", "LongTensor", "ModelConfig", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "classifier_dropout", "config", "def", "deprecated_arguments", "device", "dropout", "elif", "else", "forward", "hasattr", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "seq_length", "shape", "super", "to", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view"], "mpt/modeling_mpt.py:MptForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "clip/modeling_clip.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "clip/modeling_clip.py:clip_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "clip/modeling_clip.py:_get_vector_norm": ["Tensor", "True", "_get_vector_norm", "def", "dim", "keepdim", "normed_tensor", "pow", "return", "square_tensor", "sum", "sum_tensor", "tensor", "torch"], "clip/modeling_clip.py:CLIPVisionModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "clip/modeling_clip.py:CLIPTextModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "clip/modeling_clip.py:CLIPOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "clip/modeling_clip.py:CLIPVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "Model", "ModelVisionConfig", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "clip/modeling_clip.py:CLIPTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelTextConfig", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "clip/modeling_clip.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "True", "attention_mask", "attn_output", "attn_weights", "bool", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "output_attentions", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "clip/modeling_clip.py:CLIPAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelTextConfig", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "clip/modeling_clip.py:CLIPMLP": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "clip/modeling_clip.py:CLIPEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelMLP", "ModelTextConfig", "ModelVisionConfig", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "clip/modeling_clip.py:CLIPPreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelForImageClassification", "ModelMLP", "ModelModel", "ModelTextEmbeddings", "ModelTextModelWithProjection", "ModelVisionEmbeddings", "ModelVisionModelWithProjection", "None", "PreTrainedModel", "True", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "base_model_prefix", "bias", "class", "class_embedding", "classifier", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_embedding", "v_proj", "vision_config", "vision_embed_dim", "visual_projection", "weight", "zero_"], "clip/modeling_clip.py:CLIPEncoder": ["BaseModelOutput", "False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "self", "super", "torch"], "clip/modeling_clip.py:CLIPTextTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "Model", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "ValueError", "You", "__init__", "_attn_implementation", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eos_token_id", "eps", "final_layer_norm", "flash_attention_2", "forward", "have", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "raise", "return", "self", "shape", "size", "specify", "super", "to", "torch", "view"], "clip/modeling_clip.py:CLIPTextModel": ["BaseModelOutputWithPooling", "False", "Model", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "__init__", "_no_split_modules", "_supports_flash_attn", "attention_mask", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "value"], "clip/modeling_clip.py:CLIPVisionTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "Model", "ModelEncoder", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Optional", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layrnorm", "raise", "return", "self", "specify", "super", "to", "torch"], "clip/modeling_clip.py:CLIPVisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Model", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionTransformer", "Module", "None", "Optional", "__init__", "_no_split_modules", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "self", "super", "torch", "vision_model"], "clip/modeling_clip.py:CLIPModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelEncoderLayer", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "Model_loss", "None", "Optional", "Parameter", "Tensor", "TypeError", "__init__", "_from_config", "_get_vector_norm", "_no_split_modules", "_supports_flash_attn", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "can_return_tuple", "class", "config", "def", "device", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "not", "of", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "return", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "type", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "clip/modeling_clip.py:CLIPTextModelWithProjection": ["BaseModelOutputWithPooling", "False", "Linear", "Model", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "ModelTextModelOutput", "Module", "None", "Optional", "Tensor", "__init__", "_from_config", "_no_split_modules", "_supports_flash_attn", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "input_ids", "last_hidden_state", "nn", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "return", "self", "set_input_embeddings", "super", "text_embeds", "text_model", "text_outputs", "text_projection", "token_embedding", "torch", "value"], "clip/modeling_clip.py:CLIPVisionModelWithProjection": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "Model", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionModelOutput", "Module", "None", "Optional", "__init__", "_from_config", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "image_embeds", "interpolate_pos_encoding", "last_hidden_state", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "pooled_output", "pooler_output", "post_init", "projection_dim", "r", "return", "self", "super", "torch", "vision_model", "vision_outputs", "visual_projection"], "clip/modeling_clip.py:CLIPForImageClassification": ["BaseModelOutputWithPooling", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelVisionModel", "None", "Optional", "Tensor", "__init__", "_from_config", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "dim", "else", "forward", "hidden_size", "hidden_states", "if", "is", "labels", "last_hidden_state", "logits", "loss", "loss_function", "main_input_name", "mean", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch", "vision_config", "vision_model"], "zamba2/modeling_zamba2.py:Zamba2RMSNormGated": ["ModelRMSNormGated", "Module", "None", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "functional", "gate", "group_count", "group_size", "hidden_size", "hidden_states", "hidden_states_group", "if", "input_dtype", "is", "keepdim", "last_dim", "mean", "nn", "not", "ones", "pow", "prefix_dims", "return", "rsqrt", "self", "shape", "silu", "super", "to", "torch", "variance", "variance_epsilon", "view", "weight"], "zamba2/modeling_zamba2.py:Zamba2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "zamba2/modeling_zamba2.py:Zamba2HybridDynamicCache": ["Any", "False", "LongTensor", "ModelConfig", "ModelHybridDynamicCache", "None", "Optional", "Tensor", "_", "__getitem__", "__init__", "__len__", "_buffers", "_modules", "_parameters", "append", "batch_size", "beam_idx", "cache_kwargs", "cache_position", "cat", "clamp", "class", "config", "conv_kernel_size", "conv_state", "conv_states", "def", "device", "dict", "dim", "dims", "dtype", "else", "float16", "for", "get_seq_length", "has_previous_state", "hidden_size", "hybrid", "i", "if", "in", "index_select", "int", "intermediate_size", "is_compileable", "key_cache", "key_states", "layer_idx", "layers_block_type", "len", "mamba_d_conv", "mamba_d_state", "mamba_expand", "mamba_headdim", "mamba_ngroups", "n_mamba_heads", "new_conv_state", "not", "num_hidden_layers", "numel", "or", "range", "reorder_cache", "reset", "return", "roll", "self", "shape", "shifts", "ssm_state_size", "ssm_states", "str", "tensor", "to", "torch", "transformer_layers", "tuple", "update", "update_conv_state", "value_cache", "value_states", "zero_", "zeros"], "zamba2/modeling_zamba2.py:Zamba2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "zamba2/modeling_zamba2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "zamba2/modeling_zamba2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "zamba2/modeling_zamba2.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "zamba2/modeling_zamba2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "zamba2/modeling_zamba2.py:Zamba2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Identity", "Linear", "ModelAttention", "ModelConfig", "ModelHybridDynamicCache", "Module", "ModuleList", "None", "Optional", "Sequential", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "adapter_layer_idx", "adapter_rank", "append", "apply_rotary_pos_emb", "attention_dropout", "attention_head_dim", "attention_hidden_size", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "block_id", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "enumerate", "for", "forward", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "hybrid_layer_ids", "i", "if", "in", "index", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_block_map", "layer_dic", "layer_idx", "linear_k_adapter", "linear_k_adapter_list", "linear_q_adapter", "linear_q_adapter_list", "linear_v_adapter", "linear_v_adapter_list", "max_position_embeddings", "new_name", "nn", "not", "num_attention_heads", "num_fwd_mem_blocks", "num_key_value_groups", "num_key_value_heads", "num_mem_blocks", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "range", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "use_mem_rope", "use_shared_attention_adapter", "v_proj", "value", "value_states", "version", "view"], "zamba2/modeling_zamba2.py:pad_tensor_by_size": ["Model", "Model_shape", "Model_size", "Model_tensor_by_size", "Tensor", "constant", "def", "else", "functional", "if", "input_tensor", "int", "len", "mode", "nn", "return", "shape", "torch", "value"], "zamba2/modeling_zamba2.py:reshape_into_chunks": ["Model", "Model_into_chunks", "chunk_size", "def", "else", "if", "input_tensor", "len", "pad_size", "pad_tensor_by_size", "return", "shape"], "zamba2/modeling_zamba2.py:segment_sum": ["Model_sum", "None", "bool", "chunk_size", "cumsum", "def", "device", "diagonal", "dim", "dtype", "expand", "inf", "input_tensor", "mask", "masked_fill", "ones", "return", "size", "tensor_segsum", "torch", "tril"], "zamba2/modeling_zamba2.py:Zamba2MambaMixer": ["A", "AILab", "A_cumsum", "A_log", "B", "B_decay_contraction", "C", "C_reshaped", "C_times_states", "Conv1d", "D", "D_residual", "Dao", "Falling", "False", "G", "G_intermediate", "L", "Linear", "M", "M_intermediate", "ModelConfig", "ModelHybridDynamicCache", "ModelMambaMixer", "ModelRMSNormGated", "Module", "None", "Optional", "Parameter", "SiLU", "Tensor", "The", "To", "True", "Y_diag", "Y_off", "_", "__init__", "act", "activation", "add_bias_linear", "all", "and", "arange", "attention_mask", "available", "back", "batch_size", "because", "bias", "bmm", "cache_params", "cat", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk_size", "clamp", "class", "clone", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_dim", "conv_kernel_size", "conv_state", "conv_states", "copy_", "cuda", "cuda_kernels_forward", "cumsum", "dA", "dB", "dBx", "d_mlp", "d_to_remove", "decay_chunk", "decay_states", "def", "device", "dim", "dims", "dt", "dt_bias", "dt_limit", "dt_limit_kwargs", "dt_softplus", "dtype", "else", "eps", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "group_size", "groups", "groups_time_state_size", "has_previous_state", "head_dim", "headdim", "hidden_size", "hidden_states", "hidden_states_B_C", "hidden_states_B_C_t", "hidden_states_reshaped", "https", "if", "implementation", "in", "in_channels", "in_proj", "in_projected_states", "input_not_masked", "input_states", "install", "int", "intermediate_size", "is", "is_fast_path_available", "kernel_size", "layer_idx", "log", "logger", "mamba", "mamba_chunk_scan_combined", "mamba_d_conv", "mamba_d_state", "mamba_expand", "mamba_headdim", "mamba_ngroups", "mamba_split_conv1d_scan_combined", "n_groups", "n_mamba_heads", "naive", "ndim", "new_states", "ngroups", "nn", "norm", "norm_before_gate", "not", "num_heads", "of", "one", "ones", "or", "out", "out_channels", "out_proj", "outproj_bias", "outproj_weight", "output_size", "pad", "pad_size", "pad_tensor_by_size", "padding", "path", "permute", "previous_states", "projected_states", "projection_size", "repeat_interleave", "reshape", "reshape_into_chunks", "result", "return", "return_final_states", "rmsnorm_eps", "rmsnorm_weight", "roll", "scan_output", "segment_sum", "selective_state_update", "self", "seq_idx", "seq_len", "shape", "shifts", "silu", "softplus", "spaces", "split", "split_projection_dim", "squeeze", "ssm_state", "ssm_state_size", "ssm_states", "ssm_states_reshaped", "state", "state_decay_out", "state_decay_out_permuted", "states", "states_permuted", "sum", "super", "swish", "t", "the", "time_step", "time_step_limit", "time_step_max", "time_step_min", "to", "torch", "torch_forward", "training", "transpose", "type", "unsqueeze", "use_conv_bias", "use_mem_eff_path", "variance_epsilon", "view", "warning_once", "weight", "x", "y", "z", "zeros", "zeros_like"], "zamba2/modeling_zamba2.py:Zamba2MLP": ["ACT2FN", "False", "Identity", "Linear", "ModelConfig", "ModelMLP", "Module", "ModuleList", "None", "Optional", "Sequential", "__init__", "act_fn", "adapter_rank", "add_bias_linear", "append", "bias", "block_id", "chunk", "class", "config", "def", "dim", "down_proj", "else", "enumerate", "for", "forward", "gate_up_proj", "gate_up_proj_adapter", "gate_up_proj_adapter_list", "gate_up_state", "hidden_act", "hidden_size", "hidden_state", "hybrid_layer_ids", "i", "if", "in", "index", "int", "intermediate_size", "layer_block_map", "layer_dic", "layer_idx", "nn", "num_fwd_mem_blocks", "num_mem_blocks", "output", "range", "return", "self", "super", "torch", "value"], "zamba2/modeling_zamba2.py:Zamba2AttentionDecoderLayer": ["False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelAttention", "ModelAttentionDecoderLayer", "ModelConfig", "ModelHybridDynamicCache", "ModelMLP", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_hidden_size", "attention_mask", "block_id", "bool", "class", "concatenate", "config", "def", "deprecate_kwarg", "dim", "eps", "feed_forward", "forward", "hidden_size", "hidden_states", "hybrid_layer_ids", "if", "input_layernorm", "int", "kwargs", "layer_idx", "len", "new_name", "nn", "num_fwd_mem_blocks", "num_gs", "original_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "pre_ff_layernorm", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "version"], "zamba2/modeling_zamba2.py:Zamba2MambaDecoderLayer": ["False", "FloatTensor", "LongTensor", "ModelConfig", "ModelHybridDynamicCache", "ModelMambaDecoderLayer", "ModelMambaMixer", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_params", "cache_position", "causal_mask", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is", "kwargs", "layer_idx", "mamba", "new_name", "nn", "not", "original_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "residual", "return", "rms_norm_eps", "self", "self_attn_weights", "super", "torch", "transformer_hidden_states", "tuple", "use_cache", "version"], "zamba2/modeling_zamba2.py:Zamba2HybridLayer": ["False", "FloatTensor", "Linear", "LongTensor", "ModelAttentionDecoderLayer", "ModelHybridDynamicCache", "ModelHybridLayer", "ModelMambaDecoderLayer", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "causal_mask", "class", "def", "deprecate_kwarg", "forward", "hidden_states", "if", "int", "layer_idx", "layer_outputs", "linear", "mamba", "mamba_decoder", "new_name", "nn", "original_hidden_states", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "return", "self", "self_attn_weights", "shared_transformer", "super", "torch", "transformer_hidden_states", "tuple", "use_cache", "version"], "zamba2/modeling_zamba2.py:Zamba2PreTrainedModel": ["A", "A_log", "D", "ModelAttentionDecoderLayer", "ModelConfig", "ModelMambaDecoderLayer", "ModelMambaMixer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "arange", "base_model_prefix", "clamp", "class", "config", "copy_", "data", "def", "dt", "dt_bias", "exp", "expm1", "fill_", "if", "inv_dt", "isinstance", "log", "math", "min", "model", "module", "n_mamba_heads", "num_heads", "past_key_values", "rand", "self", "super", "supports_gradient_checkpointing", "time_step_floor", "time_step_max", "time_step_min", "torch"], "zamba2/modeling_zamba2.py:Zamba2Model": ["AttentionMaskConverter", "BaseModelOutputWithPast", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "ModelAttentionDecoderLayer", "ModelConfig", "ModelHybridDynamicCache", "ModelHybridLayer", "ModelMambaDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__call__", "__init__", "_attn_implementation", "_gradient_checkpointing_func", "_layer_type", "_tied_weights_keys", "_unmask_unattended", "_update_causal_mask", "adapter_id", "adapter_pattern", "all_hidden_states", "all_self_attns", "and", "append", "arange", "at", "attention_mask", "attentions", "attn_adapter_pattern", "auto_docstring", "batch_size", "bias", "block", "block_id", "blocks", "bool", "both", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "clone", "compile", "config", "cuda", "cycle", "def", "device", "diagonal", "dim", "down_proj", "dtype", "either", "elif", "else", "embed_tokens", "enumerate", "eps", "eq", "expand", "extended", "feed_forward", "fill_value", "final_layernorm", "finfo", "first_transformer_layer_id", "flash_attention_2", "for", "forward", "full", "gate_up_proj", "gate_up_proj_adapter_list", "get_layers", "get_seq_length", "gradient", "gradient_checkpointing", "has_previous_state", "hidden_size", "hidden_states", "hybrid", "hybrid_layer_ids", "i", "if", "in", "incompatible", "input_ids", "input_layernorm", "input_tensor", "inputs_embeds", "is", "iter", "k", "k_proj", "last_hidden_state", "layer", "layer_id", "layer_idx", "layer_outputs", "layer_type", "layers", "layers_block_type", "len", "linear_k_adapter_list", "linear_layers", "linear_q_adapter_list", "linear_v_adapter_list", "logger", "main_keys_pattern", "mamba", "mamba_layers", "mask_length", "masked_fill", "max_position_embeddings", "min", "min_dtype", "must", "next", "nn", "not", "npu", "num_hidden_layers", "num_mem_blocks", "o_proj", "one", "original_hidden_states", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "pre_ff_layernorm", "prefix_pattern", "q_proj", "r", "raise", "range", "re", "rescaled", "reshape", "return", "return_dict", "rf", "rms_norm_eps", "rope_theta", "rotary_emb", "same", "sdpa", "self", "self_attn", "sequence_length", "set", "shape", "shared_transformer", "specify", "str", "super", "target_length", "the", "time", "to", "to_tuple", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_long_context", "use_mem_rope", "use_return_dict", "use_shared_attention_adapter", "using", "v_proj", "vocab_size", "warning_once", "weight", "with", "xpu"], "zamba2/modeling_zamba2.py:Zamba2ForCausalLM": ["CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelHybridDynamicCache", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "contiguous", "cumsum", "def", "device", "dtype", "elif", "else", "empty_past_kv", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "key", "kwargs", "labels", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "masked_fill_", "model", "model_inputs", "nn", "not", "num_logits_to_keep", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "shape", "slice", "slice_indices", "super", "torch", "tuple", "update", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "zamba2/modeling_zamba2.py:Zamba2ForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "list", "logger", "logits", "long", "loss", "loss_fct", "may", "model", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "tokens", "torch", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "janus/modeling_janus.py:JanusPreTrainedModel": ["False", "LlamaDecoderLayer", "ModelConfig", "ModelPreTrainedModel", "ModelVisionEncoderLayer", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "base_model_prefix", "causal_mask", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "janus/modeling_janus.py:JanusVQVAEOutput": ["FloatTensor", "ModelOutput", "ModelVQVAEOutput", "None", "Optional", "class", "decoded_pixel_values", "embedding_loss", "r", "torch"], "janus/modeling_janus.py:JanusBaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "janus/modeling_janus.py:JanusCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "janus/modeling_janus.py:JanusVisionEmbeddings": ["Conv2d", "Embedding", "False", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "_", "__init__", "align_corners", "and", "arange", "bicubic", "bool", "class", "config", "def", "dim", "dtype", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "out_channels", "padding", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "pos_embeds", "position_embedding", "position_ids", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "valid", "view", "weight", "width"], "janus/modeling_janus.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "janus/modeling_janus.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "janus/modeling_janus.py:JanusVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "Identity", "LayerNorm", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "_", "__init__", "_attn_implementation", "and", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "by", "class", "config", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "output", "proj_dropout", "projection_dropout", "projection_layer", "q_norm", "q_proj", "qk_norm", "query_states", "raise", "reshape", "return", "scale", "scaling", "self", "seq_len", "size", "super", "torch", "training", "transpose", "use_qk_norm", "v_proj", "value_states", "view"], "janus/modeling_janus.py:JanusVisionMLP": ["ACT2FN", "Dropout", "Linear", "ModelVisionConfig", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "dropout1", "dropout2", "fc1", "fc2", "forward", "hidden_act", "hidden_dropout_rate", "hidden_size", "hidden_states", "int", "intermediate_size", "mlp_ratio", "nn", "return", "self", "super", "torch"], "janus/modeling_janus.py:JanusVisionEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoderLayer", "ModelVisionMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "janus/modeling_janus.py:JanusVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "janus/modeling_janus.py:JanusAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_attn_implementation", "_shape", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bsz", "by", "cat", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key_states", "kwargs", "mixed_qkv", "must", "nn", "not", "num_attention_heads", "num_heads", "permute", "projection", "q_bias", "qkv", "qkv_bias", "query_states", "raise", "requires_grad", "reshape", "return", "scale", "scaling", "self", "seq_len", "size", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_bias", "value_states", "view", "zeros", "zeros_like"], "janus/modeling_janus.py:JanusMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "janus/modeling_janus.py:JanusEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "janus/modeling_janus.py:JanusVisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "attentions", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "get_input_embeddings", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "main_input_name", "nn", "pixel_values", "pooled_output", "pooler_output", "post_init", "post_layernorm", "raise", "return", "self", "specify", "super", "to", "torch", "tuple"], "janus/modeling_janus.py:JanusVisionAlignerMLP": ["ACT2FN", "Linear", "ModelVisionAlignerMLP", "ModelVisionConfig", "Module", "ModuleList", "_", "__init__", "activation_fn", "class", "config", "def", "depth", "fc1", "for", "forward", "hidden_act", "hidden_layers", "hidden_size", "hidden_states", "in", "layer", "nn", "projection_dim", "range", "return", "self", "super"], "janus/modeling_janus.py:JanusVQVAEVectorQuantizer": ["Embedding", "F", "FloatTensor", "LongTensor", "ModelVQVAEConfig", "ModelVQVAEVectorQuantizer", "Module", "Tensor", "True", "__init__", "argmin", "batch_size", "bd", "beta", "bn", "class", "config", "contiguous", "def", "detach", "dim", "distances", "dn", "einsum", "emb_dim", "embed_dim", "embedding", "embedding_dim", "forward", "get_codebook_entry", "getattr", "hidden_state", "hidden_state_flattened", "hidden_state_quant", "image_tokens", "int", "keepdim", "loss", "mean", "min_encoding_indices", "nn", "normalize", "num_embeddings", "num_patches", "p", "permute", "quant_state_dims", "return", "self", "shape", "sum", "super", "torch", "transpose", "view", "weight"], "janus/modeling_janus.py:JanusVQVAEResnetBlock": ["Conv2d", "Dropout", "False", "GroupNorm", "ModelVQVAEResnetBlock", "Module", "None", "True", "__init__", "affine", "class", "config", "conv1", "conv2", "conv_shortcut", "def", "dropout", "else", "eps", "forward", "hidden_states", "if", "in_channels", "is", "kernel_size", "nin_shortcut", "nn", "norm1", "norm2", "num_channels", "num_groups", "out_channels", "padding", "residual", "return", "self", "sigmoid", "stride", "super", "torch", "use_conv_shortcut"], "janus/modeling_janus.py:JanusVQVAEAttnBlock": ["Conv2d", "F", "GroupNorm", "ModelVQVAEAttnBlock", "Module", "True", "__init__", "affine", "attn_output", "attn_weights", "batch_size", "bmm", "channels", "class", "def", "dim", "eps", "forward", "height", "hidden_states", "in_channels", "int", "k", "kernel_size", "key_states", "nn", "norm", "num_channels", "num_groups", "padding", "permute", "proj_out", "q", "query_states", "reshape", "residual", "return", "self", "shape", "softmax", "stride", "super", "torch", "v", "value_states", "width"], "janus/modeling_janus.py:JanusVQVAEConvDownsample": ["Conv2d", "F", "ModelVQVAEConvDownsample", "Module", "__init__", "class", "constant", "conv", "def", "forward", "hidden_states", "in_channels", "kernel_size", "mode", "nn", "pad", "padding", "return", "self", "stride", "super", "value"], "janus/modeling_janus.py:JanusVQVAEConvUpsample": ["Conv2d", "F", "ModelVQVAEConvUpsample", "Module", "__init__", "class", "conv", "def", "forward", "hidden_states", "in_channels", "interpolate", "kernel_size", "mode", "nearest", "nn", "padding", "return", "scale_factor", "self", "stride", "super", "torch"], "janus/modeling_janus.py:JanusVQVAEMidBlock": ["ModelVQVAEAttnBlock", "ModelVQVAEConfig", "ModelVQVAEMidBlock", "ModelVQVAEResnetBlock", "Module", "Tensor", "__init__", "attn_1", "block_1", "block_2", "channels", "class", "config", "def", "forward", "hidden_states", "in_channels", "int", "nn", "out_channels", "return", "self", "super", "torch"], "janus/modeling_janus.py:JanusVQVAEEncoder": ["Conv2d", "GroupNorm", "LongTensor", "ModelVQVAEAttnBlock", "ModelVQVAEConvDownsample", "ModelVQVAEEncoder", "ModelVQVAEMidBlock", "ModelVQVAEResnetBlock", "Module", "ModuleList", "True", "__init__", "affine", "append", "attn", "base_channels", "block", "block_in", "block_out", "channel_multiplier", "class", "config", "conv_in", "conv_out", "def", "double_latent", "down", "downsample", "else", "eps", "for", "forward", "hidden_state", "hidden_states", "i_block", "i_level", "if", "in", "in_channel_multiplier", "in_channels", "kernel_size", "last_hidden_state", "latent_channels", "len", "mid", "nn", "norm_out", "num_channels", "num_groups", "num_res_blocks", "num_resolutions", "out_channels", "padding", "pixel_values", "range", "return", "self", "sigmoid", "stride", "super", "torch", "tuple"], "janus/modeling_janus.py:JanusVQVAEDecoder": ["Conv2d", "FloatTensor", "GroupNorm", "ModelVQVAEAttnBlock", "ModelVQVAEConvUpsample", "ModelVQVAEDecoder", "ModelVQVAEMidBlock", "ModelVQVAEResnetBlock", "Module", "ModuleList", "True", "__init__", "affine", "append", "attn", "base_channels", "block", "block_in", "block_out", "channel_multiplier", "class", "config", "conv_in", "conv_out", "def", "eps", "for", "forward", "hidden_state", "i_block", "i_level", "if", "in", "in_channels", "kernel_size", "latent_channels", "len", "mid", "nn", "norm_out", "num_channels", "num_groups", "num_res_blocks", "num_resolutions", "out_channels", "padding", "range", "return", "reversed", "self", "sigmoid", "stride", "super", "torch", "up", "upsample"], "janus/modeling_janus.py:JanusVQVAE": ["Conv2d", "Expected", "False", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelVQVAE", "ModelVQVAEAttnBlock", "ModelVQVAEConfig", "ModelVQVAEDecoder", "ModelVQVAEEncoder", "ModelVQVAEOutput", "ModelVQVAEResnetBlock", "ModelVQVAEVectorQuantizer", "ValueError", "__init__", "_no_split_modules", "auto_docstring", "batch_size", "but", "can_return_tuple", "class", "codebook_entry", "config", "decode", "decoded_pixel_values", "decoder", "def", "emb_loss", "embed_dim", "embedding_loss", "encode", "encoder", "eval", "f", "forward", "get_codebook_entry", "got", "gradient_checkpointing", "have", "hidden_states", "if", "image_tokens", "indices", "latent_channels", "main_input_name", "nn", "pixel_values", "post_init", "post_quant_conv", "quant", "quant_conv", "quant_state_dims", "quantize", "raise", "return", "self", "shape", "super", "to", "torch", "tuple", "view"], "janus/modeling_janus.py:JanusVQVAEAlignerMLP": ["ACT2FN", "Linear", "ModelVQVAEAlignerMLP", "ModelVQVAEConfig", "Module", "ModuleList", "_", "__init__", "activation_fn", "class", "config", "def", "embed_dim", "fc1", "for", "forward", "hidden_act", "hidden_layers", "hidden_states", "in", "layer", "nn", "num_hidden_layers", "projection_dim", "range", "return", "self", "super"], "janus/modeling_janus.py:JanusVQVAEHead": ["ACT2FN", "Linear", "ModelVQVAEConfig", "ModelVQVAEHead", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "forward", "hidden_act", "hidden_states", "image_token_embed_dim", "nn", "num_embeddings", "proj_out", "projection_dim", "return", "self", "super", "tensor", "torch", "vision_head"], "janus/modeling_janus.py:JanusModel": ["AutoModel", "Cache", "Embedding", "False", "FloatTensor", "Image", "LongTensor", "ModelBaseModelOutputWithPast", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelVQVAE", "ModelVQVAEAlignerMLP", "ModelVQVAEHead", "ModelVisionAlignerMLP", "ModelVisionModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_from_config", "aligner", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cache_position", "can_return_tuple", "cannot", "class", "config", "def", "device", "do", "dtype", "either", "else", "embed_dim", "expand_as", "f", "features", "forward", "from_config", "generation_aligner", "generation_embeddings", "generation_head", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "gradient_checkpointing", "hidden_states", "if", "image", "image_attention_mask", "image_embeds", "image_features", "image_hidden_states", "image_token_id", "input_ids", "inputs_embeds", "int", "is", "kwargs", "language_model", "last_hidden_state", "lm_output", "logits_to_keep", "long", "masked_scatter", "match", "must", "n_image_features", "n_image_tokens", "nn", "not", "num_embeddings", "numel", "one", "past_key_values", "pixel_values", "position_ids", "post_init", "raise", "reshape", "return", "same", "self", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "the", "time", "to", "tokens", "torch", "unsqueeze", "use_cache", "value", "vision_config", "vision_model", "vq_config", "vqmodel"], "janus/modeling_janus.py:JanusForConditionalGeneration": ["CFG", "Cache", "ClassifierFreeGuidanceLogitsProcessor", "Ensure", "Expected", "False", "FloatTensor", "GREEDY_SEARCH", "GenerateDecoderOnlyOutput", "Generation", "GenerationMixin", "GenerationMode", "Got", "Image", "Linear", "LogitsProcessorList", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Passing", "SAMPLE", "Setting", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_can_compile_fullgraph", "_expand_inputs_for_generation", "_get_cache", "_get_initial_cache_position", "_get_logits_processor", "_prepare_model_inputs", "_prepare_special_tokens", "_tied_weights_keys", "_update_model_kwargs_for_generation", "_validate_model_kwargs", "activated", "and", "append", "argmax", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "beam", "bias", "boi_token_id", "bool", "bos_token_id", "but", "by", "cache_implementation", "cache_position", "can_return_tuple", "cat", "class", "clone", "config", "copy", "currently", "de", "decode", "decode_image_tokens", "decoded_image", "decoder_attentions", "decoder_hidden_states", "deepcopy", "def", "default", "device", "dim", "do_sample", "dtype", "else", "embed_tokens", "embeds", "encoder_input_ids", "expand_size", "f", "float", "for", "forward", "generate", "generated_tokens", "generation_aligner", "generation_config", "generation_embeddings", "generation_head", "generation_kwargs", "generation_mode", "get", "get_generation_mode", "get_input_embeddings", "got", "greedy", "guidance_scale", "hidden_size", "hidden_state", "hidden_states", "i", "ids", "if", "image_hidden_states", "image_tokens", "in", "incompatible", "input", "input_ids", "input_ids_seq_length", "input_tokens", "inputs", "inputs_embeds", "int", "is", "isinstance", "kwargs", "kwargs_has_attention_mask", "labels", "language_model", "last_hidden_state", "len", "lm_head", "logger", "logits", "logits_processor", "logits_to_keep", "loss", "loss_function", "mask", "masked_fill_", "max", "max_cache_len", "max_length", "mode", "model", "model_input_name", "model_inputs", "model_kwargs", "multinomial", "next_token", "next_token_scores", "nn", "no_grad", "not", "num_beams", "num_image_tokens", "num_return_sequences", "num_samples", "of", "one", "or", "output_attentions", "output_hidden_states", "output_logits", "output_scores", "outputs", "pad_token_id", "past_key_values", "permute", "pixel_values", "pop", "position_ids", "post_init", "prefix_allowed_tokens_fn", "prepare_embeddings_for_image_generation", "prepare_inputs_for_generation", "probs", "provided", "r", "raise", "range", "raw_logits", "raw_scores", "repeat", "required", "return", "return_dict_in_generate", "sampling", "scores", "search", "self", "seq_len", "sequences", "set_input_embeddings", "setting", "shape", "should", "slice", "slice_indices", "softmax", "squeeze", "static", "super", "supported", "text", "text_config", "that", "to", "torch", "unsqueeze", "update", "use_cache", "validate", "value", "vision_model", "vocab_size", "vqmodel", "warning", "weight", "zeros"], "gpt_bigcode/modeling_gpt_bigcode.py:upcast_masked_softmax": ["Model_masked_softmax", "Tensor", "def", "dim", "dtype", "float", "functional", "input_dtype", "mask", "mask_value", "nn", "return", "scale", "softmax", "softmax_dtype", "to", "torch", "where", "x"], "gpt_bigcode/modeling_gpt_bigcode.py:upcast_softmax": ["Model_softmax", "Tensor", "def", "dim", "dtype", "float", "functional", "input_dtype", "nn", "return", "scale", "softmax", "softmax_dtype", "to", "torch", "x"], "gpt_bigcode/modeling_gpt_bigcode.py:masked_softmax": ["Model_softmax", "Tensor", "def", "dim", "functional", "mask", "mask_value", "nn", "return", "softmax", "torch", "where", "x"], "gpt_bigcode/modeling_gpt_bigcode.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "gpt_bigcode/modeling_gpt_bigcode.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states", "view"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeAttention": ["ALL_ATTENTION_FUNCTIONS", "Attention", "Cache", "Callable", "Dropout", "EncoderDecoderCache", "False", "If", "Linear", "Model", "Module", "Multi", "None", "NotImplementedError", "Optional", "Please", "Query", "Tensor", "True", "Union", "ValueError", "__init__", "_attn_implementation", "and", "as", "attention", "attention_interface", "attention_mask", "attention_softmax_in_fp32", "attn_dropout", "attn_output", "attn_pdrop", "attn_weights", "be", "bool", "by", "c_attn", "c_proj", "cache_position", "class", "config", "contiguous", "cross", "cross_attention", "cross_attention_cache", "curr_past_key_value", "def", "defined", "dim", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "encoder_attention_mask", "encoder_hidden_states", "f", "for", "forward", "get", "got", "hasattr", "have", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "input_shape", "instantiate", "is", "is_causal", "is_cross_attention", "is_updated", "isinstance", "key", "keys", "kv_dim", "kv_heads", "kwargs", "layer_idx", "layer_past", "layers", "make", "mask_value", "multi_query", "must", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "or", "output_attentions", "q_attn", "query", "raise", "reshape", "resid_dropout", "resid_pdrop", "return", "scale_attention_softmax_in_fp32", "scale_attn_weights", "scaling", "self", "self_attention_cache", "shape", "split", "split_size", "super", "supported", "sure", "the", "to", "torch", "training", "transpose", "tuple", "unsqueeze", "update", "use_cache", "used", "value", "values", "view", "weights", "with"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeMLP": ["ACT2FN", "Dropout", "FloatTensor", "Linear", "Model", "Module", "Optional", "__init__", "act", "activation_function", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "resid_pdrop", "return", "self", "super", "torch", "tuple"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeBlock": ["Cache", "Cross", "False", "If", "LayerNorm", "MQA", "Model", "ModelAttention", "ModelMLP", "Module", "None", "NotImplementedError", "Optional", "Tensor", "True", "Union", "ValueError", "__init__", "add_cross_attention", "are", "attention", "attention_mask", "attn", "attn_output", "attn_outputs", "be", "bool", "by", "cache_position", "class", "config", "cross", "cross_attn_outputs", "crossattention", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_hidden_states", "for", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "implemented", "inner_dim", "instantiated", "is", "is_cross_attention", "kwargs", "layer_idx", "layer_norm_epsilon", "layer_past", "layers", "ln_1", "ln_2", "ln_cross_attn", "mlp", "multi_query", "n_inner", "nn", "not", "output_attentions", "outputs", "passed", "raise", "residual", "return", "self", "setting", "super", "to", "torch", "tuple", "use_cache", "with"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodePreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelBlock", "ModelConfig", "ModelMLP", "None", "PreTrainedModel", "True", "__init__", "_init_weights", "_is_hf_initialized", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "c_proj", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "inputs", "is", "isinstance", "kwargs", "math", "mean", "module", "n_layer", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "sqrt", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "Model", "ModelBlock", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Tensor", "Transformers", "Union", "ValueError", "You", "__init__", "_attn_implementation", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "arange", "assert", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bias", "block", "bool", "cache_position", "can_return_tuple", "causal_mask", "class", "config", "create_causal_mask", "cross_attentions", "def", "defined", "deprecated", "device", "dim", "drop", "dtype", "e", "either", "elif", "else", "embd_pdrop", "embed_dim", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eps", "exactly", "flash_attention_2", "for", "forward", "from_legacy_cache", "g", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "h", "has", "have", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "ln_f", "logger", "max_position_embeddings", "max_positions", "multi_query", "must", "n_layer", "new_embeddings", "nn", "not", "num_hidden_layers", "of", "one", "ones", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "pass", "past_key_values", "past_seen_tokens", "persistent", "position_embeds", "position_ids", "post_init", "r", "raise", "range", "register_buffer", "removed", "return", "return_dict", "self", "set_input_embeddings", "shape", "should", "size", "specify", "super", "to", "token_type_embeds", "token_type_ids", "torch", "tril", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v4", "view", "vocab_size", "warning_once", "will", "wpe", "wte"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_function", "n_embd", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "Linear", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "kwargs", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "n_embd", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "sequence_length", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "gpt_bigcode/modeling_gpt_bigcode.py:GPTBigCodeForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "elif", "else", "forward", "hasattr", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "codevector_perplexity", "contrastive_loss", "diversity_loss", "hidden_states", "loss", "projected_quantized_states", "projected_states", "r", "torch", "tuple"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerSamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerRotaryPositionalEmbedding": ["ModelRotaryPositionalEmbedding", "Module", "None", "__init__", "and", "arange", "base", "cached_rotary_positional_embedding", "cached_sequence_length", "cat", "class", "config", "cos", "cos_embeddings", "def", "dim", "dtype", "einsum", "embeddings", "float", "forward", "freqs", "hidden_size", "hidden_states", "i", "if", "ij", "int64", "inv_freq", "is", "j", "nn", "not", "num_attention_heads", "register_buffer", "return", "rotary_embedding_base", "self", "sequence_length", "shape", "sin", "sin_embeddings", "stack", "super", "time_stamps", "torch", "type_as"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerRelPositionalEmbedding": ["ModelRelPositionalEmbedding", "Module", "None", "Tensor", "__init__", "arange", "cat", "class", "config", "cos", "d_model", "def", "device", "dim", "div_term", "dtype", "end_idx", "exp", "expand", "extend_pe", "flip", "float", "forward", "hidden_size", "hidden_states", "if", "int64", "is", "log", "math", "max_len", "max_source_positions", "nn", "not", "or", "pe", "pe_negative", "pe_positive", "position", "relative_position_embeddings", "return", "self", "sin", "size", "start_idx", "super", "tensor", "to", "torch", "unsqueeze", "x", "zeros"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerFeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerConvolutionModule": ["ACT2FN", "BatchNorm1d", "Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelConvolutionModule", "Module", "SAME", "ValueError", "__init__", "a", "activation", "batch_norm", "be", "bias", "class", "config", "conformer_conv_dropout", "conv_depthwise_kernel_size", "def", "depthwise_conv", "dim", "dropout", "for", "forward", "glu", "groups", "hidden_act", "hidden_size", "hidden_states", "if", "kernel_size", "layer_norm", "nn", "number", "odd", "padding", "pointwise_conv1", "pointwise_conv2", "raise", "return", "self", "should", "stride", "super", "transpose"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_apply_relative_embeddings", "_apply_rotary_embedding", "attention_dropout", "attention_mask", "batch_size", "be", "bias", "bool", "cat", "class", "config", "cos", "def", "defined", "device", "dim", "dropout", "dtype", "else", "forward", "has", "head_size", "hidden_size", "hidden_states", "if", "is", "key", "linear_k", "linear_out", "linear_pos", "linear_q", "linear_v", "math", "matmul", "ndim", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "p", "pos_bias_u", "pos_bias_v", "position_embeddings_type", "probs", "proj_relative_position_embeddings", "q_with_bias_u", "q_with_bias_v", "query", "query_key_states", "raise", "relative", "relative_position_embeddings", "reshape", "return", "rotary", "rotated_states", "rotated_states_begin", "rotated_states_end", "scores", "scores_ac", "scores_bd", "scores_bd_padded", "scores_bd_padded_shape", "self", "sequence_length", "shape", "sin", "size", "softmax", "sqrt", "super", "to", "torch", "transpose", "tuple", "value", "value_states", "view", "view_as", "when", "zero_pad", "zeros"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelConvolutionModule", "ModelEncoderLayer", "ModelFeedForward", "ModelSelfAttention", "None", "Optional", "Tensor", "__init__", "attention_dropout", "attention_mask", "attn_weigts", "bool", "class", "config", "conv_module", "def", "dropout", "embed_dim", "ffn1", "ffn1_layer_norm", "ffn2", "ffn2_layer_norm", "final_layer_norm", "forward", "hidden_size", "hidden_states", "nn", "output_attentions", "relative_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_dropout", "self_attn_layer_norm", "super", "torch"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalConvEmbedding", "ModelRelPositionalEmbedding", "ModelRotaryPositionalEmbedding", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "embed_positions", "enumerate", "eps", "expand", "expand_attention_mask", "finfo", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "i", "if", "in", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "min", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings_type", "rand", "range", "relative", "relative_position_embeddings", "repeat", "return", "return_dict", "rotary", "self", "shape", "skip_the_layer", "super", "synced_gpus", "to", "torch", "training", "tuple", "unsqueeze", "v"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerGumbelVectorQuantizer": ["FloatTensor", "Linear", "ModelGumbelVectorQuantizer", "Module", "None", "Parameter", "True", "ValueError", "__init__", "_compute_perplexity", "argmax", "batch_size", "be", "by", "class", "codevector_dim", "codevector_idx", "codevector_probs", "codevector_soft_dist", "codevectors", "codevectors_per_group", "concatenation", "config", "conv_dim", "def", "dim", "divisible", "else", "exp", "expand", "f", "flatten", "float", "for", "forward", "functional", "gumbel_softmax", "hard", "hidden_size", "hidden_states", "if", "is", "log", "marginal_probs", "mask", "mask_extended", "mask_time_indices", "mean", "must", "new_zeros", "nn", "not", "num_codevector_groups", "num_codevectors_per_group", "num_groups", "num_vars", "perplexity", "probs", "raise", "return", "scatter_", "self", "sequence_length", "shape", "softmax", "staticmethod", "sum", "super", "tau", "temperature", "torch", "training", "type_as", "unsqueeze", "view", "weight_proj", "where", "zeros_like"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerAdapter": ["LayerNorm", "Linear", "ModelAdapter", "ModelAdapterLayer", "Module", "ModuleList", "None", "_", "__init__", "and", "class", "config", "def", "else", "for", "forward", "hidden_size", "hidden_states", "if", "in", "is", "layer", "layerdrop", "layerdrop_prob", "layers", "nn", "not", "np", "num_adapter_layers", "or", "output_hidden_size", "proj", "proj_layer_norm", "random", "range", "return", "self", "super", "training", "transpose"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerAdapterLayer": ["Conv1d", "ModelAdapterLayer", "Module", "__init__", "adapter_kernel_size", "adapter_stride", "class", "config", "conv", "def", "dim", "forward", "functional", "glu", "hidden_states", "nn", "output_hidden_size", "padding", "return", "self", "stride", "super"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerPreTrainedModel": ["Conv1d", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelFeatureProjection", "ModelForPreTraining", "ModelGumbelVectorQuantizer", "ModelPositionalConvEmbedding", "ModelPreTrainedModel", "ModelSelfAttention", "None", "Optional", "PreTrainedModel", "True", "Union", "_", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_is_hf_initialized", "a", "adapter_stride", "add_adapter", "arange", "attention_mask", "b", "base_model_prefix", "batch_size", "bias", "bool", "class", "codevectors", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "data", "def", "device", "dim", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "groups", "hasattr", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "module", "nn", "non_padded_lengths", "normal_", "not", "num_adapter_layers", "output_lengths", "pos_bias_u", "pos_bias_v", "project_hid", "project_q", "projection", "range", "reset_parameters", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "supports_gradient_checkpointing", "to", "torch", "uniform_", "weight", "weight_proj", "xavier_uniform_", "zero_", "zeros", "zip"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerModel": ["False", "FloatTensor", "LongTensor", "ModelAdapter", "ModelBaseModelOutput", "ModelConfig", "ModelEncoder", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_freeze_parameters", "_get_feature_vector_attention_mask", "_mask_hidden_states", "adapter", "add_adapter", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "dtype", "elif", "else", "encoder", "encoder_outputs", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "freeze_feature_encoder", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "uniform_", "use_return_dict"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerForPreTraining": ["BoolTensor", "Dropout", "False", "FloatTensor", "Linear", "Model", "ModelConfig", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelGumbelVectorQuantizer", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "add_adapter", "all", "any", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cat", "class", "codevector_dim", "codevector_perplexity", "compute_contrastive_logits", "config", "contrastive_logits_temperature", "contrastive_loss", "cosine_similarity", "cross_entropy", "def", "dim", "diversity_loss", "diversity_loss_weight", "dropout_features", "dtype", "else", "extract_features", "feat_quantizer_dropout", "feature_extractor", "flatten", "float", "forward", "freeze_feature_encoder", "functional", "hidden_size", "hidden_states", "if", "inf", "input_values", "int", "is", "logits", "long", "loss", "mask_time_indices", "neg_is_pos", "negative_features", "negative_quantized_features", "nn", "not", "num_codevector_groups", "num_codevectors", "num_codevectors_per_group", "output_attentions", "output_hidden_states", "outputs", "permute", "post_init", "predicted_features", "proj_codevector_dim", "project_hid", "project_q", "projected_quantized_states", "projected_states", "quantized_features", "quantizer", "r", "reduction", "reshape", "return", "return_dict", "sampled_negative_indices", "self", "sequence_length", "set_gumbel_temperature", "shape", "size", "staticmethod", "sum", "super", "target", "target_features", "temperature", "to", "torch", "transformer_features", "transpose", "tuple", "type_as", "use_return_dict", "view", "weight"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerForCTC": ["CausalLMOutput", "Dropout", "False", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "define", "dim", "does", "dropout", "dtype", "else", "enabled", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "forward", "freeze_feature_encoder", "from_pretrained", "functional", "hasattr", "head", "hidden_size", "hidden_states", "if", "input_lengths", "input_values", "instantiate", "is", "labels", "labels_mask", "language", "lm_head", "log_probs", "log_softmax", "logits", "long", "loss", "masked_select", "max", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "post_init", "r", "raise", "reduction", "return", "return_dict", "s", "self", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "to", "torch", "transpose", "trying", "tuple", "use_return_dict", "values", "vocab_size", "vocabulary", "with", "your", "zero_infinity"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerForSequenceClassification": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Sequence", "SequenceClassifierOutput", "Tensor", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "dim", "does", "else", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "view"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerForAudioFrameClassification": ["Audio", "CrossEntropyLoss", "False", "Linear", "Model", "ModelForAudioFrameClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "TokenClassifierOutput", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "adapters", "add_adapter", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "axis", "bool", "class", "classification", "classifier", "config", "def", "dim", "does", "else", "feature_extractor", "for", "forward", "frame", "freeze_base_model", "freeze_feature_encoder", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "init_weights", "input_values", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "param", "parameters", "r", "raise", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "view"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:AMSoftmaxLoss": ["CrossEntropyLoss", "Model", "Module", "Parameter", "True", "__init__", "bool", "class", "cos_theta", "def", "dim", "flatten", "forward", "functional", "hidden_states", "input_dim", "labels", "logits", "loss", "margin", "mm", "nn", "normalize", "num_labels", "one_hot", "onehot", "psi", "randn", "requires_grad", "return", "scale", "self", "super", "torch", "weight", "where"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:TDNNLayer": ["Detected", "Linear", "LoRA", "LoraLayer", "Model", "Module", "ReLU", "Tensor", "You", "__init__", "activation", "applied", "be", "bias", "class", "config", "conv1d", "def", "dilation", "due", "else", "exclude", "forward", "from", "functional", "hidden_states", "if", "in_conv_dim", "is_peft_available", "isinstance", "kernel", "kernel_size", "layer_id", "modules", "nn", "on", "optimization", "out_conv_dim", "return", "s", "self", "should", "super", "t", "target", "tdnn_dilation", "tdnn_dim", "tdnn_kernel", "to", "torch", "transpose", "view", "warn", "warnings", "weight", "weights", "won"], "wav2vec2_conformer/modeling_wav2vec2_conformer.py:Wav2Vec2ConformerForXVector": ["AMSoftmaxLoss", "False", "Linear", "LongTensor", "Model", "ModelForXVector", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "TDNNLayer", "Tensor", "True", "Union", "XVectorOutput", "_HIDDEN_STATES_START_POSITION", "__init__", "_conv_out_length", "_freeze_parameters", "_get_feat_extract_output_lengths", "_get_tdnn_output_lengths", "append", "attention_mask", "attentions", "auto_docstring", "bool", "cat", "class", "classifier", "config", "def", "dim", "else", "embeddings", "enumerate", "feat_extract_output_lengths", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "functional", "hidden_size", "hidden_states", "i", "if", "in", "init_weights", "input_length", "input_lengths", "input_values", "int", "is", "kernel_size", "labels", "layer_weights", "len", "length", "logits", "loss", "mean", "mean_features", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "objective", "ones", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "parameters", "projector", "r", "range", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "statistic_pooling", "std", "std_features", "stride", "sum", "super", "tdnn", "tdnn_dim", "tdnn_kernel", "tdnn_layer", "tdnn_layers", "tdnn_output_lengths", "torch", "tuple", "use_return_dict", "use_weighted_layer_sum", "view", "xvector_output_dim"], "mlcd/modeling_mlcd.py:MLCDMLP": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "mlcd/modeling_mlcd.py:MLCDRotaryEmbedding": ["False", "Model", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "expand", "flatten", "float", "forward", "hpos_ids", "int", "inv_freq", "max", "max_grid_size", "nn", "num_patches_height", "num_patches_width", "outer", "persistent", "pos_ids", "register_buffer", "return", "rotary_pos_emb", "rotary_pos_emb_full", "self", "seq", "stack", "super", "theta", "torch", "unsqueeze", "wpos_ids"], "mlcd/modeling_mlcd.py:MLCDVisionEmbeddings": ["Conv2d", "False", "FloatTensor", "Model", "ModelVisionConfig", "Module", "Parameter", "Tensor", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "dtype", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "mlcd/modeling_mlcd.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "mlcd/modeling_mlcd.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "mlcd/modeling_mlcd.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "mlcd/modeling_mlcd.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "mlcd/modeling_mlcd.py:MLCDAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "Model", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "apply_rotary_pos_emb_vision", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "cos", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "key_states", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "out_proj", "permute", "position_embeddings", "q_proj", "query_states", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "sin", "super", "torch", "training", "tuple", "unsqueeze", "v_proj", "value_states", "view"], "mlcd/modeling_mlcd.py:MLCDEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelMLP", "ModelVisionConfig", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "position_embeddings", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "mlcd/modeling_mlcd.py:MLCDEncoder": ["BaseModelOutput", "False", "FloatTensor", "Model", "ModelLayer", "ModelVisionConfig", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "position_embeddings", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "mlcd/modeling_mlcd.py:MLCDVisionTransformer": ["BaseModelOutputWithPooling", "FloatTensor", "LayerNorm", "Model", "ModelEncoder", "ModelRotaryEmbedding", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Optional", "Parameter", "Union", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "cat", "class", "class_pos_emb", "config", "cos", "def", "device", "dim", "else", "emb", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "num_attention_heads", "num_patches_height", "num_patches_width", "output_attentions", "output_hidden_states", "patch_size", "pixel_values", "pooled_output", "pooler_output", "position_embeddings", "post_layernorm", "pre_layrnorm", "raise", "randn", "return", "return_dict", "rotary_pos_emb", "self", "shape", "sin", "specify", "super", "to", "torch", "tuple", "use_return_dict", "vision_rotary_embedding"], "mlcd/modeling_mlcd.py:MLCDPreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelMLP", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "None", "PreTrainedModel", "True", "_init_weights", "_supports_flash_attn", "_supports_sdpa", "and", "base_model_prefix", "bias", "class", "class_embedding", "class_pos_emb", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_attention_heads", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "pos_emb_std", "q_proj", "self", "std", "supports_gradient_checkpointing", "v_proj", "weight", "zero_"], "mlcd/modeling_mlcd.py:MLCDVisionModel": ["BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "_no_split_modules", "auto_docstring", "bool", "class", "config", "def", "else", "embeddings", "forward", "get_input_embeddings", "if", "is", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "vision_model"], "vits/modeling_vits.py:VitsModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "r", "sequence_lengths", "spectrogram", "torch", "tuple", "waveform"], "vits/modeling_vits.py:VitsTextEncoderOutput": ["FloatTensor", "ModelOutput", "ModelTextEncoderOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "prior_log_variances", "prior_means", "r", "torch", "tuple"], "vits/modeling_vits.py:fused_add_tanh_sigmoid_multiply": ["Model_add_tanh_sigmoid_multiply", "acts", "def", "in_act", "input_a", "input_b", "num_channels", "return", "s_act", "sigmoid", "t_act", "tanh", "torch"], "vits/modeling_vits.py:_unconstrained_rational_quadratic_spline": ["False", "_rational_quadratic_spline", "_unconstrained_rational_quadratic_spline", "constant", "def", "exp", "functional", "inputs", "inside_interval_mask", "log", "log_abs_det", "min_bin_height", "min_bin_width", "min_derivative", "nn", "np", "outputs", "outside_interval_mask", "pad", "return", "reverse", "tail_bound", "torch", "unnormalized_derivatives", "unnormalized_heights", "unnormalized_widths", "zeros_like"], "vits/modeling_vits.py:_rational_quadratic_spline": ["Input", "Minimal", "None", "RuntimeError", "ValueError", "_rational_quadratic_spline", "a", "all", "b", "bin", "bin_idx", "bin_locations", "bins", "c", "constant", "cumheights", "cumsum", "cumwidths", "def", "delta", "denominator", "derivative_numerator", "derivatives", "dim", "discriminant", "domain", "else", "f", "for", "functional", "gather", "height", "heights", "if", "input_bin_widths", "input_cumheights", "input_cumwidths", "input_delta", "input_derivatives", "input_derivatives_plus_one", "input_heights", "inputs", "intermediate1", "intermediate2", "intermediate3", "invalid", "is", "its", "large", "log", "log_abs_det", "lower_bound", "max", "min", "min_bin_height", "min_bin_width", "min_derivative", "mode", "nn", "not", "num_bins", "number", "numerator", "of", "or", "outputs", "pad", "pow", "raise", "return", "reverse", "root", "shape", "softmax", "softplus", "sqrt", "sum", "tail_bound", "the", "theta", "theta_one_minus_theta", "to", "too", "torch", "transform", "unnormalized_derivatives", "unnormalized_heights", "unnormalized_widths", "upper_bound", "value", "width", "widths", "within"], "vits/modeling_vits.py:VitsWaveNet": ["Conv1d", "Dropout", "IntTensor", "ModelConfig", "ModelWaveNet", "Module", "ModuleList", "None", "__init__", "acts", "append", "class", "cond_layer", "cond_offset", "config", "def", "dilation", "dropout", "else", "for", "forward", "fused_add_tanh_sigmoid_multiply", "global_conditioning", "global_states", "hasattr", "hidden_size", "hidden_states", "i", "if", "in", "in_channels", "in_layer", "in_layers", "inputs", "int", "is", "kernel_size", "layer", "name", "nn", "not", "num_channels_tensor", "num_layers", "out_channels", "outputs", "padding", "padding_mask", "parametrizations", "range", "remove_weight_norm", "res_acts", "res_skip_acts", "res_skip_channels", "res_skip_layer", "res_skip_layers", "return", "self", "speaker_embedding_size", "super", "torch", "utils", "wavenet_dilation_rate", "wavenet_dropout", "wavenet_kernel_size", "weight", "weight_norm", "zeros_like"], "vits/modeling_vits.py:VitsPosteriorEncoder": ["Conv1d", "ModelConfig", "ModelPosteriorEncoder", "ModelWaveNet", "Module", "None", "__init__", "class", "config", "conv_pre", "conv_proj", "def", "dim", "exp", "flow_size", "forward", "global_conditioning", "hidden_size", "inputs", "log_stddev", "mean", "nn", "num_layers", "out_channels", "padding_mask", "posterior_encoder_num_wavenet_layers", "randn_like", "return", "sampled", "self", "spectrogram_bins", "split", "stats", "super", "torch", "wavenet"], "vits/modeling_vits.py:HifiGanResidualBlock": ["Conv1d", "ModelGanResidualBlock", "Module", "ModuleList", "_", "__init__", "apply_weight_norm", "channels", "class", "conv1", "conv2", "convs1", "convs2", "def", "dilation", "for", "forward", "functional", "get_padding", "hasattr", "hidden_states", "i", "if", "in", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "nn", "padding", "parametrizations", "range", "remove_weight_norm", "residual", "return", "self", "stride", "super", "utils", "weight_norm", "zip"], "vits/modeling_vits.py:VitsHifiGan": ["Conv1d", "ConvTranspose1d", "False", "FloatTensor", "HifiGanResidualBlock", "ModelConfig", "ModelHifiGan", "Module", "ModuleList", "None", "Optional", "__init__", "append", "apply_weight_norm", "bias", "channels", "class", "cond", "config", "conv_post", "conv_pre", "def", "dilation", "enumerate", "flow_size", "for", "forward", "functional", "global_conditioning", "hasattr", "hidden_states", "i", "if", "in", "is", "j", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "nn", "not", "num_kernels", "num_upsamples", "padding", "parametrizations", "r", "range", "remove_weight_norm", "res_state", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "self", "speaker_embedding_size", "spectrogram", "stride", "super", "tanh", "torch", "upsample_initial_channel", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "utils", "waveform", "weight_norm", "zip"], "vits/modeling_vits.py:VitsResidualCouplingLayer": ["Conv1d", "False", "ModelConfig", "ModelResidualCouplingLayer", "ModelWaveNet", "Module", "None", "__init__", "cat", "class", "config", "conv_post", "conv_pre", "def", "dim", "else", "exp", "first_half", "flow_size", "forward", "global_conditioning", "half_channels", "hidden_size", "hidden_states", "if", "inputs", "log_determinant", "log_stddev", "mean", "nn", "not", "num_layers", "outputs", "padding_mask", "prior_encoder_num_wavenet_layers", "return", "reverse", "second_half", "self", "split", "sum", "super", "torch", "wavenet", "zeros_like"], "vits/modeling_vits.py:VitsResidualCouplingBlock": ["False", "ModelConfig", "ModelResidualCouplingBlock", "ModelResidualCouplingLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "append", "class", "config", "def", "else", "flip", "flow", "flows", "for", "forward", "global_conditioning", "if", "in", "inputs", "nn", "not", "padding_mask", "prior_encoder_num_flows", "range", "return", "reverse", "reversed", "self", "super", "torch"], "vits/modeling_vits.py:VitsDilatedDepthSeparableConv": ["Conv1d", "Dropout", "LayerNorm", "ModelConfig", "ModelDilatedDepthSeparableConv", "Module", "ModuleList", "None", "__init__", "append", "channels", "class", "config", "convs_dilated", "convs_pointwise", "def", "depth_separable_num_layers", "dilation", "dropout", "dropout_rate", "duration_predictor_kernel_size", "for", "forward", "functional", "gelu", "global_conditioning", "groups", "hidden_size", "hidden_states", "i", "if", "in", "in_channels", "inputs", "is", "kernel_size", "nn", "norms_1", "norms_2", "not", "num_layers", "out_channels", "padding", "padding_mask", "range", "return", "self", "super", "transpose"], "vits/modeling_vits.py:VitsConvFlow": ["Conv1d", "False", "ModelConfig", "ModelConvFlow", "ModelDilatedDepthSeparableConv", "Module", "None", "__init__", "_unconstrained_rational_quadratic_spline", "batch_size", "cat", "channels", "class", "config", "conv_dds", "conv_pre", "conv_proj", "def", "depth_separable_channels", "dim", "duration_predictor_flow_bins", "duration_predictor_tail_bound", "else", "filter_channels", "first_half", "forward", "global_conditioning", "half_channels", "hidden_size", "hidden_states", "if", "inputs", "length", "log_abs_det", "log_determinant", "math", "nn", "not", "num_bins", "outputs", "padding_mask", "permute", "reshape", "return", "reverse", "second_half", "self", "shape", "split", "sqrt", "sum", "super", "tail_bound", "torch", "unnormalized_derivatives", "unnormalized_heights", "unnormalized_widths"], "vits/modeling_vits.py:VitsElementwiseAffine": ["False", "ModelConfig", "ModelElementwiseAffine", "Module", "None", "Parameter", "__init__", "channels", "class", "config", "def", "depth_separable_channels", "else", "exp", "forward", "global_conditioning", "if", "inputs", "log_determinant", "log_scale", "nn", "not", "outputs", "padding_mask", "return", "reverse", "self", "sum", "super", "torch", "translate", "zeros"], "vits/modeling_vits.py:VitsStochasticDurationPredictor": ["Conv1d", "False", "ModelConvFlow", "ModelDilatedDepthSeparableConv", "ModelElementwiseAffine", "ModelStochasticDurationPredictor", "Module", "ModuleList", "None", "True", "_", "__init__", "append", "cat", "clamp_min", "class", "cond", "config", "conv_dds", "conv_pre", "conv_proj", "def", "detach", "device", "dim", "dropout_rate", "dtype", "duration_predictor_dropout", "duration_predictor_num_flows", "durations", "else", "embed_dim", "filter_channels", "first_half", "flip", "flow", "flows", "for", "forward", "functional", "global_conditioning", "hidden_size", "hidden_states", "if", "in", "inputs", "is", "latents", "latents_posterior", "list", "log", "log_determinant", "log_determinant_posterior_sum", "log_determinant_sum", "log_duration", "logq", "logsigmoid", "math", "nll", "nn", "noise_scale", "not", "padding_mask", "pi", "post_conv_dds", "post_conv_pre", "post_conv_proj", "post_flows", "randn", "random_posterior", "range", "return", "reverse", "reversed", "second_half", "self", "sigmoid", "size", "speaker_embedding_size", "split", "sum", "super", "to", "torch"], "vits/modeling_vits.py:VitsDurationPredictor": ["Conv1d", "Dropout", "LayerNorm", "ModelDurationPredictor", "Module", "None", "__init__", "class", "cond", "config", "conv_1", "conv_2", "def", "detach", "dropout", "duration_predictor_dropout", "duration_predictor_filter_channels", "duration_predictor_kernel_size", "eps", "filter_channels", "forward", "global_conditioning", "hidden_size", "if", "inputs", "is", "kernel_size", "layer_norm_eps", "nn", "norm_1", "norm_2", "not", "padding", "padding_mask", "proj", "relu", "return", "self", "speaker_embedding_size", "super", "torch", "transpose"], "vits/modeling_vits.py:VitsAttention": ["Attention", "False", "Head", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "_", "__init__", "_absolute_position_to_relative_position", "_get_relative_embeddings", "_relative_position_to_absolute_position", "_shape", "a", "and", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_heads", "be", "bias", "bmm", "bool", "bsz", "but", "by", "class", "config", "contiguous", "def", "dim", "divisible", "dropout", "else", "emb_rel_k", "emb_rel_v", "embed_dim", "f", "for", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_relative_embeddings", "key_states", "key_value_states", "layer", "layer_head_mask", "length", "mask", "matmul", "max", "must", "nn", "not", "num_attention_heads", "num_heads", "of", "out_proj", "output_attentions", "p", "pad", "pad_length", "proj_shape", "q_proj", "query_states", "raise", "randn", "rel_pos_bias", "relative_embeddings", "relative_logits", "relative_weights", "reshape", "return", "scaling", "self", "seq_len", "should", "single", "size", "slice_end_position", "slice_start_position", "softmax", "src_len", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "use_bias", "v_proj", "value_relative_embeddings", "value_states", "view", "weights", "window_size", "x", "x_final", "x_flat"], "vits/modeling_vits.py:VitsFeedForward": ["ACT2FN", "Conv1d", "Dropout", "ModelFeedForward", "Module", "None", "__init__", "act_fn", "activation_dropout", "class", "config", "conv_1", "conv_2", "def", "dropout", "else", "ffn_dim", "ffn_kernel_size", "forward", "functional", "hidden_act", "hidden_size", "hidden_states", "if", "is", "isinstance", "nn", "not", "pad", "pad_left", "pad_right", "padding", "padding_mask", "permute", "return", "self", "str", "super"], "vits/modeling_vits.py:VitsEncoderLayer": ["Dropout", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelFeedForward", "None", "Optional", "Tensor", "__init__", "attention", "attention_mask", "attn_weights", "bool", "class", "config", "def", "dropout", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "layer_norm", "layer_norm_eps", "nn", "output_attentions", "outputs", "padding_mask", "residual", "return", "self", "super", "torch"], "vits/modeling_vits.py:VitsEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "_prepare_4d_attention_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout_probability", "dtype", "else", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "if", "in", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer_outputs", "layerdrop", "layers", "nn", "not", "np", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "padding_mask", "random", "range", "return", "return_dict", "self", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "uniform", "v"], "vits/modeling_vits.py:VitsTextEncoder": ["Conv1d", "Embedding", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelTextEncoder", "ModelTextEncoderOutput", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "attention_mask", "attentions", "bool", "class", "config", "def", "dim", "else", "embed_tokens", "encoder", "encoder_outputs", "flow_size", "forward", "hidden_size", "hidden_states", "if", "input_ids", "kernel_size", "last_hidden_state", "math", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "padding_mask", "prior_log_variances", "prior_means", "project", "return", "return_dict", "self", "split", "sqrt", "stats", "super", "torch", "transpose", "tuple", "vocab_size"], "vits/modeling_vits.py:VitsPreTrainedModel": ["Conv1d", "ConvTranspose1d", "Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelElementwiseAffine", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "a", "b", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "emb_rel_k", "emb_rel_v", "fill_", "groups", "head_dim", "hidden_size", "if", "in_channels", "init", "initializer_range", "input_ids", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "log_scale", "main_input_name", "math", "mean", "module", "nn", "normal_", "not", "num_attention_heads", "padding_idx", "self", "sqrt", "std", "supports_gradient_checkpointing", "translate", "uniform_", "weight", "window_size", "zero_"], "vits/modeling_vits.py:VitsModel": ["Any", "Embedding", "FloatTensor", "Model", "ModelConfig", "ModelDurationPredictor", "ModelHifiGan", "ModelModel", "ModelModelOutput", "ModelPosteriorEncoder", "ModelPreTrainedModel", "ModelResidualCouplingBlock", "ModelStochasticDurationPredictor", "ModelTextEncoder", "None", "NotImplementedError", "Optional", "Set", "Tensor", "Training", "True", "Union", "ValueError", "_", "__init__", "and", "arange", "attention_mask", "attentions", "attn", "attn_mask", "auto_docstring", "batch_size", "bool", "ceil", "clamp_min", "class", "config", "cum_duration", "cumsum", "decoder", "def", "device", "dtype", "duration", "duration_predictor", "else", "embed_speaker", "embed_tokens", "exp", "f", "fill_value", "flow", "forward", "full", "functional", "get_encoder", "hidden_states", "if", "in", "indices", "input_ids", "input_length", "input_padding_mask", "int", "is", "isinstance", "labels", "last_hidden_state", "latents", "length_scale", "log_duration", "long", "mask_dtype", "matmul", "max", "nn", "noise_scale", "noise_scale_duration", "not", "np", "num_speakers", "of", "ones_like", "output_attentions", "output_hidden_states", "output_length", "output_padding_mask", "outputs", "pad", "padded_indices", "padding_mask", "post_init", "posterior_encoder", "predicted_lengths", "prior_latents", "prior_log_variances", "prior_means", "prod", "r", "raise", "randn_like", "range", "return", "return_dict", "reverse", "self", "sequence_lengths", "shape", "size", "speaker_embedding_size", "speaker_embeddings", "speaker_id", "speaking_rate", "spectrogram", "squeeze", "sum", "super", "supported", "text_encoder", "text_encoder_output", "the", "to", "torch", "transpose", "tuple", "unsqueeze", "upsample_rates", "use_return_dict", "use_stochastic_duration_prediction", "valid_indices", "view", "waveform", "weight", "yet"], "encodec/modeling_encodec.py:EncodecOutput": ["FloatTensor", "LongTensor", "ModelOutput", "None", "Optional", "audio_codes", "audio_values", "class", "r", "torch"], "encodec/modeling_encodec.py:EncodecEncoderOutput": ["FloatTensor", "LongTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "audio_codes", "audio_scales", "class", "int", "last_frame_pad_length", "r", "torch"], "encodec/modeling_encodec.py:EncodecDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "audio_values", "class", "r", "torch"], "encodec/modeling_encodec.py:EncodecConv1d": ["Conv1d", "False", "GroupNorm", "ModelConv1d", "Module", "Tensor", "ValueError", "__init__", "_get_extra_padding_for_conv1d", "_pad1d", "and", "be", "been", "causal", "ceil", "class", "config", "conv", "def", "dilation", "dtype", "elif", "else", "end", "extra_pad", "extra_padding", "f", "float", "forward", "functional", "got", "has", "hasattr", "hidden_states", "ideal_length", "if", "in", "in_channels", "initialized", "int", "int64", "kernel_size", "length", "logger", "max", "max_pad", "mode", "must", "n_frames", "nn", "norm", "norm_type", "not", "of", "one", "out_channels", "pad", "pad_mode", "padded", "padding_left", "padding_right", "padding_total", "paddings", "parametrizations", "persistent", "raise", "reflect", "register_buffer", "return", "self", "shape", "staticmethod", "str", "stride", "super", "tensor", "time_group_norm", "to", "torch", "tuple", "use_causal_conv", "utils", "value", "warning", "weight_norm", "with", "zero"], "encodec/modeling_encodec.py:EncodecConvTranspose1d": ["ConvTranspose1d", "GroupNorm", "ModelConvTranspose1d", "Module", "ValueError", "__init__", "be", "causal", "ceil", "class", "config", "conv", "convolutions", "def", "elif", "else", "end", "f", "for", "forward", "got", "hasattr", "hidden_states", "if", "in", "in_channels", "int", "kernel_size", "makes", "math", "must", "nn", "norm", "norm_type", "not", "of", "one", "only", "or", "out_channels", "padding_left", "padding_right", "padding_total", "parametrizations", "raise", "return", "self", "sense", "shape", "stride", "super", "time_group_norm", "trim_right_ratio", "use_causal_conv", "utils", "weight_norm"], "encodec/modeling_encodec.py:EncodecLSTM": ["LSTM", "ModelConfig", "ModelLSTM", "Module", "__init__", "class", "config", "def", "dimension", "forward", "hidden_states", "int", "lstm", "nn", "num_lstm_layers", "permute", "return", "self", "super"], "encodec/modeling_encodec.py:EncodecResnetBlock": ["ELU", "Identity", "ModelConfig", "ModelConv1d", "ModelResnetBlock", "Module", "ModuleList", "Number", "ValueError", "__init__", "block", "class", "compress", "config", "def", "dilation", "dilations", "dim", "else", "enumerate", "for", "forward", "hidden", "hidden_states", "i", "if", "in", "in_chs", "int", "kernel", "kernel_size", "kernel_sizes", "layer", "len", "list", "match", "nn", "number", "of", "out_chs", "raise", "residual", "residual_kernel_size", "return", "self", "shortcut", "should", "sizes", "super", "use_conv_shortcut", "zip"], "encodec/modeling_encodec.py:EncodecEncoder": ["ELU", "ModelConfig", "ModelConv1d", "ModelEncoder", "ModelLSTM", "ModelResnetBlock", "Module", "ModuleList", "__init__", "audio_channels", "class", "config", "current_scale", "def", "dilation_growth_rate", "for", "forward", "hidden_size", "hidden_states", "in", "j", "kernel_size", "last_kernel_size", "layer", "layers", "model", "nn", "num_filters", "num_residual_layers", "range", "ratio", "return", "reversed", "scaling", "self", "stride", "super", "upsampling_ratios"], "encodec/modeling_encodec.py:EncodecDecoder": ["ELU", "ModelConfig", "ModelConv1d", "ModelConvTranspose1d", "ModelDecoder", "ModelLSTM", "ModelResnetBlock", "Module", "ModuleList", "__init__", "audio_channels", "class", "config", "current_scale", "def", "dilation_growth_rate", "for", "forward", "hidden_size", "hidden_states", "in", "int", "j", "kernel_size", "last_kernel_size", "layer", "layers", "len", "model", "nn", "num_filters", "num_residual_layers", "range", "ratio", "return", "scaling", "self", "stride", "super", "upsampling_ratios"], "encodec/modeling_encodec.py:EncodecEuclideanCodebook": ["ModelConfig", "ModelEuclideanCodebook", "Module", "Tensor", "True", "__init__", "class", "clone", "cluster_size", "codebook_dim", "codebook_size", "config", "decode", "def", "dim", "dist", "embed", "embed_avg", "embed_ind", "embedding", "encode", "functional", "hidden_states", "indices", "inited", "keepdim", "max", "nn", "pow", "quantize", "register_buffer", "reshape", "return", "scaled_states", "self", "shape", "sum", "super", "t", "torch", "view", "zeros"], "encodec/modeling_encodec.py:EncodecVectorQuantization": ["ModelConfig", "ModelEuclideanCodebook", "ModelVectorQuantization", "Module", "__init__", "class", "codebook", "config", "decode", "def", "embed_in", "embed_ind", "encode", "hidden_states", "nn", "permute", "quantize", "return", "self", "super"], "encodec/modeling_encodec.py:EncodecResidualVectorQuantizer": ["ModelConfig", "ModelResidualVectorQuantizer", "ModelVectorQuantization", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_indices", "and", "append", "bandwidth", "bw_per_q", "class", "codebook_size", "codes", "config", "decode", "def", "device", "embeddings", "encode", "enumerate", "float", "floor", "for", "frame_rate", "get_num_quantizers_for_bandwidth", "i", "if", "in", "indices", "int", "is", "layer", "layers", "log2", "math", "max", "nn", "not", "num_quantizers", "out_indices", "quantized", "quantized_out", "range", "residual", "return", "self", "stack", "super", "tensor", "torch"], "encodec/modeling_encodec.py:EncodecPreTrainedModel": ["Conv1d", "ConvTranspose1d", "GroupNorm", "LSTM", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedAudioTokenizerBase", "_init_weights", "a", "b", "base_model_prefix", "bias", "class", "config", "constant_", "data", "def", "elif", "fill_", "for", "groups", "if", "in", "in_channels", "init", "input_values", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "main_input_name", "math", "module", "name", "named_parameters", "nn", "not", "param", "reset_parameters", "self", "sqrt", "uniform_", "weight", "xavier_uniform_", "zero_"], "encodec/modeling_encodec.py:EncodecModel": ["BoolTensor", "Duration", "Expected", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelEncoder", "ModelEncoderOutput", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelResidualVectorQuantizer", "None", "Number", "Optional", "RuntimeError", "Select", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "_", "__init__", "_decode_frame", "_encode_frame", "_linear_overlap_add", "a", "abs", "an", "and", "append", "audio", "audio_codes", "audio_scales", "audio_values", "auto_docstring", "bandwidth", "be", "bigger", "bits_per_codebook", "bool", "but", "cannot", "channels", "chunk", "chunk_length", "chunk_length_s", "chunk_stride", "class", "codebook_size", "codes", "config", "decode", "decoded_frames", "decoder", "def", "device", "did", "dim", "doesn", "dtype", "duration", "element", "else", "embeddings", "empty", "encode", "encoded_frame", "encoded_frames", "encoder", "enumerate", "f", "float", "for", "forward", "frame", "frame_length", "frames", "functional", "get_encoder", "got", "i", "if", "in", "input_length", "input_values", "int", "is", "keepdim", "last_frame", "last_frame_pad_length", "len", "length", "linspace", "list", "log2", "longer", "mask", "math", "mean", "min", "minimum", "model", "mono", "must", "nn", "normalize", "not", "of", "offset", "one", "ones_like", "or", "out", "outputs", "pad", "padding_mask", "post_init", "pow", "power", "quantizer", "r", "raise", "range", "return", "return_dict", "sampling_rate", "scale", "scales", "self", "shape", "specified", "specify", "sqrt", "stack", "staticmethod", "stride", "sum", "sum_weight", "super", "support", "t", "target_bandwidths", "than", "the", "time_vec", "torch", "total_size", "transpose", "tuple", "value", "view", "weight", "zero", "zeros", "zip"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LongTensor", "Model", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "Model", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "Model", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLSelfOutput": ["Dropout", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLAttention": ["False", "FloatTensor", "LayerNorm", "Model", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "intermediate", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "layer_norm_eps", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_layer_norm", "set", "super", "torch", "tuple", "union", "value"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLOutput": ["Linear", "Model", "Module", "__init__", "class", "config", "def", "dense", "forward", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLIntermediate": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "LayerNorm", "Model", "ModelAttention", "ModelIntermediate", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_norm_eps", "layer_output", "layers", "model", "nn", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LayerNorm", "Model", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eps", "for", "forward", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_eps", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLMHead", "ModelLayer", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "roberta", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLPooler": ["Linear", "Model", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "Model", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLLMHead": ["LayerNorm", "Linear", "Model", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLClassificationHead": ["Dropout", "Linear", "Model", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "Model", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "RobertaLMHeadModel", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "RobertaForMaskedLM", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "problem_type", "r", "regression", "return", "return_dict", "roberta", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "r", "reshaped_logits", "return", "return_dict", "roberta", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "active_labels", "active_logits", "active_loss", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "ignore_index", "init_weights", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "r", "return", "return_dict", "roberta", "self", "sequence_output", "super", "tensor", "token_type_ids", "torch", "tuple", "type_as", "view", "where"], "xlm_roberta_xl/modeling_xlm_roberta_xl.py:XLMRobertaXLForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "init_weights", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "qa_outputs", "return", "return_dict", "roberta", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "gemma3/modeling_gemma3.py:Gemma3ModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "gemma3/modeling_gemma3.py:Gemma3CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "gemma3/modeling_gemma3.py:Gemma3TextScaledWordEmbedding": ["Embedding", "False", "ModelTextScaledWordEmbedding", "Tensor", "__init__", "class", "def", "dtype", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "persistent", "register_buffer", "return", "self", "super", "tensor", "to", "torch", "weight"], "gemma3/modeling_gemma3.py:Gemma3MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "ModelTextConfig", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_activation", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "gemma3/modeling_gemma3.py:Gemma3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "gemma3/modeling_gemma3.py:Gemma3RotaryEmbedding": ["False", "ModelRotaryEmbedding", "ModelTextConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gemma3/modeling_gemma3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "gemma3/modeling_gemma3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "gemma3/modeling_gemma3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "gemma3/modeling_gemma3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_dim", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softcap", "softmax", "tanh", "to", "torch", "training", "transpose", "tuple", "value", "value_states"], "gemma3/modeling_gemma3.py:Gemma3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelRMSNorm", "ModelTextConfig", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_logit_softcapping", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dim", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_sliding", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_pre_attn_scalar", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "use_bidirectional_attention", "v_proj", "value_states", "version", "view"], "gemma3/modeling_gemma3.py:Gemma3DecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelTextConfig", "None", "Optional", "Tensor", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is_sliding", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_embeddings_global", "position_embeddings_local", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "pre_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "gemma3/modeling_gemma3.py:Gemma3PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMultiModalProjector", "ModelPreTrainedModel", "PreTrainedModel", "RMSNorm", "SiglipEncoderLayer", "SiglipMultiheadAttentionPoolingHead", "SiglipVisionEmbeddings", "True", "__class__", "__name__", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "elif", "hidden_states", "if", "in", "isinstance", "mm_input_projection_weight", "module", "past_key_values", "self", "super", "supports_gradient_checkpointing", "weight", "zero_"], "gemma3/modeling_gemma3.py:_bidirectional_window_overlay": ["Callable", "_bidirectional_window_overlay", "abs", "batch_idx", "bool", "def", "head_idx", "inner_mask", "int", "kv_idx", "q_idx", "return", "sliding_window"], "gemma3/modeling_gemma3.py:Gemma3TextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "False", "FloatTensor", "LongTensor", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTextConfig", "ModelTextModel", "ModelTextScaledWordEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_bidirectional_window_overlay", "all_hidden_states", "all_self_attns", "and", "arange", "args", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "checkpointing", "class", "config", "copy", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "deepcopy", "def", "default", "device", "dict", "dtype", "else", "embed_scale", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "lambda", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "logger", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "or_mask_function", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings_global", "position_embeddings_local", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rope_local_base_freq", "rope_scaling", "rope_theta", "rope_type", "rotary_emb", "rotary_emb_local", "self", "shape", "sliding_attention", "sliding_mask_kwargs", "sliding_window", "specify", "super", "tensor", "torch", "training", "unsqueeze", "use_bidirectional_attention", "use_cache", "vocab_size", "warning_once", "with"], "gemma3/modeling_gemma3.py:Gemma3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "None", "Optional", "Tensor", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "final_logit_softcapping", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "tanh", "torch", "use_cache", "vocab_size", "weight"], "gemma3/modeling_gemma3.py:Gemma3MultiModalProjector": ["AvgPool2d", "ModelConfig", "ModelMultiModalProjector", "ModelRMSNorm", "Module", "Parameter", "Tensor", "_", "__init__", "avg_pool", "batch_size", "class", "config", "contiguous", "def", "eps", "flatten", "forward", "hidden_size", "image_size", "int", "kernel_size", "layer_norm_eps", "matmul", "mm_input_projection_weight", "mm_soft_emb_norm", "mm_tokens_per_image", "nn", "normed_vision_outputs", "patch_size", "patches_per_image", "pooled_vision_outputs", "projected_vision_outputs", "reshape", "reshaped_vision_outputs", "return", "self", "seq_length", "shape", "stride", "super", "text_config", "tokens_per_side", "torch", "transpose", "type_as", "vision_config", "vision_outputs", "zeros"], "gemma3/modeling_gemma3.py:token_type_ids_mask_function": ["Callable", "Model_type_ids", "Model_type_ids_at_kv_idx", "Model_type_ids_mask_function", "None", "Optional", "Tensor", "batch_idx", "bool", "def", "head_idx", "if", "image_group_ids", "image_group_ids_at_kv_idx", "inner_mask", "int", "is", "is_image_block", "kv_idx", "q_idx", "return", "safe_idx", "same_image_block", "shape", "torch", "where"], "gemma3/modeling_gemma3.py:create_causal_mask_mapping": ["Cache", "False", "FloatTensor", "Model_causal_mask_mapping", "Model_masks_for_generate", "None", "Optional", "PretrainedConfig", "Tensor", "ValueError", "a", "and", "as", "attention_mask", "bool", "cache_position", "config", "cumsum", "def", "device", "dict", "dim", "full_like", "functional", "get_text_config", "if", "image_group_ids", "input", "input_embeds", "int", "is", "is_image", "is_initialized", "is_previous_image", "is_training", "kwargs", "mask_kwargs", "may_have_image_input", "model", "new_image_start", "nn", "not", "or", "or_mask_function", "pad", "past_key_values", "pixel_values", "position_ids", "raise", "required", "return", "to", "token_type_ids", "token_type_ids_mask_function", "torch", "training", "value", "when", "where"], "gemma3/modeling_gemma3.py:Gemma3Model": ["AutoModel", "Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "accepts_loss_kwargs", "all", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "causal_mask_mapping", "class", "clone", "config", "create_causal_mask_mapping", "decoder", "def", "device", "dict", "do", "dtype", "else", "exactly", "expand_as", "f", "features", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_seq_length", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_token_id", "input_ids", "inputs_embeds", "is", "is_training", "isinstance", "labels", "language_model", "last_hidden_state", "llm_input_ids", "lm_kwargs", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "past_seen_tokens", "pixel_values", "position_ids", "post_init", "r", "raise", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "token_type_ids", "tokens", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_config", "vision_outputs", "vision_tower", "vocab_size"], "gemma3/modeling_gemma3.py:Gemma3ForConditionalGeneration": ["Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PretrainedConfig", "Tensor", "True", "Union", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "accepts_loss_kwargs", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "contiguous", "create_causal_mask_mapping", "create_masks_for_generate", "decoder", "def", "device", "dict", "else", "flat_labels", "flat_logits", "float", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "in", "input_embeds", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "k", "kwargs", "labels", "language_model", "lm_head", "lm_kwargs", "logits", "logits_to_keep", "loss", "loss_fct", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "shift_attention_mask", "shift_labels", "shift_logits", "slice", "slice_indices", "staticmethod", "super", "text_config", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "v", "value", "view", "vision_tower", "vocab_size", "weight"], "gemma3/modeling_gemma3.py:Gemma3ForSequenceClassification": ["Cache", "Cannot", "False", "FloatTensor", "Linear", "LongTensor", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_checkpoint_conversion_mapping", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "can_return_tuple", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "get_input_embeddings", "handle", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int32", "is", "kwargs", "labels", "language_model", "last_hidden_state", "last_non_pad_token", "logger", "logits", "loss", "loss_function", "may", "model", "multi_modal_projector", "nn", "no", "non_pad_mask", "not", "num_labels", "pad_token_id", "padding", "past_key_values", "pixel_values", "pooled_logits", "position_ids", "post_init", "r", "raise", "return", "score", "self", "set_input_embeddings", "shape", "sizes", "super", "text_config", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer_outputs", "unexpected", "use_cache", "using", "value", "vision_tower", "warning_once", "will", "with"], "gemma3/modeling_gemma3.py:Gemma3TextForSequenceClassification": ["GenericForSequenceClassification", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextForSequenceClassification", "class", "config"], "big_bird/modeling_big_bird.py:BigBirdEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "rescale_embeddings", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "big_bird/modeling_big_bird.py:BigBirdSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "cache_position", "class", "config", "context_layer", "contiguous", "current_states", "def", "deprecate_kwarg", "dim", "dropout", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "functional", "get_seq_length", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "keys", "layer_idx", "layers", "math", "matmul", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "past_key_value", "past_key_values", "permute", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "update", "use_bias", "value", "value_layer", "values", "version", "view"], "big_bird/modeling_big_bird.py:BigBirdBlockSparseAttention": ["Error", "Key", "Linear", "Make", "ModelBlockSparseAttention", "Model_block_sparse_attention", "Module", "None", "Query", "The", "Value", "ValueError", "_", "_Model_block_rand_mask", "_Model_block_rand_mask_with_head", "__init__", "_create_rand_mask_from_inputs", "_get_rand_attn_plan", "_get_single_block_row_attention", "a", "all_head_size", "and", "append", "arange", "are", "array", "attention", "attention_head_size", "attention_probs", "attn_mask_penalty", "attn_probs_view", "attn_weights", "axis", "band_mask", "band_product", "batch_size", "be", "bhkd", "bhlk", "bhlqd", "bhlqk", "bias", "blk_rw_idx", "block", "block_id", "block_size", "blocked_key_matrix", "blocked_query_matrix", "blocked_value_matrix", "blocks", "blq", "bmm", "break", "bsz", "but", "cat", "class", "concatenate", "config", "context_layer", "contiguous", "continue", "curr_r_cnt", "def", "device", "dim", "dimensions", "div", "dtype", "einsum", "elif", "else", "end", "exp_blocked_key_matrix", "exp_blocked_value_matrix", "extend", "f", "first", "first_attn_weights", "first_band_product", "first_context_layer", "first_product", "flatten", "flattened_indices", "flattened_params", "float", "floor", "for", "forward", "from", "from_block_size", "from_blocked_mask", "from_mask", "from_seq_len", "from_seq_length", "from_start_block_id", "functional", "gathered_key", "gathered_value", "global_block_bottom", "global_block_left", "global_block_right", "global_block_top", "h", "heads", "hidden", "hidden_size", "hidden_states", "i", "i1", "i2", "identical", "if", "illegal_blocks", "in", "index", "index_select", "indices", "indices_shift", "inner_band_product", "inp_1", "inp_2", "int", "int32", "is", "key", "key_layer", "last", "last_attn_weights", "last_band_product", "last_context_layer", "last_idx", "last_product", "len", "length", "list", "long", "math", "max_plan_idx", "max_position_embeddings", "max_seqlen", "middle_query_matrix", "middle_seq", "minimum", "multiple", "must", "n_heads", "n_rand_blocks", "ndim", "needs", "new_ones", "nh", "nn", "not", "np", "num_attention_heads", "num_blocks", "num_heads", "num_indices_to_gather", "num_indices_to_pick_from", "num_rand_blocks", "num_random_blocks", "num_windows", "number", "of", "out", "out_flattened", "output_attentions", "p1", "p2", "params", "perm_block", "permutation", "pl_id", "plan", "plan_block_length", "plan_from_length", "plan_idx", "plan_num_rand_blocks", "q_idx", "query", "query_layer", "r", "raise", "rand_attn", "rand_band_product", "rand_mask", "random", "range", "reshape", "return", "right_slice", "rnd_r_cnt", "rounding_mode", "rsqrt_d", "same", "second_attn_weights", "second_context_layer", "second_key_mat", "second_last_attn_weights", "second_last_context_layer", "second_last_key_mat", "second_last_product", "second_last_rand_pad", "second_last_seq_pad", "second_last_value_mat", "second_product", "second_rand_pad", "second_seq_pad", "second_value_mat", "seed", "selected_random_blocks", "self", "seqlen", "sequence", "shape", "shift", "sided", "size", "softmax", "sqrt", "stack", "start", "staticmethod", "sum", "super", "sure", "tensor", "that", "the", "they", "to", "to_block_list", "to_block_size", "to_blocked_mask", "to_end_block_id", "to_mask", "to_seq_len", "to_seq_length", "to_start_block_id", "torch", "torch_bmm_nd", "torch_bmm_nd_transpose", "torch_gather_b2", "training", "transpose", "two", "unsqueeze", "unsqueeze_", "use_bias", "value", "value_layer", "view", "vs", "w1", "w2", "window_block_left", "window_block_right", "zeros", "zip"], "big_bird/modeling_big_bird.py:BigBirdSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "big_bird/modeling_big_bird.py:BigBirdAttention": ["False", "Model", "ModelAttention", "ModelBlockSparseAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "ValueError", "__init__", "a", "as", "attention_mask", "attention_output", "attention_type", "attn_weights", "band_mask", "be", "block_sparse", "but", "cache_position", "can", "cannot", "class", "config", "decoder", "def", "deprecate_kwarg", "dtype", "either", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "eval", "f", "forward", "from_blocked_mask", "from_mask", "head_mask", "hidden_states", "if", "in", "is", "key", "layer_idx", "new_name", "nn", "not", "only", "or", "original_full", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "query", "raise", "return", "seed", "self", "self_outputs", "set", "set_attention_type", "str", "super", "to", "to_blocked_mask", "to_mask", "training", "used", "value", "version", "when"], "big_bird/modeling_big_bird.py:BigBirdIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "big_bird/modeling_big_bird.py:BigBirdOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "big_bird/modeling_big_bird.py:BigBirdLayer": ["False", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "True", "TypeError", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "attention_type", "band_mask", "be", "block_sparse", "blocked_encoder_mask", "but", "by", "cache_position", "can", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "deprecate_kwarg", "either", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "from_blocked_mask", "from_mask", "has", "hasattr", "head_mask", "hidden_states", "if", "in", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_idx", "layer_output", "layers", "model", "new_name", "not", "only", "or", "original_full", "output", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "raise", "return", "seed", "self", "self_attention_outputs", "seq_len_dim", "set", "set_attention_type", "setting", "should", "str", "super", "to", "to_blocked_mask", "to_mask", "used", "value", "version", "with"], "big_bird/modeling_big_bird.py:BigBirdEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Passing", "Setting", "Transformers", "True", "Union", "ValueError", "You", "__init__", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "attention_mask", "attention_type", "attentions", "band_mask", "be", "block_sparse", "blocked_encoder_mask", "but", "cache_position", "can", "checkpointing", "class", "config", "cross_attentions", "def", "deprecated", "e", "either", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "from_mask", "g", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_outputs", "logger", "nn", "not", "num_hidden_layers", "of", "only", "or", "original_full", "output_attentions", "output_hidden_states", "pass", "past_key_values", "raise", "range", "removed", "return", "return_dict", "seed", "self", "set", "set_attention_type", "should", "str", "super", "to", "to_mask", "training", "tuple", "use_cache", "v", "v4", "value", "warning_once", "will", "with"], "big_bird/modeling_big_bird.py:BigBirdPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "big_bird/modeling_big_bird.py:BigBirdLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "big_bird/modeling_big_bird.py:BigBirdOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "big_bird/modeling_big_bird.py:BigBirdOnlyNSPHead": ["Linear", "ModelOnlyNSPHead", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "big_bird/modeling_big_bird.py:BigBirdPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "big_bird/modeling_big_bird.py:BigBirdPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bert", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "big_bird/modeling_big_bird.py:BigBirdForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "big_bird/modeling_big_bird.py:BigBirdForQuestionAnsweringModelOutput": ["FloatTensor", "ModelForQuestionAnsweringModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "end_logits", "hidden_states", "loss", "pooler_output", "r", "start_logits", "torch", "tuple"], "big_bird/modeling_big_bird.py:BigBirdModel": ["Attention", "BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "Changing", "False", "FloatTensor", "Input", "Linear", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Sequence", "Setting", "Tanh", "Tensor", "True", "Union", "ValueError", "When", "You", "_", "__init__", "_pad_to_block_size", "a", "activation", "add_cross_attention", "add_pooling_layer", "additional", "and", "are", "as", "at", "attention", "attention_mask", "attention_type", "attentions", "auto_docstring", "automatically", "band_mask", "batch_size", "be", "blk", "block", "block_size", "block_sparse", "blocked_encoder_mask", "blq", "blqk", "bool", "both", "buffer", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "but", "cache_position", "can", "cannot", "cat", "class", "config", "create_band_mask_from_inputs", "create_masks_for_block_sparse_attn", "cross_attentions", "decoder", "def", "device", "dim", "dtype", "einsum", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "exp_blocked_to_pad", "expand", "extended_attention_mask", "f", "forward", "from", "from_blocked_mask", "from_mask", "functional", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "global", "hasattr", "have", "head_mask", "hidden_size", "hidden_states", "ids", "if", "in", "input_ids", "input_ids_padding", "input_shape", "inputs_embeds", "inputs_embeds_padding", "int", "invert_attention_mask", "is", "is_decoder", "isinstance", "kwargs", "last_hidden_state", "length", "logger", "long", "max_tokens_to_attend", "min", "multiple", "must", "new_full", "nn", "not", "num", "num_hidden_layers", "num_random_blocks", "of", "ones", "only", "or", "original_full", "output_attentions", "output_hidden_states", "pad", "pad_token_id", "padded", "padding_len", "past_key_values", "past_key_values_length", "pooler", "pooler_output", "position_ids", "possible", "post_init", "r", "raise", "return", "return_dict", "same", "self", "seq_len", "seq_length", "sequence", "sequence_length", "sequence_output", "set", "set_attention_type", "set_input_embeddings", "shape", "size", "sliding", "specify", "staticmethod", "str", "super", "the", "then", "time", "to", "to_blocked_mask", "to_mask", "token_type_ids", "tokens", "torch", "tuple", "type", "unsqueeze_", "use_cache", "use_return_dict", "using", "value", "view", "warn_if_padding_and_no_attention_mask", "warning", "warning_once", "while", "with", "word_embeddings", "zeros"], "big_bird/modeling_big_bird.py:BigBirdForPreTraining": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "None", "Optional", "True", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bert", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "loss", "loss_fct", "new_embeddings", "next_sentence_label", "next_sentence_loss", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "big_bird/modeling_big_bird.py:BigBirdForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "The", "Union", "ValueError", "__init__", "_tied_weights_keys", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bert", "bi", "bias", "bool", "cat", "class", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "big_bird/modeling_big_bird.py:BigBirdForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "FloatTensor", "GenerationMixin", "If", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "a", "add", "as", "attention_mask", "attentions", "auto_docstring", "bert", "bias", "bool", "cache_position", "class", "cls", "config", "cross_attentions", "decoder", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "use_return_dict", "vocab_size", "want", "warning", "weight", "you"], "big_bird/modeling_big_bird.py:BigBirdClassificationHead": ["ACT2FN", "Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "x"], "big_bird/modeling_big_bird.py:BigBirdForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "FloatTensor", "LongTensor", "MSELoss", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "big_bird/modeling_big_bird.py:BigBirdForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "big_bird/modeling_big_bird.py:BigBirdForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "big_bird/modeling_big_bird.py:BigBirdForQuestionAnsweringHead": ["Dropout", "Linear", "ModelForQuestionAnsweringHead", "ModelIntermediate", "ModelOutput", "Module", "__init__", "class", "config", "def", "dropout", "encoder_output", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "intermediate", "nn", "num_labels", "output", "qa_outputs", "return", "self", "super"], "big_bird/modeling_big_bird.py:BigBirdForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "ModelForQuestionAnswering", "ModelForQuestionAnsweringHead", "ModelForQuestionAnsweringModelOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "clamp", "class", "config", "contiguous", "def", "device", "dim", "dtype", "else", "end_logits", "end_loss", "end_positions", "eq", "forward", "head_mask", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "int", "is", "len", "logits", "logits_mask", "loss", "loss_fct", "mask", "maxlen", "not", "num_labels", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "pooler_output", "position_ids", "post_init", "prepare_question_mask", "q_lengths", "qa_classifier", "question_lengths", "r", "return", "return_dict", "self", "sep_token_id", "seqlen", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "staticmethod", "super", "to", "token_type_ids", "torch", "total_loss", "tuple", "unsqueeze_", "use_return_dict", "where"], "ovis2/modeling_ovis2.py:Ovis2ModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "ovis2/modeling_ovis2.py:Ovis2CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "ovis2/modeling_ovis2.py:Ovis2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "ovis2/modeling_ovis2.py:Ovis2VisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "ovis2/modeling_ovis2.py:Ovis2VisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "ModelRMSNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "__init__", "arange", "class", "config", "def", "dtype", "embed_dim", "embeddings", "expand", "flatten", "forward", "hidden_size", "image_size", "in_channels", "kernel_size", "nn", "num_channels", "num_patches", "num_positions", "out_channels", "padding", "patch_embedding", "patch_embeds", "patch_size", "persistent", "pixel_values", "position_embedding", "position_ids", "register_buffer", "return", "rms_norm", "rms_norm_eps", "self", "stride", "super", "target_dtype", "to", "torch", "transpose", "valid", "weight"], "ovis2/modeling_ovis2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "ovis2/modeling_ovis2.py:Ovis2VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "qkv_bias", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "ovis2/modeling_ovis2.py:Ovis2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "ovis2/modeling_ovis2.py:Ovis2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "qkv_bias", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "ovis2/modeling_ovis2.py:Ovis2VisionEncoderLayer": ["GradientCheckpointingLayer", "ModelAttention", "ModelMLP", "ModelRMSNorm", "ModelVisionConfig", "ModelVisionEncoderLayer", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention", "attention_mask", "attn_output", "class", "config", "def", "ffn", "forward", "hidden_size", "hidden_states", "kwargs", "mlp_output", "norm_hidden_states", "return", "rms_norm1", "rms_norm2", "rms_norm_eps", "self", "super", "torch"], "ovis2/modeling_ovis2.py:Ovis2VisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "can_return_tuple", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "ovis2/modeling_ovis2.py:Ovis2VisionTransformer": ["BaseModelOutput", "False", "ModelRMSNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionTransformer", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "embeddings", "encoder", "encoder_outputs", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "inputs_embeds", "kwargs", "last_hidden_state", "nn", "pixel_values", "return", "rms_norm", "rms_norm_eps", "self", "super", "torch"], "ovis2/modeling_ovis2.py:Ovis2VisualEmbeddingTable": ["Embedding", "ModelVisualEmbeddingTable", "Tensor", "class", "def", "dtype", "forward", "if", "in", "int16", "int32", "int64", "int8", "long", "matmul", "nn", "return", "self", "super", "torch", "visual_tokens", "weight"], "ovis2/modeling_ovis2.py:Ovis2PreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "ModelVisionAttention", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_cache_class", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "ovis2/modeling_ovis2.py:hard_softmax": ["Model_softmax", "Tensor", "True", "def", "detach", "dim", "index", "int", "keepdim", "legacy_contiguous_format", "logits", "max", "memory_format", "ret", "return", "scatter_", "softmax", "torch", "y_Model", "y_soft", "zeros_like"], "ovis2/modeling_ovis2.py:Ovis2VisionModel": ["False", "FloatTensor", "LayerNorm", "Linear", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Tensor", "Token", "True", "ValueError", "__init__", "a", "be", "bias", "class", "config", "constant", "def", "dim", "elif", "forward", "functional", "gumbel_argmax", "gumbel_softmax", "hard", "hard_softmax", "head_linear", "head_norm", "hidden_dim", "hidden_size", "hidden_stride", "if", "int", "kwargs", "last_hidden_state", "length", "logits", "math", "must", "nn", "num_images", "num_visual_indicator_tokens", "outputs", "pad", "pad_size", "perfect", "permute", "pixel_values", "prob_token", "raise", "reshape", "return", "self", "seq_len", "sequence", "shape", "softmax", "sqrt", "sqrt_l", "square", "st_argmax", "super", "tokenize_function", "torch", "transformer", "tuple", "vocab_size"], "ovis2/modeling_ovis2.py:Ovis2Model": ["AutoModel", "Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelVisionModel", "ModelVisualEmbeddingTable", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_checkpoint_conversion_mapping", "all", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "device", "dim", "do", "dtype", "else", "enumerate", "exactly", "expand_as", "f", "features", "for", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_size", "hidden_states", "i", "if", "image", "image_features", "image_hidden_states", "image_token_id", "img_seq_len", "in", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "language_model", "last_hidden_state", "layout", "logits_to_keep", "long", "mask", "masked_scatter", "match", "must", "n_image_features", "n_image_tokens", "not", "num_visual_indicator_tokens", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "padding_tensor", "past_key_values", "pixel_values", "position_ids", "post_init", "raise", "requires_grad", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "value", "vision_config", "vision_tower", "visual_embeddings_table", "visual_indicator", "visual_indicator_features", "visual_indicator_id", "visual_indicator_token_ids", "visual_vocab_size", "vocab_size", "zeros"], "ovis2/modeling_ovis2.py:Ovis2ForConditionalGeneration": ["AttributeError", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Not", "Optional", "Tensor", "True", "Union", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "for", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "needed", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "raise", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "value", "vision_tower", "vocab_size", "weight"], "convnextv2/modeling_convnextv2.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "convnextv2/modeling_convnextv2.py:ConvNextV2DropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "convnextv2/modeling_convnextv2.py:ConvNextV2GRN": ["FloatTensor", "ModelGRN", "Module", "Parameter", "True", "__init__", "bias", "class", "def", "dim", "forward", "global_features", "hidden_states", "int", "keepdim", "linalg", "mean", "nn", "norm_features", "ord", "return", "self", "super", "torch", "vector_norm", "weight", "zeros"], "convnextv2/modeling_convnextv2.py:ConvNextV2LayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "convnextv2/modeling_convnextv2.py:ConvNextV2Embeddings": ["FloatTensor", "Make", "Model2d", "ModelEmbeddings", "ModelLayerNorm", "Module", "Tensor", "ValueError", "__init__", "channel", "channels_first", "class", "config", "configuration", "data_format", "def", "dimension", "embeddings", "eps", "forward", "hidden_sizes", "if", "in", "kernel_size", "layernorm", "match", "nn", "num_channels", "of", "one", "patch_embeddings", "patch_size", "pixel", "pixel_values", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "values", "with"], "convnextv2/modeling_convnextv2.py:ConvNextV2Layer": ["ACT2FN", "Identity", "Linear", "Model2d", "ModelDropPath", "ModelGRN", "ModelLayer", "ModelLayerNorm", "Module", "Tensor", "__init__", "act", "class", "config", "def", "dim", "drop_path", "dwModel", "else", "eps", "features", "forward", "grn", "groups", "hidden_act", "if", "kernel_size", "layernorm", "nn", "padding", "permute", "pwModel1", "pwModel2", "residual", "return", "self", "super", "torch"], "convnextv2/modeling_convnextv2.py:ConvNextV2Stage": ["Model2d", "ModelLayer", "ModelLayerNorm", "ModelStage", "Module", "ModuleList", "None", "Tensor", "__init__", "channels_first", "class", "config", "data_format", "def", "depth", "dim", "downsampling_layer", "drop_path", "drop_path_rates", "else", "eps", "features", "for", "forward", "if", "in", "in_channels", "j", "kernel_size", "layer", "layers", "nn", "or", "out_channels", "range", "return", "self", "stride", "super", "torch"], "convnextv2/modeling_convnextv2.py:ConvNextV2Encoder": ["BaseModelOutputWithNoAttention", "False", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_hidden_states", "append", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "drop_path_rate", "drop_path_rates", "else", "for", "forward", "hidden_sizes", "hidden_states", "i", "if", "in", "in_channels", "is", "last_hidden_state", "layer_module", "linspace", "nn", "not", "num_stages", "out_channels", "out_chs", "output_hidden_states", "prev_chs", "range", "return", "self", "split", "stage", "stages", "stride", "sum", "super", "tolist", "torch", "x"], "convnextv2/modeling_convnextv2.py:ConvNextV2PreTrainedModel": ["LayerNorm", "Linear", "Model", "Model2d", "ModelConfig", "ModelGRN", "ModelLayer", "ModelLayerNorm", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "convnextv2/modeling_convnextv2.py:ConvNextV2Model": ["BaseModelOutputWithNoAttention", "BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "ValueError", "You", "__init__", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_sizes", "hidden_states", "if", "is", "last_hidden_state", "layer_norm_eps", "layernorm", "mean", "nn", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_init", "raise", "return", "self", "specify", "super", "to", "torch"], "convnextv2/modeling_convnextv2.py:ConvNextV2ForImageClassification": ["BaseModelOutputWithPoolingAndNoAttention", "False", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "__init__", "accepts_loss_kwargs", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "kwargs", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "outputs", "pixel_values", "pooled_logits", "pooled_output", "pooler_output", "post_init", "r", "return", "self", "super", "torch"], "convnextv2/modeling_convnextv2.py:ConvNextV2Backbone": ["BackboneMixin", "BackboneOutput", "BaseModelOutputWithPoolingAndNoAttention", "False", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelLayerNorm", "ModelPreTrainedModel", "ModuleDict", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "_out_features", "append", "auto_docstring", "bool", "can_return_tuple", "channels", "channels_first", "class", "config", "data_format", "def", "else", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_state", "hidden_states", "hidden_states_norms", "if", "in", "is", "nn", "num_channels", "num_features", "out_features", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "self", "stage", "stage_names", "super", "torch", "tuple", "zip"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoVisionEmbeddings": ["Conv2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "dtype", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embedding", "randn", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "view", "weight", "width"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "ModelQFormerEmbeddings", "ModelQFormerMultiHeadAttention", "ModelQFormerSelfOutput", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "blip", "class", "class_embedding", "config", "data", "def", "elif", "factor", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "position_embedding", "query_tokens", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight", "zero_"], "instructblipvideo/modeling_instructblipvideo.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "torch", "training", "transpose", "value"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_attn_implementation", "_shape", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bsz", "by", "cat", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key_states", "kwargs", "mixed_qkv", "must", "nn", "not", "num_attention_heads", "num_heads", "permute", "projection", "q_bias", "qkv", "qkv_bias", "query_states", "raise", "requires_grad", "reshape", "return", "scale", "scaling", "self", "seq_len", "size", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_bias", "value_states", "view", "zeros", "zeros_like"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoVisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "attentions", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "get_input_embeddings", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "main_input_name", "nn", "pixel_values", "pooled_output", "pooler_output", "post_init", "post_layernorm", "raise", "return", "self", "specify", "super", "to", "torch", "tuple"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerMultiHeadAttention": ["Dropout", "Embedding", "False", "Linear", "ModelQFormerMultiHeadAttention", "Module", "None", "Softmax", "The", "TransformersKwargs", "Unpack", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_map", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_probs_dropped", "attention_scores", "attention_scores_dtype", "attn_gradients", "bhld", "bhlr", "bhrd", "class", "config", "context_layer", "contiguous", "d", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_size", "encoder_hidden_states", "forward", "get_attention_map", "get_attn_gradients", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "key", "key_layer", "kwargs", "long", "lrd", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "raise", "register_hook", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "save_attention", "save_attention_map", "save_attn_gradients", "self", "seq_length", "size", "sqrt", "super", "the", "to", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "view", "x"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelQFormerSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerAttention": ["False", "FloatTensor", "ModelQFormerAttention", "ModelQFormerMultiHeadAttention", "ModelQFormerSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_output", "attn_output", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_cross_attention", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "union", "value"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerIntermediate": ["ACT2FN", "Linear", "ModelQFormerIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerOutput": ["Dropout", "LayerNorm", "Linear", "ModelQFormerOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerLayer": ["False", "GradientCheckpointingLayer", "ModelQFormerAttention", "ModelQFormerIntermediate", "ModelQFormerLayer", "ModelQFormerOutput", "None", "TransformersKwargs", "True", "Unpack", "ValueError", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "be", "cat", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_frequency", "crossattention", "def", "device", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "feed_forward_chunk", "feed_forward_chunk_query", "for", "forward", "given", "has_cross_attention", "head_mask", "hidden_states", "if", "intermediate", "intermediate_output", "intermediate_query", "is", "is_cross_attention", "kwargs", "layer_idx", "layer_output", "layer_output_text", "layers", "must", "output", "output_query", "query_attention_output", "query_length", "raise", "return", "self", "seq_len_dim", "shape", "super", "to", "torch"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "False", "ModelQFormerEncoder", "ModelQFormerLayer", "Module", "ModuleList", "None", "TransformersKwargs", "Unpack", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "query_length", "range", "return", "self", "super"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelQFormerEmbeddings", "Module", "None", "__init__", "absolute", "arange", "cat", "class", "clone", "config", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "is", "layer_norm_eps", "layernorm", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "query_embeds", "register_buffer", "return", "self", "seq_length", "size", "super", "to", "torch", "vocab_size", "weight", "word_embeddings"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoQFormerModel": ["BaseModelOutput", "BaseModelOutputWithPoolingAndCrossAttentions", "False", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelQFormerConfig", "ModelQFormerEmbeddings", "ModelQFormerEncoder", "ModelQFormerLayer", "ModelQFormerModel", "ModelQFormerMultiHeadAttention", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "Wrong", "You", "_", "__init__", "_can_record_outputs", "_prune_heads", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "check_model_inputs", "class", "config", "cross_attentions", "crossattention", "def", "device", "dim", "dtype", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "f", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "has_query", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "index", "input_ids", "input_shape", "int", "invert_attention_mask", "is", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "layer_name", "list", "mask", "not", "num_hidden_layers", "ones", "or", "pooled_output", "pooler_output", "position_ids", "post_init", "prune_heads", "query_embeds", "query_length", "r", "raise", "return", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "to", "torch", "tuple", "value", "when", "word_embeddings"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoForConditionalGenerationModelOutput": ["Any", "FloatTensor", "ModelForConditionalGenerationModelOutput", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "in", "k", "keys", "language_model_outputs", "logits", "loss", "not", "qformer_outputs", "r", "return", "self", "to_tuple", "torch", "tuple", "vision_outputs"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoModel": ["AutoModel", "False", "FlashAttentionKwargs", "FloatTensor", "GPU", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGenerationModelOutput", "ModelModel", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "None", "Optional", "Parameter", "Please", "Tensor", "The", "True", "Union", "Unpack", "__init__", "_hf_hook", "_keep_in_fp32_modules", "_no_split_modules", "_preprocess_accelerate", "_tie_weights", "a", "accelerate", "all", "and", "are", "attention_mask", "auto_docstring", "batch_size", "behavior", "blob", "blog", "bool", "can_return_tuple", "cat", "channel", "class", "com", "config", "contains", "creating", "cuda", "decoder", "decoder_attention_mask", "decoder_input_ids", "def", "details", "device", "device_count", "device_map", "dictionary", "dim", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_hidden_states", "environment", "expand", "expand_as", "extend", "for", "forward", "frames", "from_config", "get_input_embeddings", "get_placeholder_mask", "github", "hasattr", "height", "hf_device_map", "hidden_size", "https", "huggingface", "if", "image_attention_mask", "image_embeds", "image_token_id", "in", "input_ids", "inputs_embeds", "interpolate_pos_encoding", "io_same_device", "is", "kwargs", "language_model", "language_model_inputs", "language_model_outputs", "language_projection", "large", "lead", "len", "logger", "long", "main", "main_input_name", "masked_scatter", "may", "md", "models", "more", "multi", "nn", "not", "num_query_tokens", "on", "ones", "ones_like", "output_attentions", "output_hidden_states", "outputs", "pass", "pixel_values", "post_init", "qformer", "qformer_attention_mask", "qformer_config", "qformer_input_ids", "qformer_outputs", "query_attention_mask", "query_embeds", "query_output", "query_outputs", "query_tokens", "r", "refer", "remove", "repeat_interleave", "reshape", "return", "return_dict", "running", "script", "self", "set_input_embeddings", "shape", "shared", "size", "special_image_mask", "super", "tensor", "text_config", "that", "the", "this", "to", "torch", "tuple", "unexpected", "unsqueeze", "use_cache", "use_decoder_only_language_model", "use_return_dict", "using", "value", "video_token_id", "vision_config", "vision_model", "vision_outputs", "warning", "when", "width", "you", "your", "zeros"], "instructblipvideo/modeling_instructblipvideo.py:InstructBlipVideoForConditionalGeneration": ["AutoModelForCausalLM", "AutoModelForSeq2SeqLM", "False", "FloatTensor", "GPU", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelForConditionalGenerationModelOutput", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "Module", "None", "Optional", "Parameter", "Please", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_from_config", "_hf_hook", "_keep_in_fp32_modules", "_no_split_modules", "_preprocess_accelerate", "_tie_weights", "a", "accelerate", "all", "and", "are", "attention_mask", "auto_docstring", "batch_size", "behavior", "blob", "blog", "bool", "bos_token_id", "can_return_tuple", "cat", "channel", "class", "com", "config", "contains", "creating", "cuda", "decoder", "decoder_attention_mask", "decoder_input_ids", "def", "details", "device", "device_count", "device_map", "dictionary", "dim", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_hidden_states", "environment", "expand", "expand_as", "extend", "for", "forward", "frames", "from_config", "generate", "generate_kwargs", "get_decoder", "get_encoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_placeholder_mask", "get_video_features", "github", "hasattr", "height", "hf_device_map", "hidden_size", "https", "huggingface", "if", "image_attention_mask", "image_embeds", "in", "input_ids", "inputs", "inputs_embeds", "interpolate_pos_encoding", "io_same_device", "is", "is_encoder_decoder", "kwargs", "labels", "language_model", "language_model_inputs", "language_model_outputs", "language_projection", "large", "lead", "len", "logger", "logits", "long", "loss", "loss_function", "main", "main_input_name", "masked_scatter", "may", "md", "models", "more", "multi", "new_embeddings", "nn", "no_grad", "not", "num_query_tokens", "on", "ones", "ones_like", "output_attentions", "output_hidden_states", "outputs", "pass", "pixel_values", "post_init", "qformer", "qformer_attention_mask", "qformer_config", "qformer_input_ids", "qformer_outputs", "query_attention_mask", "query_embeds", "query_output", "query_outputs", "query_tokens", "r", "refer", "remove", "repeat", "repeat_interleave", "reshape", "return", "return_dict", "running", "script", "self", "set_input_embeddings", "set_output_embeddings", "shape", "shared", "size", "special_image_mask", "start_tokens", "super", "tensor", "text_config", "that", "the", "this", "to", "to_tuple", "torch", "tuple", "unexpected", "unsqueeze", "use_cache", "use_decoder_only_language_model", "use_return_dict", "using", "value", "video_token_id", "video_token_index", "video_tokens", "vision_config", "vision_model", "vision_outputs", "vocab_size", "warning", "when", "width", "you", "your", "zeros"], "megatron_bert/modeling_megatron_bert.py:MegatronBertEmbeddings": ["Dropout", "Embedding", "False", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "megatron_bert/modeling_megatron_bert.py:MegatronBertSelfAttention": ["Cache", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "True", "ValueError", "_", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bhld", "bhlr", "bhrd", "bool", "cache_position", "class", "config", "context_layer", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_hidden_states", "f", "forward", "functional", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "key", "key_layer", "key_length", "keys", "layer_idx", "layers", "long", "lrd", "math", "matmul", "max_position_embeddings", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "past_key_value", "past_key_values", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "query_length", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "self_attention_cache", "seq_length", "shape", "size", "softmax", "sqrt", "super", "tensor", "the", "to", "torch", "transpose", "tuple", "update", "value", "value_layer", "values", "version", "view"], "megatron_bert/modeling_megatron_bert.py:MegatronBertSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "nn", "residual", "return", "self", "super", "torch"], "megatron_bert/modeling_megatron_bert.py:MegatronBertAttention": ["Cache", "False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "cache_position", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "encoder_hidden_states", "eps", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "key", "layer_idx", "layer_norm_eps", "len", "ln", "ln_outputs", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "version"], "megatron_bert/modeling_megatron_bert.py:MegatronBertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "megatron_bert/modeling_megatron_bert.py:MegatronBertOutput": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "megatron_bert/modeling_megatron_bert.py:MegatronBertLayer": ["AttributeError", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "LayerNorm", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "TypeError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "bool", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "deprecate_kwarg", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_idx", "layer_norm_eps", "layer_output", "layers", "ln", "ln_output", "model", "new_name", "nn", "not", "output", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "raise", "return", "self", "self_attention_outputs", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "version", "with"], "megatron_bert/modeling_megatron_bert.py:MegatronBertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "You", "__init__", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "attention_mask", "attentions", "be", "bool", "cache_position", "checkpointing", "class", "config", "cross_attentions", "def", "deprecated", "e", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eps", "for", "forward", "from_legacy_cache", "g", "gradient", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_eps", "layer_outputs", "ln", "logger", "nn", "not", "num_hidden_layers", "of", "output_attentions", "output_hidden_states", "pass", "past_key_values", "range", "removed", "return", "return_dict", "self", "should", "super", "torch", "training", "tuple", "use_cache", "v", "v4", "warning_once", "will", "with"], "megatron_bert/modeling_megatron_bert.py:MegatronBertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "megatron_bert/modeling_megatron_bert.py:MegatronBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "megatron_bert/modeling_megatron_bert.py:MegatronBertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "megatron_bert/modeling_megatron_bert.py:MegatronBertOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "megatron_bert/modeling_megatron_bert.py:MegatronBertOnlyNSPHead": ["Linear", "ModelOnlyNSPHead", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "megatron_bert/modeling_megatron_bert.py:MegatronBertPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "megatron_bert/modeling_megatron_bert.py:MegatronBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "and", "base_model_prefix", "bert", "bias", "class", "config", "data", "def", "elif", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "megatron_bert/modeling_megatron_bert.py:MegatronBertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "False", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cache_position", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_key_values_length", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForPreTraining": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "None", "Optional", "True", "Union", "__init__", "_tied_weights_keys", "add_binary_head", "and", "attention_mask", "attentions", "auto_docstring", "bert", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "next_sentence_label", "next_sentence_loss", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "view", "vocab_size"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bert", "bias", "bool", "cache_position", "class", "cls", "config", "cross_attentions", "decoder", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "use_return_dict", "vocab_size", "want", "warning", "you"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "The", "Union", "ValueError", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bert", "bi", "bias", "bool", "cat", "class", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "you"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForNextSentencePrediction": ["CrossEntropyLoss", "FloatTensor", "FutureWarning", "LongTensor", "ModelForNextSentencePrediction", "ModelModel", "ModelOnlyNSPHead", "ModelPreTrainedModel", "NextSentencePredictorOutput", "None", "Optional", "The", "Union", "__init__", "a", "and", "argument", "attention_mask", "attentions", "auto_docstring", "be", "bert", "bool", "class", "cls", "config", "def", "deprecated", "else", "forward", "future", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "pop", "position_ids", "post_init", "r", "removed", "return", "return_dict", "self", "seq_relationship_scores", "super", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "version", "view", "warn", "warnings", "will"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "class", "classifier", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "megatron_bert/modeling_megatron_bert.py:MegatronBertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bert", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "ffn_hidden_size", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "return", "self", "super", "up_proj", "x"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashTopkRouter": ["F", "False", "Linear", "ModelTopkRouter", "Module", "__init__", "bias", "class", "classifier", "config", "def", "dim", "e_score_correction_bias", "float32", "forward", "gather", "get_topk_indices", "getattr", "hidden_size", "hidden_states", "k", "linear", "moe_topk", "n_routed_experts", "nn", "no_grad", "or", "register_buffer", "return", "routed_scaling_factor", "router_bias", "router_logits", "scores", "scores_for_choice", "self", "softmax", "sorted", "super", "top_k", "topk", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight", "zero_expert_num", "zeros"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashMoE": ["Identity", "ModelMLP", "ModelMoE", "ModelTopkRouter", "Module", "ModuleList", "Tensor", "_", "__init__", "class", "config", "def", "dtype", "expert", "expert_ffn_hidden_size", "expert_idx", "expert_input", "expert_mask", "expert_output", "expert_weights", "experts", "final_hidden_states", "for", "forward", "functional", "hidden_states", "if", "in", "index_add_", "intermediate_size", "len", "mask", "moe", "n_routed_experts", "nn", "num_classes", "numel", "one_hot", "orig_shape", "permute", "r", "range", "return", "router", "self", "shape", "super", "token_indices", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight_indices", "weighted_output", "where", "zero_expert_num", "zeros_like"], "longcat_flash/modeling_longcat_flash.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "longcat_flash/modeling_longcat_flash.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "longcat_flash/modeling_longcat_flash.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "longcat_flash/modeling_longcat_flash.py:apply_rotary_pos_emb_interleave": ["Model_rotary_pos_emb_interleave", "None", "b", "cos", "d", "def", "h", "k", "k_embed", "position_ids", "q", "q_embed", "r", "reshape", "return", "rotate_half", "s", "shape", "sin", "transpose", "unsqueeze", "unsqueeze_dim", "view"], "longcat_flash/modeling_longcat_flash.py:yarn_get_mscale": ["Model_get_mscale", "def", "if", "log", "math", "mscale", "return", "scale"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashMLA": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "F", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelMLA", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "apply_rotary_pos_emb_interleave", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "cache_kwargs", "cache_position", "cat", "class", "compressed_kv", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dim", "dropout", "eager", "eager_attention_forward", "else", "expand", "factor", "flash_attention_2", "forward", "get", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_pass", "k_rot", "key_shape", "key_states", "kv_a_layernorm", "kv_a_proj_with_mqa", "kv_b_proj", "kv_lora_rank", "kwargs", "layer_idx", "mla_scale_kv_lora", "mla_scale_q_lora", "mscale", "mscale_all_dim", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "pad", "past_key_value", "past_key_values", "position_embeddings", "q_a_layernorm", "q_a_proj", "q_b_proj", "q_lora_rank", "q_pass", "q_proj", "q_rot", "q_states", "qk_head_dim", "qk_nope_head_dim", "qk_rope_head_dim", "query_shape", "query_states", "reshape", "return", "rope_scaling", "rope_theta", "scaling", "scaling_factor", "self", "seq_length", "shape", "sin", "split", "super", "torch", "training", "transpose", "tuple", "update", "v_head_dim", "value_states", "version", "view", "yarn_get_mscale"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashDecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "GradientCheckpointingLayer", "LongTensor", "ModelDecoderLayer", "ModelMLA", "ModelMLP", "ModelMoE", "ModelRMSNorm", "ModuleList", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "eps", "for", "forward", "hidden_size", "hidden_states", "i", "in", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "mlps", "nn", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "shortcut_mlp_output", "super", "torch", "tuple", "use_cache"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashPreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelMLA", "ModelPreTrainedModel", "ModelTopkRouter", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "classifier", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_keys_to_ignore_on_load_unexpected", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "head_dim", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "model", "mtp", "must", "nn", "norm", "not", "num_hidden_layers", "num_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "longcat_flash/modeling_longcat_flash.py:LongcatFlashForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_keys_to_ignore_on_load_unexpected", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "mtp", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "clap/modeling_clap.py:interpolate": ["Model", "None", "batch_size", "classes_num", "def", "hidden_states", "ratio", "repeat", "reshape", "return", "shape", "time_length", "upsampled"], "clap/modeling_clap.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "height", "hidden_states", "num_channels", "permute", "return", "shape", "view", "width"], "clap/modeling_clap.py:window_reverse": ["Model_reverse", "Model_size", "Models", "contiguous", "def", "height", "num_channels", "permute", "return", "shape", "view", "width"], "clap/modeling_clap.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "labels", "len", "logits", "nn", "return", "torch"], "clap/modeling_clap.py:ClapTextModelOutput": ["FloatTensor", "ModelOutput", "ModelTextModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "clap/modeling_clap.py:ClapAudioModelOutput": ["FloatTensor", "ModelAudioModelOutput", "ModelOutput", "None", "Optional", "attentions", "audio_embeds", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "clap/modeling_clap.py:ClapOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "audio_embeds", "audio_model_output", "class", "def", "else", "for", "getattr", "if", "in", "k", "keys", "logits_per_audio", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple"], "clap/modeling_clap.py:ClapDropPath": ["ModelDropPath", "Module", "None", "__init__", "class", "def", "device", "div", "drop_prob", "dtype", "floor_", "forward", "hidden_states", "if", "keep_prob", "ndim", "nn", "not", "or", "output", "rand", "random_tensor", "return", "self", "shape", "super", "torch", "training"], "clap/modeling_clap.py:ClapAudioAFFBlock": ["AdaptiveAvgPool2d", "BatchNorm2d", "Conv2d", "ModelAudioAFFBlock", "ModelAudioConfig", "Module", "ReLU", "Sequential", "Sigmoid", "True", "__init__", "aff_block_r", "attention_input", "channels", "class", "config", "def", "downsize_ratio", "forward", "fused_layer_output", "global_att", "hidden_states", "inplace", "int", "inter_channels", "kernel_size", "local_att", "nn", "output", "padding", "patch_embeds_hidden_size", "r", "residual", "return", "self", "sigmoid", "stride", "super"], "clap/modeling_clap.py:ClapAudioPatchEmbed": ["Conv2d", "Identity", "Input", "LayerNorm", "ModelAudioAFFBlock", "ModelAudioConfig", "ModelAudioPatchEmbed", "Module", "None", "ValueError", "_", "__init__", "and", "audio", "batch_size", "channel_map", "class", "config", "constant", "contiguous", "def", "doesn", "else", "enable_fusion", "enable_patch_layer_norm", "f", "features", "flatten", "flatten_patch_embeds", "forward", "functional", "fusion_model", "fusion_type", "global_hidden_states", "grid_size", "height", "hidden_states", "if", "img_size", "int", "is_longer_idx", "isinstance", "kernel_size", "len", "local_hidden_states", "local_width", "match", "mel_conv2d", "model", "nn", "norm", "num_channels", "num_patches", "or", "output_width", "pad", "padding", "patch_embed_input_channels", "patch_embeds_hidden_size", "patch_size", "patch_stride", "permute", "proj", "raise", "return", "scale_factor", "self", "shape", "size", "spec_size", "stride", "super", "t", "torch", "transpose", "view", "width"], "clap/modeling_clap.py:ClapAudioSelfAttention": ["Dropout", "False", "FloatTensor", "Iterable", "Linear", "ModelAudioSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "The", "ValueError", "__init__", "a", "abc", "all_head_size", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "collections", "config", "context_layer", "contiguous", "coords", "coords_flatten", "coords_h", "coords_w", "def", "dim", "dropout", "else", "f", "flatten", "forward", "functional", "head_mask", "heads", "hidden", "hidden_shape", "hidden_states", "if", "ij", "indexing", "int", "is", "isinstance", "key", "key_layer", "mask_shape", "math", "matmul", "meshgrid", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_channels", "num_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "register_buffer", "relative_coords", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "return", "self", "shape", "size", "softmax", "sqrt", "stack", "sum", "super", "the", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view", "window_size", "zeros"], "clap/modeling_clap.py:ClapAudioSelfOutput": ["Dropout", "Linear", "ModelAudioSelfOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "clap/modeling_clap.py:ClapAudioAttention": ["False", "FloatTensor", "ModelAudioAttention", "ModelAudioSelfAttention", "ModelAudioSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "clap/modeling_clap.py:ClapAudioIntermediate": ["ACT2FN", "Linear", "ModelAudioIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "clap/modeling_clap.py:ClapAudioOutput": ["Dropout", "Linear", "ModelAudioOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "clap/modeling_clap.py:ClapAudioLayer": ["False", "FloatTensor", "Identity", "LayerNorm", "ModelAudioAttention", "ModelAudioIntermediate", "ModelAudioLayer", "ModelAudioOutput", "ModelDropPath", "Module", "None", "Optional", "Tensor", "_", "__init__", "always_partition", "attention", "attention_output", "attention_outputs", "attention_windows", "attn_mask", "batch_size", "bool", "channels", "chunk_size_feed_forward", "class", "config", "contiguous", "count", "def", "device", "dim", "dims", "drop_path", "drop_path_rate", "dtype", "else", "eps", "for", "forward", "functional", "get_attn_mask", "head_mask", "height", "height_pad", "height_slice", "height_slices", "hidden_states", "hidden_states_windows", "if", "img_mask", "in", "input_dimensions", "input_resolution", "int", "intermediate", "is_tracing", "jit", "layer_norm_eps", "layer_output", "layer_outputs", "layernorm_after", "layernorm_before", "mask_windows", "masked_fill", "maybe_pad", "min", "nn", "not", "num_heads", "or", "output", "output_attentions", "pad", "pad_bottom", "pad_right", "pad_values", "pass", "return", "roll", "self", "set_shift_and_window_size", "shape", "shift_size", "shifted_hidden_states", "shifted_windows", "shifts", "shortcut", "size", "slice", "super", "tensor", "torch", "torch_int", "tuple", "unsqueeze", "view", "was_padded", "width", "width_pad", "width_slice", "width_slices", "window_partition", "window_reverse", "window_size", "zeros"], "clap/modeling_clap.py:ClapAudioStage": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAudioLayer", "ModelAudioStage", "ModuleList", "None", "Optional", "Tensor", "__init__", "always_partition", "blocks", "bool", "class", "config", "def", "depth", "dim", "downsample", "drop_path", "drop_path_rate", "else", "enumerate", "for", "forward", "head_mask", "height", "height_downsampled", "hidden_states", "hidden_states_before_downsampling", "i", "if", "in", "input_dimensions", "input_resolution", "int", "is", "layer_head_mask", "layer_module", "layer_outputs", "nn", "norm_layer", "not", "num_heads", "output_attentions", "output_dimensions", "pointing", "range", "return", "self", "shift_size", "stage_outputs", "super", "torch", "tuple", "width", "width_downsampled", "window_size"], "clap/modeling_clap.py:ClapAudioPatchMerging": ["False", "LayerNorm", "Linear", "ModelAudioPatchMerging", "Module", "None", "Tensor", "__init__", "batch_size", "bias", "cat", "class", "def", "dim", "forward", "functional", "height", "if", "input_dimensions", "input_feature", "input_feature_0", "input_feature_1", "input_feature_2", "input_feature_3", "input_resolution", "int", "maybe_pad", "nn", "norm", "norm_layer", "num_channels", "or", "pad", "pad_values", "reduction", "return", "self", "shape", "should_pad", "super", "torch", "tuple", "view", "width"], "clap/modeling_clap.py:ClapAudioEncoder": ["AdaptiveAvgPool1d", "BaseModelOutputWithPooling", "BatchNorm2d", "False", "FloatTensor", "LayerNorm", "ModelAudioEncoder", "ModelAudioModelOutput", "ModelAudioPatchEmbed", "ModelAudioPatchMerging", "ModelAudioStage", "Module", "ModuleList", "None", "Optional", "True", "Union", "ValueError", "_", "__init__", "align_corners", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "always_partition", "and", "attentions", "avgpool", "batch", "batch_norm", "batch_size", "be", "bicubic", "bool", "c_freq_bin", "channels", "class", "config", "contiguous", "cpu", "def", "depth", "depths", "device", "dim", "downsample", "drop_path", "drop_path_rate", "elif", "else", "enable_fusion", "enumerate", "equal", "flatten", "for", "forward", "frames_num", "freq", "freq_length", "freq_ratio", "freq_shape", "functional", "gradient_checkpointing", "grid_size", "head_mask", "hidden_size", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "input", "input_dimensions", "input_features", "input_resolution", "input_resolutions", "int", "interpolate", "is", "is_longer", "is_longer_list", "is_longer_list_idx", "item", "last_hidden_state", "latent_output", "layer_head_mask", "layer_module", "layer_outputs", "layers", "len", "less", "linspace", "mode", "n_channels", "n_frequencies", "n_temp", "nn", "norm", "normalized_input_features", "not", "num_attention_heads", "num_features", "num_heads", "num_layers", "num_mel_bins", "or", "output_attentions", "output_dimensions", "output_hidden_states", "output_hidden_states_before_downsampling", "patch_embed", "patch_embeds_hidden_size", "patch_stride", "permute", "pooler_output", "raise", "range", "reshape", "reshape_mel2img", "reshaped_hidden_state", "return", "return_dict", "self", "shape", "should", "size", "spec_height", "spec_size", "spec_width", "sum", "super", "swin", "temporal_shape", "than", "the", "time", "time_length", "to", "torch", "transpose", "tuple", "v", "view", "wav", "where", "x"], "clap/modeling_clap.py:ClapProjectionLayer": ["ACT2FN", "Linear", "ModelAudioConfig", "ModelProjectionLayer", "ModelTextConfig", "Module", "Union", "__init__", "activation", "class", "config", "def", "forward", "hidden_size", "hidden_states", "linear1", "linear2", "nn", "projection_dim", "projection_hidden_act", "return", "self", "super"], "clap/modeling_clap.py:ClapTextEmbeddings": ["Dropout", "Embedding", "FloatTensor", "LayerNorm", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "True", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "clap/modeling_clap.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "clap/modeling_clap.py:ClapTextSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelTextSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "bool", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_states", "kwargs", "multiple", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "query", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_states", "view"], "clap/modeling_clap.py:ClapTextSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "clap/modeling_clap.py:ClapTextAttention": ["False", "FloatTensor", "ModelTextAttention", "ModelTextSelfAttention", "ModelTextSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "clap/modeling_clap.py:ClapTextIntermediate": ["ACT2FN", "Linear", "ModelTextIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "clap/modeling_clap.py:ClapTextOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "clap/modeling_clap.py:ClapTextLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelTextAttention", "ModelTextIntermediate", "ModelTextLayer", "ModelTextOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "clap/modeling_clap.py:ClapTextEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelTextEncoder", "ModelTextLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "clap/modeling_clap.py:ClapTextPooler": ["Linear", "ModelTextPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "clap/modeling_clap.py:ClapPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "False", "LayerNorm", "Linear", "Model", "ModelAudioSelfAttention", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelTextEmbeddings", "Module", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "factor", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "is", "isinstance", "log", "logit_scale_a", "logit_scale_init_value", "logit_scale_t", "math", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "position_embeddings", "relative_position_bias_table", "self", "std", "supports_gradient_checkpointing", "token_type_embeddings", "weight", "zero_"], "clap/modeling_clap.py:ClapAudioModel": ["BaseModelOutputWithPooling", "BoolTensor", "FloatTensor", "ModelAudioConfig", "ModelAudioEncoder", "ModelAudioModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Union", "__init__", "audio_encoder", "auto_docstring", "bool", "class", "config", "def", "else", "forward", "get_input_embeddings", "if", "input_features", "is", "is_longer", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "patch_embed", "post_init", "proj", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "clap/modeling_clap.py:ClapTextModel": ["BaseModelOutputWithPooling", "BaseModelOutputWithPoolingAndCrossAttentions", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextModel", "ModelTextPooler", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "add_pooling_layer", "and", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "can_return_tuple", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "clap/modeling_clap.py:ClapModel": ["BaseModelOutputWithPooling", "BoolTensor", "F", "FloatTensor", "LongTensor", "ModelAudioConfig", "ModelAudioModel", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelProjectionLayer", "ModelTextConfig", "ModelTextModel", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "Union", "__init__", "attention_mask", "audio_config", "audio_embeds", "audio_features", "audio_loss", "audio_model", "audio_model_output", "audio_outputs", "audio_projection", "auto_docstring", "be", "bool", "but", "can_return_tuple", "caption_loss", "class", "config", "contrastive_loss", "def", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_audio_features", "get_text_features", "if", "input_features", "input_ids", "is", "is_longer", "isinstance", "keepdim", "log", "logit_scale_a", "logit_scale_audio", "logit_scale_init_value", "logit_scale_t", "logit_scale_text", "logits_per_audio", "logits_per_text", "loss", "math", "matmul", "nn", "norm", "normalize", "not", "of", "output_attentions", "output_hidden_states", "p", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "return", "return_dict", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "use_return_dict"], "clap/modeling_clap.py:ClapTextModelWithProjection": ["ModelPreTrainedModel", "ModelProjectionLayer", "ModelTextConfig", "ModelTextModel", "ModelTextModelOutput", "ModelTextModelWithProjection", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "else", "embeddings", "forward", "get_input_embeddings", "hidden_states", "if", "input_ids", "is", "last_hidden_state", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "text_embeds", "text_model", "text_outputs", "text_projection", "torch", "tuple", "use_return_dict", "value", "word_embeddings"], "clap/modeling_clap.py:ClapAudioModelWithProjection": ["BoolTensor", "FloatTensor", "ModelAudioConfig", "ModelAudioModel", "ModelAudioModelOutput", "ModelAudioModelWithProjection", "ModelPreTrainedModel", "ModelProjectionLayer", "Module", "None", "Optional", "True", "Union", "__init__", "attentions", "audio_embeds", "audio_encoder", "audio_model", "audio_outputs", "audio_projection", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "else", "forward", "get_input_embeddings", "hidden_states", "if", "input_features", "is", "is_longer", "last_hidden_state", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "patch_embed", "pooled_output", "pooler_output", "post_init", "proj", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "electra/modeling_electra.py:ElectraEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embedding_size", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "if", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "shape", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "electra/modeling_electra.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "electra/modeling_electra.py:ElectraSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "electra/modeling_electra.py:ElectraCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "electra/modeling_electra.py:ElectraSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "electra/modeling_electra.py:ElectraAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "electra/modeling_electra.py:ElectraIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "electra/modeling_electra.py:ElectraOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "electra/modeling_electra.py:ElectraLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "electra/modeling_electra.py:ElectraEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "electra/modeling_electra.py:ElectraDiscriminatorPredictions": ["Linear", "ModelDiscriminatorPredictions", "Module", "__init__", "activation", "class", "config", "def", "dense", "dense_prediction", "discriminator_hidden_states", "forward", "get_activation", "hidden_act", "hidden_size", "hidden_states", "logits", "nn", "return", "self", "squeeze", "super"], "electra/modeling_electra.py:ElectraGeneratorPredictions": ["LayerNorm", "Linear", "ModelGeneratorPredictions", "Module", "__init__", "activation", "class", "config", "def", "dense", "embedding_size", "eps", "forward", "gelu", "generator_hidden_states", "get_activation", "hidden_size", "hidden_states", "layer_norm_eps", "nn", "return", "self", "super"], "electra/modeling_electra.py:ElectraPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "electra/modeling_electra.py:ElectraForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "electra/modeling_electra.py:ElectraModel": ["BaseModelOutputWithCrossAttentions", "BaseModelOutputWithPastAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "Linear", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embedding_size", "embeddings", "embeddings_project", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "hasattr", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "nn", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "position_ids", "post_init", "prune_heads", "query_length", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "electra/modeling_electra.py:ElectraClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "activation", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "gelu", "get_activation", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "x"], "electra/modeling_electra.py:ElectraSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "ModelConfig", "ModelSequenceSummary", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "electra/modeling_electra.py:ElectraForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "discriminator_hidden_states", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "electra/modeling_electra.py:ElectraForPreTraining": ["BCEWithLogitsLoss", "Model", "ModelDiscriminatorPredictions", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "active_labels", "active_logits", "active_loss", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "config", "def", "discriminator_hidden_states", "discriminator_predictions", "discriminator_sequence_output", "else", "float", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "position_ids", "post_init", "r", "return", "return_dict", "self", "shape", "super", "token_type_ids", "torch", "tuple", "view"], "electra/modeling_electra.py:ElectraForMaskedLM": ["CrossEntropyLoss", "Linear", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelGeneratorPredictions", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "config", "def", "embedding_size", "forward", "generator_hidden_states", "generator_lm_head", "generator_predictions", "generator_sequence_output", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "view", "vocab_size", "weight", "word_embeddings"], "electra/modeling_electra.py:ElectraForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "discriminator_hidden_states", "discriminator_sequence_output", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "token_type_ids", "torch", "tuple", "view"], "electra/modeling_electra.py:ElectraForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelConfig", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "can_return_tuple", "clamp", "class", "config", "config_class", "contiguous", "def", "dim", "discriminator_hidden_states", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "electra/modeling_electra.py:ElectraForMultipleChoice": ["CrossEntropyLoss", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "discriminator_hidden_states", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_output", "sequence_summary", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "electra/modeling_electra.py:ElectraForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "If", "Linear", "Model", "ModelForCausalLM", "ModelGeneratorPredictions", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "as", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "def", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "forward", "generator_lm_head", "generator_predictions", "get_output_embeddings", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "nn", "not", "outputs", "past_key_values", "position_ids", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "glm4v/modeling_glm4v.py:Glm4vRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "glm4v/modeling_glm4v.py:Glm4VisionMlp": ["ACT2FN", "False", "Linear", "ModelisionMlp", "Module", "__init__", "act_fn", "bias", "bool", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "nn", "out_hidden_size", "return", "self", "super", "up_proj"], "glm4v/modeling_glm4v.py:Glm4vVisionPatchEmbed": ["Conv3d", "ModelVisionConfig", "ModelVisionPatchEmbed", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dtype", "embed_dim", "forward", "hidden_size", "hidden_states", "in_channels", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "glm4v/modeling_glm4v.py:Glm4vVisionRotaryEmbedding": ["False", "ModelVisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "glm4v/modeling_glm4v.py:Glm4vVisionPatchMerger": ["ACT2FN", "False", "GELU", "LayerNorm", "Linear", "ModelVisionPatchMerger", "Module", "None", "Tensor", "__init__", "act1", "act_fn", "bias", "bool", "class", "context_dim", "def", "dim", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_state", "int", "nn", "post_projection_norm", "proj", "return", "self", "str", "super", "torch", "up_proj"], "glm4v/modeling_glm4v.py:Glm4vVisionEmbeddings": ["Embedding", "F", "False", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "__init__", "adapted_pos_embed", "adapted_pos_embed_fp32", "align_corners", "arange", "bicubic", "border", "cat", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "empty", "expand", "float32", "for", "forward", "grid", "grid_sample", "h_coords", "hidden_size", "i", "if", "image_shapes", "image_size", "in", "int", "interpolated_embed_fp32", "isinstance", "len", "lengths", "list", "long", "mode", "nn", "norm_h", "norm_w", "not", "num_patches", "num_positions", "orig_size", "orig_size_sq", "padding_mode", "patch_size", "permute", "persistent", "pos_embed_2d", "pos_embed_weight", "position_embedding", "position_ids", "range", "register_buffer", "repeat", "return", "self", "shape", "squeeze", "stack", "super", "target_h", "target_w", "tensor", "to", "torch", "total_seq", "unsqueeze", "view", "w_coords", "weight"], "glm4v/modeling_glm4v.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "glm4v/modeling_glm4v.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "glm4v/modeling_glm4v.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "glm4v/modeling_glm4v.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "glm4v/modeling_glm4v.py:Glm4vVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "glm4v/modeling_glm4v.py:Glm4vVisionBlock": ["False", "GradientCheckpointingLayer", "ModelRMSNorm", "ModelVisionAttention", "ModelVisionBlock", "ModelisionMlp", "None", "Optional", "Tensor", "__init__", "attn", "bias", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "norm1", "norm2", "position_embeddings", "return", "rms_norm_eps", "rotary_pos_emb", "self", "super", "torch", "tuple"], "glm4v/modeling_glm4v.py:Glm4vTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "glm4v/modeling_glm4v.py:rotate_half_llm": ["Model_half_llm", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "glm4v/modeling_glm4v.py:apply_multimodal_rotary_pos_emb": ["Model_multimodal_rotary_pos_emb", "cat", "cos", "def", "dim", "enumerate", "for", "i", "in", "k", "k_embed", "k_pass", "k_rot", "m", "mrope_section", "q", "q_embed", "q_pass", "q_rot", "repeat_interleave", "return", "rotary_dim", "rotate_half_llm", "shape", "sin", "split", "torch", "unsqueeze", "unsqueeze_dim"], "glm4v/modeling_glm4v.py:Glm4vTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelTextAttention", "ModelTextConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "_", "__init__", "_attn_implementation", "apply_multimodal_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bsz", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "mrope_section", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "reshape", "return", "rope_scaling", "scaling", "self", "sin", "size", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "view"], "glm4v/modeling_glm4v.py:Glm4vTextMLP": ["ACT2FN", "False", "FloatTensor", "Linear", "ModelTextMLP", "Module", "__init__", "activation_fn", "bias", "chunk", "class", "config", "def", "dim", "down_proj", "forward", "gate", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch", "up_states"], "glm4v/modeling_glm4v.py:Glm4vTextDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelRMSNorm", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextMLP", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "output_attentions", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_mlp_layernorm", "post_self_attn_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache"], "glm4v/modeling_glm4v.py:Glm4vModelOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "glm4v/modeling_glm4v.py:Glm4vPreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextDecoderLayer", "ModelVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "glm4v/modeling_glm4v.py:Glm4vVisionModel": ["Conv2d", "F", "False", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "ModelVisionPatchEmbed", "ModelVisionPatchMerger", "ModelVisionRotaryEmbedding", "ModuleList", "None", "Tensor", "_", "__init__", "_no_split_modules", "append", "arange", "blk", "blocks", "cat", "class", "config", "context_dim", "cos", "cu_seqlens", "cumsum", "def", "depth", "dim", "downsample", "dtype", "else", "emb", "embeddings", "eps", "expand", "flatten", "for", "forward", "gradient_checkpointing", "grid_thw", "h", "head_dim", "hidden_act", "hidden_size", "hidden_states", "hpos_ids", "if", "image_type_ids", "in", "in_channels", "int32", "intermediate_size", "is_tracing", "jit", "kernel_size", "max", "max_grid_size", "merger", "nn", "num_heads", "out_channels", "out_hidden_size", "pad", "patch_embed", "patch_size", "permute", "pos_ids", "position_embeddings", "post_conv_layernorm", "post_init", "post_layernorm", "range", "repeat", "repeat_interleave", "reshape", "return", "rms_norm_eps", "rot_pos_emb", "rotary_pos_emb", "rotary_pos_emb_full", "self", "seqlens", "shape", "sin", "spatial_merge_size", "stack", "stride", "super", "t", "tolist", "torch", "unsqueeze", "value", "view", "w", "wpos_ids"], "glm4v/modeling_glm4v.py:Glm4vTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelRMSNorm", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "dim", "elif", "else", "embed_tokens", "eps", "exactly", "expand", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "tuple", "use_cache", "view", "vocab_size"], "glm4v/modeling_glm4v.py:Glm4vModel": ["Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelTextDecoderLayer", "ModelTextModel", "ModelVisionBlock", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "_from_config", "_no_split_modules", "accepts_loss_kwargs", "add", "all", "and", "append", "arange", "attention_mask", "attention_mask_tensor", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "cache_position", "can_return_tuple", "cat", "class", "config", "cumsum", "decoder", "def", "delta", "device", "diagonal", "dict", "dim", "dim1", "dim2", "do", "dtype", "elif", "else", "end_idx", "end_index", "enumerate", "exactly", "expand", "expand_as", "f", "features", "finfo", "flatten", "flattened_video_grid_thw", "for", "forward", "full_attention", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_seq_length", "get_video_features", "grid_thw", "group", "groupby", "h", "h_index", "hidden_states", "i", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_index", "image_mask", "image_token_id", "in", "input_ids", "input_token_type", "input_tokens", "input_type_group", "inputs_embeds", "int", "is", "is_floating_point", "is_torchdynamo_compiling", "isinstance", "item", "itertools", "keepdim", "key", "kwargs", "lambda", "language_model", "last_hidden_state", "len", "list", "llm_grid_h", "llm_grid_t", "llm_grid_w", "llm_pos_ids_list", "llm_positions", "long", "masked_fill_", "masked_scatter", "match", "max", "max_position_ids", "min", "modality_type", "mrope_position_deltas", "must", "n_image_tokens", "n_video_tokens", "ndim", "not", "numel", "of", "one", "ones", "ones_like", "or", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prefill_compiled_stage", "prefill_noncompiled_stage", "prod", "r", "raise", "range", "repeat", "repeat_interleave", "repeated_row", "reshape", "return", "rope_deltas", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "special_video_mask", "specify", "split", "split_sizes", "st_idx", "stack", "start_idx", "start_index", "sum", "super", "t", "t_idx", "t_index", "temp_frames_hw", "tensor", "text", "text_config", "text_len", "to", "token", "tokens", "tolist", "torch", "total_input_ids", "tuple", "type", "unsqueeze", "value", "video", "video_check_flg", "video_embeds", "video_end_token_id", "video_features", "video_frame_num", "video_grid_thw", "video_group_index", "video_index", "video_mask", "video_start_token_id", "video_token_id", "view", "vision_config", "visual", "w", "w_index", "x", "zeros"], "glm4v/modeling_glm4v.py:Glm4vCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "glm4v/modeling_glm4v.py:Glm4vForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "If", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_expand_dict_for_generation", "_expand_dict_for_generation_visual", "_expand_inputs_for_generation", "_get_image_nums_and_video_nums", "_repeat_interleave_samples", "_tied_weights_keys", "accepts_loss_kwargs", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "cumsum", "decoder", "def", "defined", "device", "dict", "dict_to_expand", "dim", "dtype", "elif", "else", "encoder_outputs", "expand_size", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "get_video_features", "hidden_size", "hidden_states", "if", "image_counts", "image_grid_thw", "image_nums", "image_start_token_id", "in", "input_ids", "inputs_embeds", "inside_video", "int", "is", "is_encoder_decoder", "is_image", "is_video_end", "is_video_start", "isinstance", "key", "kwargs", "labels", "language_model", "lengths", "list", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "make", "model", "model_inputs", "model_kwargs", "nn", "not", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "prod", "property", "r", "raise", "repeat", "repeat_args", "repeat_interleave", "repeat_times", "result", "return", "rope_deltas", "sample", "samples", "second_per_grid_ts", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "split", "standalone_images", "str", "sum", "super", "sure", "tensor", "text_config", "that", "torch", "tuple", "use_cache", "value", "video_counts", "video_end_token_id", "video_grid_thw", "video_level", "video_nums", "video_start_token_id", "visual", "visual_keys", "vocab_size", "weight", "x"], "exaone4/modeling_exaone4.py:Exaone4RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "exaone4/modeling_exaone4.py:Exaone4RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "exaone4/modeling_exaone4.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "exaone4/modeling_exaone4.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "exaone4/modeling_exaone4.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "exaone4/modeling_exaone4.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "exaone4/modeling_exaone4.py:Exaone4Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_sliding", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "sliding_window_pattern", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "exaone4/modeling_exaone4.py:Exaone4MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "exaone4/modeling_exaone4.py:Exaone4DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "exaone4/modeling_exaone4.py:Exaone4PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "config_class", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "exaone4/modeling_exaone4.py:Exaone4Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "enumerate", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_type", "layer_types", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "tuple", "unsqueeze", "use_cache", "vocab_size"], "exaone4/modeling_exaone4.py:Exaone4ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "exaone4/modeling_exaone4.py:Exaone4ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "exaone4/modeling_exaone4.py:Exaone4ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "exaone4/modeling_exaone4.py:Exaone4ForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "donut/modeling_donut_swin.py:DonutSwinEncoderOutput": ["FloatTensor", "ModelOutput", "ModelSwinEncoderOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "reshaped_hidden_states", "torch", "tuple"], "donut/modeling_donut_swin.py:DonutSwinModelOutput": ["FloatTensor", "ModelOutput", "ModelSwinModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "reshaped_hidden_states", "torch", "tuple"], "donut/modeling_donut_swin.py:DonutSwinImageClassifierOutput": ["FloatTensor", "ModelOutput", "ModelSwinImageClassifierOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "donut/modeling_donut_swin.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "height", "input_feature", "num_channels", "permute", "return", "shape", "view", "width"], "donut/modeling_donut_swin.py:window_reverse": ["Model_reverse", "Model_size", "Models", "contiguous", "def", "height", "num_channels", "permute", "return", "shape", "view", "width"], "donut/modeling_donut_swin.py:DonutSwinEmbeddings": ["BoolTensor", "Dropout", "False", "FloatTensor", "LayerNorm", "ModelSwinEmbeddings", "ModelSwinPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "config", "def", "dim", "dropout", "else", "embed_dim", "embeddings", "expand", "forward", "functional", "grid_size", "height", "hidden_dropout_prob", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "norm", "not", "num_channels", "num_patches", "num_positions", "output_dimensions", "patch_embeddings", "patch_grid", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "tuple", "type_as", "unsqueeze", "use_absolute_embeddings", "use_mask_token", "view", "width", "zeros"], "donut/modeling_donut_swin.py:DonutSwinPatchEmbeddings": ["Conv2d", "FloatTensor", "Iterable", "ModelSwinPatchEmbeddings", "Module", "Optional", "Tensor", "_", "__init__", "abc", "class", "collections", "config", "def", "else", "embed_dim", "embeddings", "flatten", "forward", "functional", "grid_size", "height", "hidden_size", "if", "image_size", "int", "isinstance", "kernel_size", "maybe_pad", "nn", "num_channels", "num_patches", "output_dimensions", "pad", "pad_values", "patch_size", "pixel_values", "projection", "return", "self", "shape", "stride", "super", "torch", "transpose", "tuple", "width"], "donut/modeling_donut_swin.py:DonutSwinPatchMerging": ["False", "LayerNorm", "Linear", "ModelSwinPatchMerging", "Module", "None", "Tensor", "__init__", "batch_size", "bias", "cat", "class", "def", "dim", "forward", "functional", "height", "if", "input_dimensions", "input_feature", "input_feature_0", "input_feature_1", "input_feature_2", "input_feature_3", "input_resolution", "int", "maybe_pad", "nn", "norm", "norm_layer", "num_channels", "or", "pad", "pad_values", "reduction", "return", "self", "shape", "should_pad", "super", "torch", "tuple", "view", "width"], "donut/modeling_donut_swin.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "donut/modeling_donut_swin.py:DonutSwinDropPath": ["ModelSwinDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "donut/modeling_donut_swin.py:DonutSwinSelfAttention": ["Dropout", "False", "FloatTensor", "Iterable", "Linear", "ModelSwinSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "The", "ValueError", "__init__", "a", "abc", "all_head_size", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "collections", "config", "context_layer", "contiguous", "coords", "coords_flatten", "coords_h", "coords_w", "def", "dim", "dropout", "else", "f", "flatten", "forward", "functional", "head_mask", "heads", "hidden", "hidden_shape", "hidden_states", "if", "ij", "indexing", "int", "is", "isinstance", "key", "key_layer", "mask_shape", "math", "matmul", "meshgrid", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_channels", "num_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "register_buffer", "relative_coords", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "return", "self", "shape", "size", "softmax", "sqrt", "stack", "sum", "super", "the", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view", "window_size", "zeros"], "donut/modeling_donut_swin.py:DonutSwinSelfOutput": ["Dropout", "Linear", "ModelSwinSelfOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "donut/modeling_donut_swin.py:DonutSwinAttention": ["False", "FloatTensor", "ModelSwinAttention", "ModelSwinSelfAttention", "ModelSwinSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "donut/modeling_donut_swin.py:DonutSwinIntermediate": ["ACT2FN", "Linear", "ModelSwinIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "donut/modeling_donut_swin.py:DonutSwinOutput": ["Dropout", "Linear", "ModelSwinOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "donut/modeling_donut_swin.py:DonutSwinLayer": ["False", "FloatTensor", "Identity", "LayerNorm", "ModelSwinAttention", "ModelSwinDropPath", "ModelSwinIntermediate", "ModelSwinLayer", "ModelSwinOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "always_partition", "attention", "attention_output", "attention_outputs", "attention_windows", "attn_mask", "batch_size", "bool", "channels", "chunk_size_feed_forward", "class", "config", "contiguous", "count", "def", "device", "dim", "dims", "drop_path", "drop_path_rate", "dtype", "else", "eps", "for", "forward", "functional", "get_attn_mask", "head_mask", "height", "height_pad", "height_slice", "height_slices", "hidden_states", "hidden_states_windows", "if", "img_mask", "in", "input_dimensions", "input_resolution", "int", "intermediate", "is_tracing", "jit", "layer_norm_eps", "layer_output", "layer_outputs", "layernorm_after", "layernorm_before", "mask_windows", "masked_fill", "maybe_pad", "min", "nn", "not", "num_heads", "or", "output", "output_attentions", "pad", "pad_bottom", "pad_right", "pad_values", "pass", "return", "roll", "self", "set_shift_and_window_size", "shape", "shift_size", "shifted_hidden_states", "shifted_windows", "shifts", "shortcut", "size", "slice", "super", "tensor", "torch", "torch_int", "tuple", "unsqueeze", "view", "was_padded", "width", "width_pad", "width_slice", "width_slices", "window_partition", "window_reverse", "window_size", "zeros"], "donut/modeling_donut_swin.py:DonutSwinStage": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelSwinLayer", "ModelSwinStage", "ModuleList", "None", "Optional", "Tensor", "__init__", "always_partition", "blocks", "bool", "class", "config", "def", "depth", "dim", "downsample", "drop_path", "drop_path_rate", "else", "enumerate", "for", "forward", "head_mask", "height", "height_downsampled", "hidden_states", "hidden_states_before_downsampling", "i", "if", "in", "input_dimensions", "input_resolution", "int", "is", "layer_head_mask", "layer_module", "layer_outputs", "nn", "norm_layer", "not", "num_heads", "output_attentions", "output_dimensions", "pointing", "range", "return", "self", "shift_size", "stage_outputs", "super", "torch", "tuple", "width", "width_downsampled", "window_size"], "donut/modeling_donut_swin.py:DonutSwinEncoder": ["False", "FloatTensor", "ModelSwinEncoder", "ModelSwinEncoderOutput", "ModelSwinPatchMerging", "ModelSwinStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "always_partition", "and", "attentions", "batch_size", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "dim", "downsample", "dpr", "drop_path", "drop_path_rate", "elif", "else", "embed_dim", "enumerate", "for", "forward", "gradient_checkpointing", "grid_size", "head_mask", "hidden_size", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "input_dimensions", "input_resolution", "int", "is", "item", "last_hidden_state", "layer_head_mask", "layer_module", "layer_outputs", "layers", "len", "linspace", "nn", "not", "num_heads", "num_layers", "output_attentions", "output_dimensions", "output_hidden_states", "output_hidden_states_before_downsampling", "permute", "range", "reshaped_hidden_state", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "sum", "super", "torch", "tuple", "v", "view", "x"], "donut/modeling_donut_swin.py:DonutSwinPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelSwinConfig", "ModelSwinEmbeddings", "ModelSwinPreTrainedModel", "ModelSwinSelfAttention", "ModelSwinStage", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "relative_position_bias_table", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "donut/modeling_donut_swin.py:DonutSwinModel": ["AdaptiveAvgPool1d", "BoolTensor", "False", "FloatTensor", "ModelSwinEmbeddings", "ModelSwinEncoder", "ModelSwinModel", "ModelSwinModelOutput", "ModelSwinPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "flatten", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_dimensions", "int", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "len", "nn", "not", "num_features", "num_layers", "output", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_grid", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "tuple", "use_mask_token", "use_return_dict"], "donut/modeling_donut_swin.py:DonutSwinForImageClassification": ["False", "FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelSwinForImageClassification", "ModelSwinImageClassifierOutput", "ModelSwinModel", "ModelSwinPreTrainedModel", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_features", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "post_init", "r", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "pegasus/modeling_pegasus.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "pegasus/modeling_pegasus.py:PegasusSinusoidalPositionalEmbedding": ["Embedding", "False", "FloatTensor", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_init_weight", "arange", "array", "bsz", "class", "cos", "def", "device", "dim", "dtype", "else", "embedding_dim", "empty", "for", "forward", "if", "in", "input_ids_shape", "int", "is", "j", "long", "n_pos", "nn", "no_grad", "np", "num_positions", "out", "padding_idx", "past_key_values_length", "pos", "position_enc", "position_ids", "power", "range", "requires_grad", "return", "self", "sentinel", "seq_len", "shape", "sin", "super", "torch", "weight"], "pegasus/modeling_pegasus.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "pegasus/modeling_pegasus.py:PegasusAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "pegasus/modeling_pegasus.py:PegasusEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "pegasus/modeling_pegasus.py:PegasusDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "pegasus/modeling_pegasus.py:PegasusPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weight", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "def", "device", "diagonal", "dim", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "nn", "normal_", "not", "npu", "ones", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "pegasus/modeling_pegasus.py:PegasusEncoder": ["BaseModelOutput", "Embedding", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Setting", "The", "True", "ValueError", "You", "_", "__init__", "_init_weight", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "get_position_embeddings", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "info", "input_ids", "input_shape", "inputs_embeds", "int", "is", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "math", "max_position_embeddings", "max_source_positions", "new_num_position_embeddings", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "resize_position_embeddings", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "pegasus/modeling_pegasus.py:PegasusDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "The", "Transformers", "True", "ValueError", "You", "__init__", "_init_weight", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "both", "but", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_position_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "info", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "int", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "new_num_position_embeddings", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "positions", "post_init", "r", "raise", "rand", "range", "removed", "resize_position_embeddings", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "pegasus/modeling_pegasus.py:PegasusModel": ["BaseModelOutput", "Cache", "Embedding", "FloatTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "get_position_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "last_hidden_state", "len", "max_position_embeddings", "new_num_position_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "resize_position_embeddings", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "pegasus/modeling_pegasus.py:PegasusForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "get_position_embeddings", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "max_position_embeddings", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_position_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "register_buffer", "resize_position_embeddings", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "pegasus/modeling_pegasus.py:PegasusDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "pegasus/modeling_pegasus.py:PegasusForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "decoder", "deepcopy", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "get_position_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "max_position_embeddings", "model", "new_num_position_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "resize_position_embeddings", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "longt5/modeling_longt5.py:_pad_to_multiple": ["Tensor", "_pad_to_multiple", "all", "block_len", "constant", "def", "dim", "dtype", "functional", "if", "int", "list", "mode", "ndim", "new_shape", "nn", "not", "pad", "pad_len", "pad_value", "return", "shape", "sum", "torch", "value", "x", "zeros"], "longt5/modeling_longt5.py:_split_into_blocks": ["Tensor", "_pad_to_multiple", "_split_into_blocks", "block_len", "def", "device", "dim", "dtype", "empty", "if", "in", "int", "num_blocks", "output_shape", "pad_value", "reshape", "return", "shape", "torch", "x"], "longt5/modeling_longt5.py:_concatenate_3_blocks": ["None", "Tensor", "_concatenate_3_blocks", "append", "block_dim", "blocks_list", "cat", "constant", "def", "dim", "for", "functional", "i", "in", "indices", "int", "list", "mode", "ndim", "nn", "num_blocks", "pad", "pad_value", "range", "return", "sequence_dim", "shape", "slice", "sum", "torch", "tuple", "value", "x"], "longt5/modeling_longt5.py:_make_3block_relative_position_ids": ["Tensor", "_make_3block_relative_position_ids", "arange", "block_len", "center_position_ids", "def", "dtype", "int", "int32", "position_ids", "relative_position_ids", "return", "torch", "unsqueeze"], "longt5/modeling_longt5.py:_mask_local_attention_mask": ["None", "Tensor", "_make_3block_relative_position_ids", "_mask_local_attention_mask", "abs", "block_len", "def", "device", "int", "local_attention_mask", "locality_mask", "logical_and", "relative_position_ids", "return", "to", "torch"], "longt5/modeling_longt5.py:_get_local_attention_mask": ["Tensor", "_3blocked_attention_mask", "_blocked_attention_mask", "_concatenate_3_blocks", "_get_local_attention_mask", "_mask_local_attention_mask", "_split_into_blocks", "attention_mask", "block_dim", "block_len", "def", "device", "dim", "int", "local_attention_mask", "logical_and", "return", "sequence_dim", "to", "torch", "unsqueeze"], "longt5/modeling_longt5.py:_make_global_fixed_block_ids": ["Tensor", "_global_block_ids_lower_bound", "_make_global_fixed_block_ids", "_sequence_block_ids_max", "arange", "attention_mask", "axis", "batch_size", "block_ends", "block_ids", "cumsum", "def", "device", "dim", "dtype", "else", "fixed_block_mask", "floor", "full_blocks", "global_block_ids", "global_block_size", "global_segment_ids", "handle_orphan_tokens", "if", "int", "logical_and", "mask", "max", "num_globals", "ones", "ones_like", "repeat", "return", "seq_len", "shape", "sum", "tensor", "to", "torch", "transpose", "true_block_ends", "tuple", "type", "unsqueeze", "values", "where", "zeros"], "longt5/modeling_longt5.py:_make_side_relative_position_ids": ["None", "Tensor", "_make_global_fixed_block_ids", "_make_side_relative_position_ids", "arange", "attention_mask", "block_ids", "def", "device", "global_block_size", "global_positions", "global_segment_ids", "global_seq_len", "int", "int64", "return", "shape", "side_relative_position", "torch", "type"], "longt5/modeling_longt5.py:_create_global_aggregates": ["Tensor", "_create_global_aggregates", "block_ids", "def", "device", "dtype", "einsum", "functional", "gd", "global_seq_len", "hidden_states", "int", "int64", "nd", "ng", "nn", "one_hot", "one_hot_block_ids", "return", "tensor", "torch", "type", "where"], "longt5/modeling_longt5.py:LongT5LayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "longt5/modeling_longt5.py:LongT5DenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "longt5/modeling_longt5.py:LongT5DenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseGatedActDense", "Module", "__init__", "act", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "forward", "hidden_gelu", "hidden_linear", "hidden_states", "nn", "return", "self", "super", "wi_0", "wi_1", "wo"], "longt5/modeling_longt5.py:LongT5LayerFF": ["DenseReluDense", "Dropout", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelLayerFF", "ModelLayerNorm", "Module", "__init__", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is_gated_act", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "longt5/modeling_longt5.py:LongT5Attention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "Model", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "if", "index", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "len", "list", "log", "logger", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "union", "unsqueeze", "update", "use_cache", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "longt5/modeling_longt5.py:LongT5LocalAttention": ["Embedding", "False", "Linear", "Model", "ModelConfig", "ModelLocalAttention", "Module", "None", "True", "__init__", "_concatenate_3_blocks", "_relative_position_bucket", "_split_into_blocks", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "block_dim", "block_len", "block_length", "bool", "class", "compute_bias", "config", "context_position", "contiguous", "d_kv", "d_model", "def", "device", "dim", "dropout", "dropout_rate", "dtype", "einsum", "else", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "hqk", "if", "index", "inner_dim", "int", "is", "is_decoder", "is_small", "k", "key_states", "key_value_proj_dim", "khd", "layer_head_mask", "len", "local_radius", "log", "mask", "math", "max_distance", "max_exact", "memory_position", "meta", "min", "n_heads", "nn", "not", "num_buckets", "num_heads", "o", "output_attentions", "outputs", "p", "permute", "position_bias", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "qhd", "query_states", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "seq_length", "sequence_dim", "set", "shape", "softmax", "states", "staticmethod", "super", "target_device", "to", "torch", "training", "transpose", "type", "type_as", "union", "unshape", "unsqueeze", "v", "value_states", "values", "view", "weight", "where", "zeros", "zeros_like"], "longt5/modeling_longt5.py:LongT5TransientGlobalAttention": ["Embedding", "False", "Linear", "Model", "ModelConfig", "ModelLayerNorm", "ModelTransientGlobalAttention", "Module", "None", "Tensor", "True", "__init__", "_concatenate_3_blocks", "_create_global_aggregates", "_get_local_attention_mask", "_global_seq_len", "_make_global_fixed_block_ids", "_make_side_relative_position_ids", "_relative_position_bucket", "_split_into_blocks", "abs", "and", "arange", "attention_side_bias", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "block_dim", "block_ids", "block_len", "block_length", "bool", "cat", "class", "compute_bias", "compute_side_bias", "config", "context_position", "contiguous", "d_kv", "d_model", "def", "device", "dim", "dropout", "dropout_rate", "dtype", "einsum", "else", "eps", "eq", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "global_block_size", "global_input_layer_norm", "global_inputs", "global_relative_attention_bias", "global_segment_ids", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "hqk", "if", "index", "inner_dim", "int", "is", "is_decoder", "is_small", "k", "key_states", "key_value_proj_dim", "khd", "layer_head_mask", "layer_norm_epsilon", "len", "local_attention_mask", "local_radius", "log", "mask", "math", "max_distance", "max_exact", "memory_position", "meta", "min", "n_heads", "ndim", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "permute", "position_bias", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "qhd", "query_states", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "repeat", "reps", "requires_grad", "return", "scores", "self", "seq_length", "sequence_dim", "set", "shape", "side_attention_mask", "side_bias", "side_key_states", "side_position_bias", "side_relative_position", "side_relative_position_bucket", "side_value_states", "softmax", "states", "staticmethod", "super", "target_device", "to", "torch", "training", "transpose", "type", "type_as", "union", "unshape", "unsqueeze", "v", "value_states", "values", "view", "weight", "where", "zeros", "zeros_like"], "longt5/modeling_longt5.py:LongT5LayerSelfAttention": ["Dropout", "False", "ModelAttention", "ModelLayerNorm", "ModelLayerSelfAttention", "Module", "None", "Optional", "SelfAttention", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "longt5/modeling_longt5.py:LongT5LayerLocalSelfAttention": ["Any", "Dropout", "False", "LocalSelfAttention", "ModelLayerLocalSelfAttention", "ModelLayerNorm", "ModelLocalAttention", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "class", "config", "d_model", "def", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "kwargs", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "nn", "normed_hidden_states", "output_attentions", "outputs", "position_bias", "return", "self", "super"], "longt5/modeling_longt5.py:LongT5LayerTransientGlobalSelfAttention": ["Any", "Dropout", "False", "ModelLayerNorm", "ModelLayerTransientGlobalSelfAttention", "ModelTransientGlobalAttention", "Module", "None", "Optional", "TransientGlobalSelfAttention", "__init__", "attention_mask", "attention_output", "class", "config", "d_model", "def", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "kwargs", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "nn", "normed_hidden_states", "output_attentions", "outputs", "position_bias", "return", "self", "super"], "longt5/modeling_longt5.py:LongT5LayerCrossAttention": ["Dropout", "EncDecAttention", "False", "ModelAttention", "ModelLayerCrossAttention", "ModelLayerNorm", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "longt5/modeling_longt5.py:LongT5Block": ["False", "For", "GradientCheckpointingLayer", "ModelBlock", "ModelLayerCrossAttention", "ModelLayerFF", "ModelLayerLocalSelfAttention", "ModelLayerSelfAttention", "ModelLayerTransientGlobalSelfAttention", "ModuleList", "None", "Optional", "True", "ValueError", "__init__", "and", "any", "append", "attention", "attention_layer", "attention_mask", "attention_outputs", "but", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "either", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attention_type", "encoder_decoder_position_bias", "encoder_hidden_states", "expected", "f", "finfo", "float16", "forward", "global", "got", "has_relative_attention_bias", "hidden_states", "if", "int", "is", "is_decoder", "isinf", "key_value_states", "layer", "layer_head_mask", "layer_idx", "local", "max", "mechanism", "min", "new_name", "nn", "not", "or", "output_attentions", "past_key_value", "past_key_values", "position_bias", "query_length", "raise", "return", "return_dict", "self", "self_attention_outputs", "super", "torch", "transient", "type", "use_cache", "version"], "longt5/modeling_longt5.py:LongT5PreTrainedModel": ["Consider", "DUMMY_INPUTS", "DUMMY_MASK", "False", "In", "Model", "ModelAttention", "ModelBlock", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelEncoderModel", "ModelForConditionalGeneration", "ModelLayerNorm", "ModelLocalAttention", "ModelModel", "ModelPreTrainedModel", "ModelTransientGlobalAttention", "None", "PreTrainedModel", "Recovering", "See", "True", "ValueError", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_shift_right", "_tie_or_clone_weights", "_tied_weights_keys", "_try_load_missing_tied_module", "a", "and", "args", "base_model_prefix", "be", "bias", "cat", "checkpoint", "class", "classmethod", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "def", "defined", "dim", "docs", "dummy_inputs", "elif", "else", "endswith", "f", "factor", "fill_", "for", "from", "from_pretrained", "full", "get", "getattr", "global_relative_attention_bias", "has", "has_relative_attention_bias", "hasattr", "if", "in", "information", "initializer_factor", "input_ids", "input_mask", "is", "is_torch_fx_proxy", "isinstance", "it", "k", "key", "key_value_proj_dim", "kwargs", "legacy", "len", "lm_head", "loading_info", "logger", "masked_fill_", "mean", "missing", "missing_key", "missing_keys", "model", "module", "more", "n_heads", "new_zeros", "normal_", "not", "num_heads", "o", "or", "output_loading_info", "pad_token_id", "property", "q", "raise", "relative_attention_bias", "requested_loading_info", "return", "saving", "self", "set", "shape", "shared", "shifted_input_ids", "split", "std", "sub_key", "super", "supports_gradient_checkpointing", "tensor", "the", "tie_word_embeddings", "tied", "to", "torch", "transformer", "true", "updating", "usually", "v", "warning", "weight", "wi", "wi_0", "wi_1", "wo", "your", "zero_"], "longt5/modeling_longt5.py:LongT5Stack": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "ModelBlock", "ModelLayerNorm", "ModelPreTrainedModel", "ModelStack", "ModuleList", "None", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_get_local_attention_mask", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_attentions", "all_cross_attentions", "all_hidden_states", "and", "any", "arange", "assert", "at", "attention_mask", "attentions", "batch_size", "block", "block_len", "bool", "both", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "d_model", "decoder_", "def", "device", "diagonal", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder_attention_mask", "encoder_attention_type", "encoder_batch_size", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_relative_attention_bias", "have", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_torchdynamo_compiling", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "local", "local_radius", "logger", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "min", "min_dtype", "model", "new_embeddings", "nn", "not", "npu", "num_layers", "ones", "or", "output_attentions", "output_hidden_states", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_bias", "post_init", "raise", "range", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "using_compilable_cache", "v", "valid", "view", "vocab_size", "warning_once", "weight", "with", "xpu"], "longt5/modeling_longt5.py:LongT5Model": ["BaseModelOutput", "BoolTensor", "Cache", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelStack", "Modelensor", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "True", "Union", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention", "attention_mask", "attentions", "auto_docstring", "block", "bool", "cache_position", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "len", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "prune_heads", "r", "relative_attention_bias", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "tie_encoder_decoder", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "warn", "warnings", "weight"], "longt5/modeling_longt5.py:LongT5ForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "Cache", "CrossEntropyLoss", "Embedding", "EncDecAttention", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Linear", "ModelConfig", "ModelForConditionalGeneration", "ModelPreTrainedModel", "ModelStack", "Modelensor", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "True", "Union", "__HEAD_MASK_WARNING_MSG", "__init__", "_keys_to_ignore_on_load_unexpected", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "block", "bool", "cache_position", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "device", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "labels", "last_hidden_state", "layer", "len", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "model_dim", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "r", "relative_attention_bias", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "shared", "size", "super", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warn", "warnings", "weight"], "longt5/modeling_longt5.py:LongT5EncoderModel": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "ModelConfig", "ModelEncoderModel", "ModelPreTrainedModel", "ModelStack", "Modelensor", "None", "Optional", "Union", "__init__", "_keys_to_ignore_on_load_unexpected", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "attention", "attention_mask", "auto_docstring", "bool", "class", "config", "copy", "d_model", "decoder", "deepcopy", "def", "else", "embed_tokens", "encoder", "encoder_config", "encoder_outputs", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "is", "items", "layer", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "prune_heads", "r", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "tie_encoder_decoder", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "apertus/modeling_apertus.py:ApertusMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "apertus/modeling_apertus.py:ApertusRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "apertus/modeling_apertus.py:ApertusRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "apertus/modeling_apertus.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "apertus/modeling_apertus.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "apertus/modeling_apertus.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "apertus/modeling_apertus.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "apertus/modeling_apertus.py:ApertusAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "apertus/modeling_apertus.py:ApertusDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_layernorm", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "feedforward_layernorm", "forward", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "apertus/modeling_apertus.py:ApertusPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "apertus/modeling_apertus.py:ApertusModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "apertus/modeling_apertus.py:ApertusForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "apertus/modeling_apertus.py:ApertusForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "timesformer/modeling_timesformer.py:TimesformerPatchEmbeddings": ["Conv2d", "Iterable", "ModelPatchEmbeddings", "Module", "__init__", "abc", "batch_size", "class", "collections", "config", "def", "else", "embeddings", "flatten", "forward", "height", "hidden_size", "if", "image_size", "isinstance", "kernel_size", "nn", "num_channels", "num_frames", "num_patches", "patch_size", "patch_width", "pixel_values", "projection", "reshape", "return", "self", "shape", "size", "stride", "super", "transpose", "width"], "timesformer/modeling_timesformer.py:TimesformerEmbeddings": ["Dropout", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "Parameter", "_", "__init__", "attention_type", "batch_size", "cat", "class", "cls_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "drop_rate", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "mode", "nearest", "new_pos_embed", "new_time_embeddings", "nn", "num_frames", "num_patches", "other_pos_embed", "p", "patch_embeddings", "patch_height", "patch_num", "patch_width", "permute", "pixel_values", "pos_drop", "position_embeddings", "reshape", "return", "self", "shape", "size", "space_only", "super", "time_drop", "time_embeddings", "torch", "transpose", "unsqueeze", "view", "zeros"], "timesformer/modeling_timesformer.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "timesformer/modeling_timesformer.py:TimeSformerDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "timesformer/modeling_timesformer.py:TimesformerSelfAttention": ["Dropout", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "__init__", "attention_dropout_prob", "attention_probs", "attention_probs_dropout_prob", "attn_drop", "batch_size", "bias", "bool", "class", "config", "context_layer", "def", "dim", "else", "forward", "head_dim", "hidden_size", "hidden_states", "if", "key", "nn", "num_attention_heads", "num_channels", "num_heads", "output_attentions", "outputs", "permute", "qkv", "qkv_bias", "query", "reshape", "return", "scale", "self", "shape", "softmax", "super", "transpose", "value"], "timesformer/modeling_timesformer.py:TimesformerSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch"], "timesformer/modeling_timesformer.py:TimeSformerAttention": ["False", "ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Tensor", "Union", "__init__", "attention", "attention_output", "bool", "class", "config", "def", "forward", "hidden_states", "nn", "output", "output_attentions", "outputs", "return", "self", "self_outputs", "super", "torch", "tuple"], "timesformer/modeling_timesformer.py:TimesformerIntermediate": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelIntermediate", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "else", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "timesformer/modeling_timesformer.py:TimesformerOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "timesformer/modeling_timesformer.py:TimesformerLayer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Tensor", "True", "Unknown", "ValueError", "__init__", "attention", "attention_output", "attention_type", "batch_size", "bool", "cat", "class", "cls_token", "config", "cpu", "def", "device", "divided_space_time", "drop_path", "drop_path_rate", "drop_path_rates", "elif", "else", "eps", "f", "for", "forward", "hidden_size", "hidden_states", "if", "image_size", "in", "init_cls_token", "int", "intermediate", "item", "joint_space_time", "layer_index", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "linspace", "mean", "nn", "not", "num_frames", "num_hidden_layers", "num_patch_height", "num_patch_width", "num_spatial_tokens", "output", "output_attentions", "outputs", "patch_size", "permute", "raise", "repeat", "reshape", "residual", "residual_spatial", "residual_temporal", "return", "self", "self_attention_outputs", "shape", "size", "space_only", "spatial_attention_outputs", "spatial_embedding", "super", "temporal_attention", "temporal_attention_outputs", "temporal_dense", "temporal_embedding", "temporal_layernorm", "torch", "type", "unsqueeze", "x"], "timesformer/modeling_timesformer.py:TimesformerEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "ind", "is", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "timesformer/modeling_timesformer.py:TimesformerPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "apply", "base_model_prefix", "bias", "class", "cls_token", "config", "constant_", "def", "elif", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "module", "nn", "not", "patch_embeddings", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight"], "timesformer/modeling_timesformer.py:TimesformerModel": ["BaseModelOutput", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_prune_heads", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_input_embeddings", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "nn", "not", "output_attentions", "output_hidden_states", "patch_embeddings", "pixel_values", "post_init", "prune_heads", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "timesformer/modeling_timesformer.py:TimesformerForVideoClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Identity", "ImageClassifierOutput", "Linear", "MSELoss", "Model", "ModelForVideoClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "and", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "hidden_size", "hidden_states", "if", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "torch", "tuple", "use_return_dict", "view"], "nllb_moe/modeling_nllb_moe.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "nllb_moe/modeling_nllb_moe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Tensor", "axis", "def", "dtype", "expert_indices", "expert_mask", "float", "float32", "functional", "if", "int64", "is", "len", "max", "mean", "nn", "num_experts", "one_hot", "r", "return", "router_prob_per_group_and_expert", "router_probs", "shape", "to", "tokens_per_group_and_expert", "torch", "unsqueeze", "values"], "nllb_moe/modeling_nllb_moe.py:NllbMoeScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "nllb_moe/modeling_nllb_moe.py:NllbMoeSinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "contiguous", "cos", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "detach", "device", "dim", "dtype", "else", "emb", "emb_weights", "embedding_dim", "exp", "expand", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sequence_length", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "nllb_moe/modeling_nllb_moe.py:NllbMoeTop2Router": ["CB", "Gumbel", "Linear", "LongTensor", "ModelConfig", "ModelTop2Router", "Module", "None", "Optional", "SCB", "Tensor", "True", "__init__", "_cast_classifier", "and", "argmax", "argsort", "batch_prioritized_routing", "batch_size", "bias", "bool", "capacity", "ceil", "clamp", "class", "classifier", "config", "cumsum", "def", "denom_s", "device", "dim", "distributions", "dtype", "else", "eps", "expert_capacity", "finfo", "float", "float32", "forward", "functional", "gates1", "gates2", "getattr", "gumbel", "hasattr", "hidden_dim", "hidden_size", "hidden_states", "if", "importance_scores", "inf", "input_dtype", "is", "keepdim", "len", "locations1", "locations2", "logits_except_top_1", "lt", "masked_fill", "math", "max", "min", "moe_eval_capacity_token_fraction", "nb_tokens", "nn", "non_padding", "normalize_router_prob_before_dropping", "normalize_router_probabilities", "not", "num_classes", "num_experts", "one_hot", "or", "padding_mask", "r", "rand_like", "random", "repeat", "reshape", "return", "route_tokens", "router_bias", "router_dtype", "router_ignore_padding_tokens", "router_logits", "router_probs", "rsample", "sampled", "sampling", "second_expert_policy", "self", "sequence_length", "shape", "softmax", "sorted_cumsum1", "sorted_cumsum2", "sorted_top_1_mask", "sorted_top_2_mask", "sum", "super", "to", "top_1_expert_index", "top_1_mask", "top_1_max_probs", "top_2_expert_index", "top_2_mask", "top_2_max_probs", "torch", "training", "transpose", "tuple", "unsqueeze"], "nllb_moe/modeling_nllb_moe.py:NllbMoeDenseActDense": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelDenseActDense", "Module", "Tensor", "__init__", "act", "activation_dropout", "activation_function", "and", "class", "config", "d_model", "def", "dropout", "dtype", "fc1", "fc2", "ffn_dim", "forward", "hidden_states", "if", "int", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "uint8", "weight"], "nllb_moe/modeling_nllb_moe.py:NllbMoeSparseMLP": ["Dropout", "False", "ModelConfig", "ModelDenseActDense", "ModelSparseMLP", "ModelTop2Router", "Module", "ModuleDict", "Optional", "Tensor", "__init__", "argmax", "b", "batch_size", "be", "bm", "bool", "class", "combining_weights", "config", "def", "dim", "ebm", "einsum", "else", "enumerate", "expert", "expert_", "expert_class", "expert_output", "experts", "f", "ffn_dim", "for", "forward", "hidden_dim", "hidden_states", "idx", "if", "in", "int", "masked_hidden_states", "moe_token_dropout", "nn", "num_experts", "padding_mask", "r", "range", "reshape", "return", "router", "router_mask", "router_probs", "self", "sequence_length", "shape", "sum", "super", "token_dropout", "token_indices", "top_1_expert_index", "top_1_mask", "torch", "training", "values"], "nllb_moe/modeling_nllb_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "nllb_moe/modeling_nllb_moe.py:NllbMoeAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "class", "config", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "encoder_hidden_states", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "past_key_value", "past_key_values", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "nllb_moe/modeling_nllb_moe.py:NllbMoeEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDenseActDense", "ModelEncoderLayer", "ModelSparseMLP", "None", "Tensor", "__init__", "activation_dropout", "and", "any", "attention_dropout", "attention_mask", "attn_dropout", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "else", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "ff_dropout", "ff_layer_norm", "ffn", "ffn_dim", "finfo", "float16", "forward", "hidden_states", "if", "is_sparse", "isinf", "isnan", "layer_head_mask", "max", "min", "nn", "not", "num_heads", "or", "output_attentions", "output_router_logits", "outputs", "residual", "return", "router_states", "self", "self_attn", "self_attn_layer_norm", "super", "torch"], "nllb_moe/modeling_nllb_moe.py:NllbMoeDecoderLayer": ["ACT2FN", "Cache", "Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelDenseActDense", "ModelSparseMLP", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_dropout", "bool", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention", "cross_attention_layer_norm", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "dtype", "else", "embed_dim", "encoder_attention_mask", "encoder_hidden_states", "ff_dropout", "ff_layer_norm", "ffn", "ffn_dim", "finfo", "float16", "forward", "hidden_states", "if", "int", "is", "is_decoder", "is_sparse", "isinf", "layer_head_mask", "layer_idx", "max", "min", "new_name", "nn", "not", "num_heads", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "residual", "return", "router_states", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "use_cache", "version"], "nllb_moe/modeling_nllb_moe.py:NllbMoePreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "init_std", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "nllb_moe/modeling_nllb_moe.py:NllbMoeEncoder": ["Embedding", "False", "LayerNorm", "MoEModelOutput", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_attentions", "all_router_probs", "and", "append", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_sparse_step", "encoder_states", "enumerate", "f", "flash", "flex_attention", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "is_causal", "is_sparse", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "make_flex_block_causal_mask", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "output_router_logits", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "router_probs", "same", "scale_embedding", "sdpa", "self", "should", "size", "sparse_step", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "weight"], "nllb_moe/modeling_nllb_moe.py:NllbMoeDecoder": ["Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "MoEModelOutputWithPastAndCrossAttentions", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Size", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask", "_prepare_4d_causal_attention_mask_for_sdpa", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_router_probs", "all_self_attns", "an", "and", "append", "at", "attention_mask", "attentions", "attn_mask", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "decoder_sparse_step", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "flash_attention_2", "flex_attention", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "int", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "is_sparse", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "make_flex_block_causal_mask", "mask_name", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "output_router_logits", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "query_length", "r", "raise", "rand", "range", "removed", "return", "return_dict", "router_probs", "same", "scale_embedding", "sdpa", "self", "should", "size", "skip_the_layer", "sparse_step", "specified", "specify", "sqrt", "super", "synced_gpus", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with", "zip"], "nllb_moe/modeling_nllb_moe.py:NllbMoeModel": ["Cache", "FloatTensor", "LongTensor", "MoEModelOutput", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Seq2SeqMoEModelOutput", "Tensor", "True", "Union", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_router_logits", "def", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "encoder_router_logits", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "math", "not", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "return", "return_dict", "router_probs", "scale_embedding", "self", "set_input_embeddings", "shared", "sqrt", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "value", "vocab_size", "weight"], "nllb_moe/modeling_nllb_moe.py:NllbMoeForConditionalGeneration": ["Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqMoEOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "_unpack_router_logits", "and", "append", "attention_mask", "auto_docstring", "aux_loss", "base_model_prefix", "bias", "bool", "cache_position", "cat", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_aux_loss", "decoder_expert_indexes", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_router_logits", "decoder_start_token_id", "def", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_aux_loss", "encoder_expert_indexes", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "encoder_router_logits", "expert_indexes", "for", "forward", "get_decoder", "get_encoder", "head_mask", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "is", "labels", "len", "lm_head", "lm_logits", "load_balancing_loss_func", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "pad_token_id", "past_key_values", "post_init", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "router_output", "router_outputs", "router_z_loss_coef", "self", "shift_tokens_right", "size", "stack", "super", "torch", "total_expert_indexes", "total_router_logits", "tuple", "use_cache", "view", "vocab_size", "weight"], "olmo3/modeling_olmo3.py:Olmo3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "olmo3/modeling_olmo3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "olmo3/modeling_olmo3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "olmo3/modeling_olmo3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "k", "k_embed", "k_type", "position_ids", "q", "q_embed", "q_type", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "olmo3/modeling_olmo3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "olmo3/modeling_olmo3.py:Olmo3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "assert", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_type", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "olmo3/modeling_olmo3.py:Olmo3MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "olmo3/modeling_olmo3.py:Olmo3DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "olmo3/modeling_olmo3.py:Olmo3RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "assert", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dynamic_rope_update", "elif", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "olmo3/modeling_olmo3.py:Olmo3PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "olmo3/modeling_olmo3.py:Olmo3Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleDict", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "default", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_embeddings_mapping", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rope_type", "rotary_embs", "self", "self_attn", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "olmo3/modeling_olmo3.py:Olmo3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "glm4_moe/modeling_glm4_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "glm4_moe/modeling_glm4_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "glm4_moe/modeling_glm4_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "glm4_moe/modeling_glm4_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "rope_scaling", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "use_qk_norm", "v_proj", "value_states", "version", "view"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "return", "self", "super", "up_proj", "x"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeTopkRouter": ["F", "False", "ModelConfig", "ModelTopkRouter", "Module", "Parameter", "True", "__init__", "bool", "class", "config", "def", "denominator", "dim", "dtype", "e_score_correction_bias", "empty", "expand", "float32", "forward", "gather", "get_topk_indices", "group_idx", "group_mask", "group_scores", "hidden_size", "hidden_states", "if", "k", "keepdim", "linear", "masked_fill", "n_group", "n_routed_experts", "nn", "no_grad", "norm_topk_prob", "num_experts_per_tok", "register_buffer", "reshape", "return", "routed_scaling_factor", "router_logits", "scatter_", "score_mask", "scores", "scores_for_choice", "self", "sigmoid", "sorted", "sum", "super", "top_k", "topk", "topk_group", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight", "zeros", "zeros_like"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeMoE": ["ModelMLP", "ModelMoE", "ModelTopkRouter", "Module", "ModuleList", "Tensor", "_", "__init__", "class", "config", "def", "dtype", "expert", "expert_idx", "expert_input", "expert_mask", "expert_output", "expert_weights", "experts", "final_hidden_states", "for", "forward", "functional", "gate", "hidden_states", "if", "in", "index_add_", "intermediate_size", "len", "mask", "moe", "moe_intermediate_size", "n_routed_experts", "n_shared_experts", "nn", "num_classes", "numel", "one_hot", "orig_shape", "permute", "r", "range", "residuals", "return", "self", "shape", "shared_experts", "super", "token_indices", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight_indices", "weighted_output", "where", "zeros_like"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "first_k_dense_replace", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "glm4_moe/modeling_glm4_moe.py:Glm4MoePreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelTopkRouter", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_keys_to_ignore_on_load_unexpected", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "model", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "glm4_moe/modeling_glm4_moe.py:Glm4MoeForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "flex_olmo/modeling_flex_olmo.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "flex_olmo/modeling_flex_olmo.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "flex_olmo/modeling_flex_olmo.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "k", "k_embed", "k_type", "position_ids", "q", "q_embed", "q_type", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "flex_olmo/modeling_flex_olmo.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoSparseMoeBlock": ["F", "False", "Linear", "ModelMLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "keepdim", "nn", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoDecoderLayer": ["Cache", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoPreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSparseMoeBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_Model_attn", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "index", "model", "past_key_values", "router_logits", "supports_gradient_checkpointing"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoModel": ["Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "flex_olmo/modeling_flex_olmo.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "flex_olmo/modeling_flex_olmo.py:FlexOlmoForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "class", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "flaubert/modeling_flaubert.py:create_sinusoidal_embeddings": ["False", "FloatTensor", "Model_sinusoidal_embeddings", "array", "cos", "def", "detach_", "dim", "for", "in", "j", "n_pos", "np", "out", "pos", "position_enc", "power", "range", "requires_grad", "sin", "torch"], "flaubert/modeling_flaubert.py:get_masks": ["False", "Model_masks", "None", "alen", "arange", "assert", "attn_mask", "bs", "causal", "def", "device", "dtype", "else", "if", "is", "item", "lengths", "long", "mask", "max", "not", "or", "padding_mask", "repeat", "return", "size", "slen", "torch"], "flaubert/modeling_flaubert.py:MultiHeadAttention": ["EncoderDecoderCache", "False", "Linear", "ModelHeadAttention", "Module", "None", "True", "__init__", "and", "assert", "attention_dropout", "attention_head_size", "bs", "cache", "cache_position", "class", "config", "context", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "dim", "dropout", "dtype", "else", "expand_as", "find_pruneable_heads_and_indices", "finfo", "float", "forward", "functional", "get", "head_dim", "head_mask", "heads", "if", "index", "input", "int", "is", "is_cross_attention", "is_updated", "isinstance", "k", "k_lin", "key_cache", "kv", "layer_id", "layer_idx", "len", "mask", "mask_reshape", "masked_fill_", "math", "matmul", "min", "n_heads", "nn", "not", "out_lin", "output_attentions", "outputs", "p", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "q_lin", "qlen", "return", "scores", "self", "self_attention_cache", "set", "size", "softmax", "sqrt", "super", "torch", "training", "transpose", "type_as", "union", "update", "v", "v_lin", "value_cache", "view", "weights"], "flaubert/modeling_flaubert.py:TransformerFFN": ["Linear", "ModelFFN", "Module", "__init__", "act", "apply_chunking_to_forward", "chunk_size_feed_forward", "class", "config", "def", "dim_hidden", "dropout", "else", "ff_chunk", "forward", "functional", "gelu", "gelu_activation", "if", "in_dim", "input", "lin1", "lin2", "nn", "out_dim", "p", "relu", "return", "self", "seq_len_dim", "super", "training", "x"], "flaubert/modeling_flaubert.py:FlaubertPredLayer": ["AdaptiveLogSoftmaxWithLoss", "False", "Linear", "ModelPredLayer", "Module", "None", "True", "_", "__init__", "asm", "asm_cutoffs", "asm_div_value", "bias", "class", "config", "cross_entropy", "cutoffs", "def", "dim", "div_value", "else", "emb_dim", "forward", "functional", "head_bias", "if", "in_features", "is", "log_prob", "loss", "mean", "n_classes", "n_words", "nn", "not", "outputs", "pad_index", "proj", "reduction", "return", "scores", "self", "super", "view", "x", "y"], "flaubert/modeling_flaubert.py:FlaubertSquadHeadOutput": ["FloatTensor", "LongTensor", "ModelOutput", "ModelSquadHeadOutput", "None", "Optional", "class", "cls_logits", "end_top_index", "end_top_log_probs", "loss", "r", "start_top_index", "start_top_log_probs", "torch"], "flaubert/modeling_flaubert.py:FlaubertPoolerStartLogits": ["FloatTensor", "Linear", "ModelConfig", "ModelPoolerStartLogits", "Module", "None", "Optional", "__init__", "class", "config", "def", "dense", "dtype", "else", "float16", "forward", "hidden_size", "hidden_states", "if", "is", "nn", "not", "p_mask", "return", "self", "squeeze", "super", "torch", "x"], "flaubert/modeling_flaubert.py:FlaubertPoolerEndLogits": ["FloatTensor", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelPoolerEndLogits", "Module", "None", "One", "Optional", "Tanh", "__init__", "activation", "assert", "be", "cat", "class", "config", "def", "dense_0", "dense_1", "dim", "dtype", "else", "eps", "expand", "float16", "forward", "gather", "hidden_size", "hidden_states", "hsz", "if", "is", "layer_norm_eps", "nn", "not", "of", "or", "p_mask", "return", "self", "shape", "should", "slen", "squeeze", "start_positions", "start_states", "super", "torch", "x"], "flaubert/modeling_flaubert.py:FlaubertPoolerAnswerClass": ["False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelPoolerAnswerClass", "Module", "None", "One", "Optional", "Tanh", "__init__", "activation", "assert", "be", "bias", "cat", "class", "cls_index", "cls_token_state", "config", "def", "dense_0", "dense_1", "dim", "else", "expand", "forward", "gather", "hidden_size", "hidden_states", "hsz", "if", "is", "nn", "not", "of", "or", "return", "self", "shape", "should", "squeeze", "start_positions", "start_states", "super", "torch", "x"], "flaubert/modeling_flaubert.py:FlaubertSQuADHead": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelPoolerAnswerClass", "ModelPoolerEndLogits", "ModelPoolerStartLogits", "ModelSQuADHead", "ModelSquadHeadOutput", "Module", "None", "Optional", "Union", "__init__", "and", "answer_class", "auto_docstring", "bh", "bl", "blh", "bool", "bsz", "class", "cls_index", "cls_logits", "cls_loss", "config", "def", "dim", "einsum", "else", "end_log_probs", "end_logits", "end_loss", "end_n_top", "end_positions", "end_top_index", "end_top_log_probs", "expand", "expand_as", "for", "forward", "functional", "gather", "hidden_states", "hidden_states_expanded", "hsz", "if", "in", "is", "is_impossible", "loss", "loss_fct", "loss_fct_cls", "nn", "not", "p_mask", "r", "return", "return_dict", "self", "size", "slen", "softmax", "squeeze_", "start_log_probs", "start_logits", "start_loss", "start_n_top", "start_positions", "start_states", "start_top_index", "start_top_index_exp", "start_top_log_probs", "super", "topk", "torch", "total_loss", "tuple", "unsqueeze", "view", "x"], "flaubert/modeling_flaubert.py:FlaubertSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "ModelConfig", "ModelSequenceSummary", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "flaubert/modeling_flaubert.py:FlaubertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "__init__", "_init_weights", "and", "attention_mask", "attns_list", "base_model_prefix", "bias", "class", "config", "constant_", "create_sinusoidal_embeddings", "data", "def", "dummy_inputs", "else", "emb_dim", "embed_init_std", "fill_", "if", "init", "init_std", "input_ids", "inputs", "inputs_list", "is", "isinstance", "kwargs", "langs", "langs_list", "max_position_embeddings", "mean", "module", "n_langs", "nn", "normal_", "not", "out", "padding_idx", "position_embeddings", "property", "return", "self", "sinusoidal_embeddings", "std", "super", "tensor", "torch", "transformer", "use_lang_emb", "weight", "zero_"], "flaubert/modeling_flaubert.py:FlaubertModel": ["BaseModelOutput", "Currently", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "ModuleList", "MultiHeadAttention", "None", "NotImplementedError", "Optional", "Tensor", "TransformerFFN", "Union", "__init__", "_prune_heads", "_slen", "a", "an", "and", "append", "arange", "as", "assert", "attention_dropout", "attention_mask", "attentions", "attn", "attn_mask", "attn_outputs", "auto_docstring", "be", "bool", "bs", "cache", "cache_position", "can", "causal", "class", "config", "continue", "copy", "def", "device", "dict", "dim", "dropout", "dropout_probability", "dtype", "else", "emb_dim", "embeddings", "encoder", "eos_index", "eps", "expand", "expand_as", "ffns", "for", "forward", "from_legacy_cache", "functional", "get_head_mask", "get_input_embeddings", "get_masks", "get_seq_length", "getattr", "hasattr", "head_mask", "heads", "heads_to_prune", "hidden_dim", "hidden_states", "i", "if", "in", "input_ids", "inputs_embeds", "int", "is", "is_decoder", "is_encoder", "isinstance", "item", "items", "lang_embeddings", "langs", "last_hidden_state", "layer", "layer_idx", "layer_norm1", "layer_norm2", "layer_norm_emb", "layer_norm_eps", "layerdrop", "lengths", "list", "long", "map", "mask", "max", "max_position_embeddings", "multiple", "must", "n_heads", "n_langs", "n_layers", "n_words", "new_embeddings", "nn", "not", "of", "only", "output_attentions", "output_hidden_states", "p", "pad_index", "padding_idx", "padding_mask", "persistent", "position_embeddings", "position_ids", "post_init", "pre_norm", "prune_heads", "pruned_heads", "r", "raise", "rand", "range", "register_buffer", "return", "return_dict", "self", "set_input_embeddings", "size", "slen", "str", "sum", "super", "tensor", "tensor_normalized", "to", "token_type_ids", "torch", "training", "transformer", "tuple", "unsqueeze", "use_lang_emb", "use_return_dict", "used", "v"], "flaubert/modeling_flaubert.py:FlaubertWithLMHeadModel": ["GenerationMixin", "MaskedLMOutput", "ModelModel", "ModelPreTrainedModel", "ModelPredLayer", "ModelWithLMHeadModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "cat", "class", "config", "def", "device", "dict", "dim", "dtype", "effective_batch_size", "else", "forward", "full", "full_like", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lang_id", "langs", "lengths", "logits", "long", "loss", "mask_token", "mask_token_id", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "pred_layer", "prepare_inputs_for_generation", "proj", "r", "return", "return_dict", "self", "set_output_embeddings", "shape", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict", "weight"], "flaubert/modeling_flaubert.py:FlaubertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "class", "config", "def", "dict", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "langs", "lengths", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_summary", "single_label_classification", "squeeze", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict", "view"], "flaubert/modeling_flaubert.py:FlaubertForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "class", "classifier", "config", "def", "dict", "dropout", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "langs", "lengths", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "str", "super", "token_type_ids", "torch", "transformer", "tuple", "use_return_dict", "view"], "flaubert/modeling_flaubert.py:FlaubertForQuestionAnsweringSimple": ["CrossEntropyLoss", "Linear", "ModelForQuestionAnsweringSimple", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "clamp", "class", "config", "contiguous", "def", "dict", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "langs", "len", "lengths", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "str", "super", "token_type_ids", "torch", "total_loss", "transformer", "transformer_outputs", "tuple", "use_return_dict"], "flaubert/modeling_flaubert.py:FlaubertForQuestionAnsweringOutput": ["FloatTensor", "LongTensor", "ModelForQuestionAnsweringOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cls_logits", "end_top_index", "end_top_log_probs", "hidden_states", "loss", "r", "start_top_index", "start_top_log_probs", "torch", "tuple"], "flaubert/modeling_flaubert.py:FlaubertForQuestionAnswering": ["ModelForQuestionAnswering", "ModelForQuestionAnsweringOutput", "ModelModel", "ModelPreTrainedModel", "ModelSQuADHead", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache", "class", "cls_index", "cls_logits", "config", "def", "dict", "else", "end_positions", "end_top_index", "end_top_log_probs", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_impossible", "langs", "lengths", "loss", "not", "output", "output_attentions", "output_hidden_states", "outputs", "p_mask", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "start_positions", "start_top_index", "start_top_log_probs", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict"], "flaubert/modeling_flaubert.py:FlaubertForMultipleChoice": ["CrossEntropyLoss", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "ModelSequenceSummary", "MultipleChoiceModelOutput", "None", "Optional", "Please", "Tensor", "The", "Union", "__init__", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache", "cannot", "choice", "class", "config", "def", "dict", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs", "inputs_embeds", "instead", "is", "kwargs", "labels", "langs", "lengths", "logger", "logits", "logits_proj", "loss", "loss_fct", "mask", "models", "multiple", "nn", "not", "num_choices", "num_labels", "output", "output_attentions", "output_hidden_states", "parameter", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_summary", "shape", "size", "str", "super", "the", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use", "use_return_dict", "used", "view", "warning", "with"], "mobilenet_v2/modeling_mobilenet_v2.py:make_divisible": ["Model_divisible", "None", "Optional", "def", "divisor", "if", "int", "is", "max", "min_value", "new_value", "return", "value"], "mobilenet_v2/modeling_mobilenet_v2.py:apply_depth_multiplier": ["ModelConfig", "Model_depth_multiplier", "channels", "config", "def", "depth_divisible_by", "depth_multiplier", "int", "make_divisible", "min_depth", "return", "round"], "mobilenet_v2/modeling_mobilenet_v2.py:apply_tf_padding": ["Conv2d", "Model_tf_padding", "Tensor", "constant", "conv_layer", "def", "dilation", "dilation_height", "dilation_width", "else", "features", "functional", "if", "in_height", "in_width", "int", "kernel_height", "kernel_size", "kernel_width", "max", "nn", "pad", "pad_along_height", "pad_along_width", "pad_bottom", "pad_left", "pad_right", "pad_top", "padding", "return", "shape", "stride", "stride_height", "stride_width", "torch"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2ConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Input", "ModelConfig", "ModelConvLayer", "Module", "None", "Optional", "Output", "Tensor", "True", "Union", "ValueError", "__init__", "activation", "affine", "apply_tf_padding", "are", "bias", "bool", "by", "channels", "class", "config", "convolution", "def", "dilation", "divisible", "elif", "else", "eps", "f", "features", "float", "forward", "groups", "hidden_act", "if", "in_channels", "int", "is", "isinstance", "kernel_size", "layer_norm_eps", "momentum", "nn", "normalization", "not", "num_features", "out_channels", "padding", "padding_mode", "raise", "return", "self", "str", "stride", "super", "tf_padding", "torch", "track_running_stats", "use_activation", "use_normalization", "zeros"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2InvertedResidual": ["False", "Invalid", "ModelConfig", "ModelConvLayer", "ModelInvertedResidual", "Module", "None", "Tensor", "ValueError", "__init__", "and", "class", "config", "conv_3x3", "def", "depth_divisible_by", "dilation", "else", "expand_1x1", "expand_ratio", "expanded_channels", "f", "features", "forward", "groups", "if", "in", "in_channels", "int", "kernel_size", "make_divisible", "min_depth", "nn", "not", "out_channels", "raise", "reduce_1x1", "residual", "return", "round", "self", "stride", "super", "torch", "use_activation", "use_residual"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2Stem": ["False", "ModelConfig", "ModelConvLayer", "ModelStem", "Module", "None", "Tensor", "__init__", "class", "config", "conv_3x3", "def", "else", "expand_1x1", "expanded_channels", "features", "first_conv", "first_layer_is_expansion", "forward", "groups", "if", "in_channels", "int", "is", "kernel_size", "nn", "not", "out_channels", "reduce_1x1", "return", "self", "stride", "super", "torch", "use_activation"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2PreTrainedModel": ["BatchNorm2d", "Conv2d", "False", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "Union", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2Model": ["AdaptiveAvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "ModelConfig", "ModelConvLayer", "ModelInvertedResidual", "ModelModel", "ModelPreTrainedModel", "ModelStem", "ModuleList", "None", "NotImplementedError", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "all_hidden_states", "and", "append", "apply_depth_multiplier", "auto_docstring", "bool", "channels", "class", "config", "conv_1x1", "conv_stem", "current_stride", "def", "depth_multiplier", "dilation", "else", "enumerate", "expanded_channels", "finegrained_output", "flatten", "for", "forward", "have", "heads_to_prune", "hidden_states", "i", "if", "in", "in_channels", "is", "kernel_size", "last_hidden_state", "layer", "layer_dilation", "layer_module", "layer_stride", "nn", "not", "num_channels", "out_channels", "output_channels", "output_hidden_states", "output_stride", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "r", "raise", "range", "return", "return_dict", "self", "specify", "start_dim", "stride", "strides", "super", "to", "torch", "tuple", "use_return_dict", "v", "x"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2ForImageClassification": ["Dropout", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "classifier_dropout_prob", "config", "conv_1x1", "convolution", "def", "dropout", "else", "forward", "hidden_states", "if", "inplace", "is", "labels", "last_hidden_size", "logits", "loss", "loss_function", "nn", "not", "num_labels", "out_channels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2DeepLabV3Plus": ["AdaptiveAvgPool2d", "Dropout2d", "False", "ModelConfig", "ModelConvLayer", "ModelDeepLabV3Plus", "Module", "None", "Tensor", "True", "__init__", "align_corners", "apply_depth_multiplier", "avg_pool", "bias", "bilinear", "cat", "class", "classifier", "classifier_dropout_prob", "config", "conv_aspp", "conv_pool", "conv_projection", "def", "dim", "dropout", "features", "features_aspp", "features_pool", "forward", "functional", "in_channels", "interpolate", "kernel_size", "layer_norm_eps", "mode", "nn", "num_labels", "out_channels", "output_size", "relu", "return", "self", "shape", "size", "spatial_size", "stride", "super", "torch", "use_activation", "use_normalization"], "mobilenet_v2/modeling_mobilenet_v2.py:MobileNetV2ForSemanticSegmentation": ["CrossEntropyLoss", "False", "Model", "ModelConfig", "ModelDeepLabV3Plus", "ModelForSemanticSegmentation", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SemanticSegmenterOutput", "Tensor", "The", "True", "Union", "ValueError", "__init__", "add_pooling_layer", "align_corners", "and", "attentions", "auto_docstring", "be", "bilinear", "bool", "class", "config", "def", "else", "encoder_hidden_states", "forward", "functional", "greater", "hidden_states", "if", "ignore_index", "interpolate", "is", "labels", "logits", "loss", "loss_fct", "mode", "nn", "not", "num_labels", "number", "of", "one", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "segmentation_head", "self", "semantic_loss_ignore_index", "shape", "should", "size", "super", "than", "torch", "tuple", "upsampled_logits", "use_return_dict"], "openai/modeling_openai.py:Attention": ["Conv1D", "Dropout", "False", "Model", "Model_mask", "Module", "None", "True", "ValueError", "__init__", "_attn", "a", "append", "attn_dropout", "attn_outputs", "attn_pdrop", "b", "be", "bias", "by", "c_attn", "c_proj", "cat", "class", "config", "contiguous", "def", "dim", "divisible", "else", "f", "find_pruneable_heads_and_indices", "forward", "functional", "head_mask", "heads", "if", "index", "index_attn", "is", "k", "key", "len", "math", "matmul", "merge_heads", "must", "n_head", "n_positions", "n_state", "new_x_shape", "nn", "not", "nx", "ones", "output_Models", "outputs", "permute", "persistent", "prune_conv1d_layer", "prune_heads", "pruned_heads", "q", "query", "raise", "register_buffer", "resid_dropout", "resid_pdrop", "return", "scale", "self", "set", "shape", "size", "softmax", "split", "split_heads", "split_size", "sqrt", "super", "torch", "tril", "union", "v", "value", "view", "w", "x"], "openai/modeling_openai.py:MLP": ["ACT_FNS", "Conv1D", "Dropout", "Model", "Module", "__init__", "act", "afn", "c_fc", "c_proj", "class", "config", "def", "dropout", "forward", "h", "h2", "n_embd", "n_state", "nn", "nx", "resid_pdrop", "return", "self", "super", "x"], "openai/modeling_openai.py:Block": ["Attention", "False", "LayerNorm", "MLP", "Model", "Module", "None", "__init__", "a", "attention_mask", "attn", "attn_outputs", "class", "config", "def", "eps", "forward", "h", "head_mask", "layer_norm_epsilon", "ln_1", "ln_2", "m", "mlp", "n", "n_embd", "n_positions", "nn", "nx", "output_attentions", "outputs", "return", "scale", "self", "super", "x"], "openai/modeling_openai.py:OpenAIGPTSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "ModelGPTConfig", "ModelGPTSequenceSummary", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "openai/modeling_openai.py:OpenAIGPTPreTrainedModel": ["Conv1D", "Embedding", "LayerNorm", "Linear", "ModelGPTConfig", "ModelGPTPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "transformer", "weight", "zero_"], "openai/modeling_openai.py:OpenAIGPTDoubleHeadsModelOutput": ["FloatTensor", "ModelGPTDoubleHeadsModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "mc_logits", "mc_loss", "r", "torch", "tuple"], "openai/modeling_openai.py:OpenAIGPTModel": ["BaseModelOutput", "Block", "Dropout", "Embedding", "False", "FloatTensor", "LongTensor", "ModelGPTModel", "ModelGPTPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "all_attentions", "all_hidden_states", "and", "arange", "at", "attention_mask", "attentions", "attn", "auto_docstring", "block", "bool", "both", "cannot", "class", "config", "def", "drop", "dtype", "either", "elif", "else", "embd_pdrop", "enumerate", "finfo", "for", "forward", "get_head_mask", "get_input_embeddings", "h", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "i", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "min", "n_embd", "n_layer", "n_positions", "new_embeddings", "next", "nn", "not", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "parameters", "persistent", "position_embeds", "position_ids", "positions_embed", "post_init", "prune_heads", "raise", "range", "register_buffer", "return", "return_dict", "same", "scale", "self", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_embeds", "token_type_ids", "tokens_embed", "torch", "tuple", "unsqueeze", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "openai/modeling_openai.py:OpenAIGPTLMHeadModel": ["Any", "CausalLMOutput", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelGPTLMHeadModel", "ModelGPTModel", "ModelGPTPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "def", "dict", "else", "for", "forward", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_function", "model_inputs", "n_embd", "nn", "not", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "str", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict", "value", "vocab_size", "weight"], "openai/modeling_openai.py:OpenAIGPTDoubleHeadsModel": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelGPTDoubleHeadsModel", "ModelGPTDoubleHeadsModelOutput", "ModelGPTModel", "ModelGPTPreTrainedModel", "ModelGPTSequenceSummary", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "contiguous", "def", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_head", "lm_logits", "lm_loss", "logits", "loss", "loss_fct", "mc_labels", "mc_logits", "mc_loss", "mc_token_ids", "multiple_choice_head", "n_embd", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "return_dict", "self", "shift_labels", "shift_logits", "size", "squeeze", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "openai/modeling_openai.py:OpenAIGPTForSequenceClassification": ["BCEWithLogitsLoss", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "ModelGPTForSequenceClassification", "ModelGPTModel", "ModelGPTPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "n_embd", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "sequence_length", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_return_dict", "using", "view", "warning_once", "will", "with"], "fuyu/modeling_fuyu.py:FuyuPreTrainedModel": ["Embedding", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "fuyu/modeling_fuyu.py:FuyuModel": ["AutoModel", "Batch", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "Got", "Image", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "Number", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_checkpoint_conversion_mapping", "all", "and", "arange", "as_tuple", "at", "attention_mask", "auto_docstring", "batch", "batch_idx", "batch_size", "bool", "both", "cannot", "cat", "class", "clone", "config", "continuous", "continuous_embeddings", "decoder", "def", "device", "dim", "do", "does", "dst_indices", "dtype", "either", "element", "elif", "else", "embeddings", "expand_as", "f", "features", "for", "forward", "from_config", "gather_continuous_embeddings", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_seq_length", "gradient_checkpointing", "have", "hidden_size", "ids", "if", "image", "image_features", "image_patch_input_indices", "image_patches", "image_patches_indices", "image_token_id", "in", "input_ids", "input_is", "inputs_embeds", "is", "kwargs", "language_model", "len", "list", "long", "masked_scatter", "match", "model", "must", "n_image_features", "n_image_tokens", "nn", "nonzero", "not", "num_channels", "number", "numel", "of", "or", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "pad_token_id", "padding_idx", "past_key_values", "past_key_values_length", "patch", "patch_embeddings", "patch_size", "pixel_values", "position_ids", "post_init", "r", "raise", "range", "return", "return_dict", "same", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "sizes", "special_image_mask", "specify", "squeeze", "src_indices", "sum", "super", "tensor", "text_config", "the", "time", "to", "token", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_embed_tokens", "vocab_size", "weight", "word_embeddings"], "fuyu/modeling_fuyu.py:FuyuForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_patches", "image_patches_indices", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_embed_tokens", "vocab_size", "weight"], "bit/modeling_bit.py:get_padding_value": ["False", "Model_padding_value", "None", "True", "and", "bool", "def", "dilation", "dynamic", "elif", "else", "if", "is", "isinstance", "kernel_size", "lower", "padding", "r", "return", "same", "str", "stride", "tuple", "valid"], "bit/modeling_bit.py:WeightStandardizedConv2d": ["Conv2d", "DynamicPad2d", "False", "Model", "ModelStandardizedConv2d", "None", "SAME", "True", "__init__", "batch_norm", "bias", "class", "conv2d", "def", "dilation", "else", "eps", "forward", "functional", "get_padding_value", "groups", "hidden_state", "if", "in_channel", "is", "is_dynamic", "kernel_size", "momentum", "nn", "not", "out_channels", "pad", "padding", "reshape", "reshape_as", "return", "self", "stride", "super", "training"], "bit/modeling_bit.py:BitGroupNormActivation": ["ACT2FN", "GroupNorm", "Identity", "ModelGroupNormActivation", "True", "__init__", "activation", "affine", "apply_activation", "bias", "class", "config", "def", "else", "eps", "forward", "functional", "group_norm", "hidden_act", "hidden_state", "if", "nn", "num_channels", "num_groups", "r", "return", "self", "super", "weight"], "bit/modeling_bit.py:DynamicPad2d": ["ModelPad2d", "Module", "__init__", "ceil", "class", "compute_padding", "def", "dilation", "forward", "functional", "if", "input", "input_height", "input_width", "int", "isinstance", "kernel_size", "math", "max", "nn", "or", "pad", "padding_height", "padding_width", "r", "return", "self", "size", "stride", "super", "value", "x"], "bit/modeling_bit.py:BitMaxPool2d": ["DynamicPad2d", "False", "Identity", "Iterable", "MaxPool2d", "ModelMaxPool2d", "None", "True", "__init__", "abc", "ceil_mode", "class", "collections", "def", "dilation", "else", "forward", "functional", "hidden_states", "if", "int", "isinstance", "kernel_size", "max_pool2d", "nn", "pad", "padding", "padding_value", "return", "self", "stride", "super", "use_dynamic_padding"], "bit/modeling_bit.py:BitEmbeddings": ["ConstantPad2d", "Identity", "Make", "ModelConfig", "ModelEmbeddings", "ModelGroupNormActivation", "ModelMaxPool2d", "Module", "None", "SAME", "Tensor", "ValueError", "WeightStandardizedConv2d", "__init__", "and", "channel", "class", "config", "configuration", "convolution", "def", "dimension", "else", "embedding", "embedding_dynamic_padding", "embedding_size", "eps", "forward", "global_padding", "if", "in", "is", "kernel_size", "layer_type", "match", "nn", "norm", "not", "num_channels", "of", "one", "pad", "padding", "pixel", "pixel_values", "pooler", "preactivation", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "upper", "use_dynamic_padding", "value", "values", "with"], "bit/modeling_bit.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "bit/modeling_bit.py:BitDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "bit/modeling_bit.py:make_div": ["Model_div", "def", "divisor", "if", "int", "max", "min_value", "new_value", "return", "value"], "bit/modeling_bit.py:BitPreActivationBottleneckLayer": ["False", "Identity", "ModelDownsampleConv", "ModelDropPath", "ModelGroupNormActivation", "ModelPreActivationBottleneckLayer", "Module", "None", "True", "WeightStandardizedConv2d", "__init__", "bottle_ratio", "class", "config", "conv1", "conv2", "conv3", "def", "dilation", "downsample", "drop_path", "drop_path_rate", "else", "eps", "first_dilation", "forward", "global_padding", "groups", "hidden_states", "hidden_states_preact", "if", "in_channels", "is", "is_first_layer", "make_div", "mid_channels", "nn", "norm1", "norm2", "norm3", "not", "num_channels", "or", "out_channels", "padding", "preact", "return", "self", "shortcut", "stride", "super"], "bit/modeling_bit.py:BitBottleneckLayer": ["ACT2FN", "False", "Identity", "ModelBottleneckLayer", "ModelDownsampleConv", "ModelDropPath", "ModelGroupNormActivation", "Module", "None", "WeightStandardizedConv2d", "__init__", "activation", "apply_activation", "bottle_ratio", "class", "config", "conv1", "conv2", "conv3", "def", "dilation", "downsample", "drop_path", "drop_path_rate", "else", "eps", "first_dilation", "forward", "global_padding", "groups", "hidden_act", "hidden_states", "if", "in_channels", "is", "is_first_layer", "make_div", "mid_chs", "nn", "norm1", "norm2", "norm3", "not", "num_channels", "or", "out_channels", "padding", "preact", "return", "self", "shortcut", "stride", "super"], "bit/modeling_bit.py:BitDownsampleConv": ["False", "Identity", "ModelDownsampleConv", "ModelGroupNormActivation", "Module", "True", "WeightStandardizedConv2d", "__init__", "apply_activation", "class", "config", "conv", "def", "else", "eps", "forward", "global_padding", "if", "in_channels", "nn", "norm", "num_channels", "out_channels", "padding", "preact", "return", "self", "stride", "super", "x"], "bit/modeling_bit.py:BitStage": ["ModelBottleneckLayer", "ModelPreActivationBottleneckLayer", "ModelStage", "Module", "None", "Sequential", "Tensor", "_", "__init__", "_get_updated_hyperparameters", "add_module", "bottle_ratio", "bottleneck", "class", "config", "def", "depth", "dilation", "drop_path_rate", "else", "enumerate", "first_dilation", "for", "forward", "hidden_state", "if", "in", "in_channels", "input", "is_first_layer", "layer", "layer_cls", "layer_dropout", "layer_idx", "layer_type", "layers", "nn", "out_channels", "prev_chs", "r", "range", "return", "self", "str", "stride", "super"], "bit/modeling_bit.py:BitEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "_get_updated_hyperparameters", "add_module", "bool", "class", "config", "current_depth", "current_hidden_size", "current_stride", "def", "depth", "depths", "dilation", "drop_path_rate", "else", "embedding_size", "enumerate", "for", "forward", "hidden_sizes", "hidden_state", "hidden_states", "if", "in", "is", "last_hidden_state", "layer_dropout", "layer_dropouts", "linspace", "make_div", "nn", "not", "np", "out_channels", "output_hidden_states", "output_stride", "prev_chs", "return", "return_dict", "self", "split", "stage", "stage_idx", "stage_module", "stages", "str", "stride", "sum", "super", "tolist", "torch", "tuple", "v", "width_factor", "x", "zip"], "bit/modeling_bit.py:BitPreTrainedModel": ["BatchNorm2d", "Conv2d", "GroupNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "None", "PreTrainedModel", "_", "_calculate_fan_in_and_fan_out", "_init_weights", "_no_split_modules", "a", "base_model_prefix", "bias", "bound", "class", "config", "constant_", "def", "elif", "else", "fan_in", "fan_out", "if", "init", "is", "isinstance", "kaiming_normal_", "kaiming_uniform_", "main_input_name", "math", "mode", "module", "nn", "nonlinearity", "not", "pixel_values", "relu", "self", "sqrt", "uniform_", "weight"], "bit/modeling_bit.py:BitModel": ["AdaptiveAvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "Identity", "ModelEmbeddings", "ModelEncoder", "ModelGroupNormActivation", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embedder", "embedding_output", "encoder", "encoder_outputs", "forward", "hidden_sizes", "hidden_states", "if", "is", "last_hidden_state", "layer_type", "nn", "norm", "not", "num_channels", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "preactivation", "return", "return_dict", "self", "super", "use_return_dict"], "bit/modeling_bit.py:BitForImageClassification": ["Flatten", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Sequential", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "use_return_dict"], "bit/modeling_bit.py:BitBackbone": ["BackboneMixin", "BackboneOutput", "False", "Model", "ModelBackbone", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedding_size", "enumerate", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_features", "out_features", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "stage", "stage_names", "super", "use_return_dict"], "vit/modeling_vit.py:ViTEmbeddings": ["BoolTensor", "Dropout", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "else", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "randn", "reshape", "return", "self", "seq_length", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_mask_token", "view", "width", "zeros"], "vit/modeling_vit.py:ViTPatchEmbeddings": ["Conv2d", "Expected", "False", "Input", "Iterable", "Make", "ModelConfig", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "bool", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "flatten", "forward", "got", "height", "hidden_size", "if", "image", "image_size", "in", "interpolate_pos_encoding", "isinstance", "kernel_size", "match", "model", "nn", "not", "num_channels", "num_patches", "of", "one", "or", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "torch", "transpose", "values", "width", "with"], "vit/modeling_vit.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "vit/modeling_vit.py:ViTSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "vit/modeling_vit.py:ViTSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "vit/modeling_vit.py:ViTAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "vit/modeling_vit.py:ViTIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "vit/modeling_vit.py:ViTOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "vit/modeling_vit.py:ViTLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "vit/modeling_vit.py:ViTEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "vit/modeling_vit.py:ViTPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "dtype", "elif", "fill_", "float32", "hidden_states", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "vit/modeling_vit.py:ViTModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "BoolTensor", "False", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dict", "dtype", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "expected_dtype", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "int", "interpolate_pos_encoding", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "projection", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch", "use_mask_token", "weight"], "vit/modeling_vit.py:ViTPooler": ["ACT2FN", "Linear", "ModelConfig", "ModelPooler", "Module", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "pooler_act", "pooler_output_size", "return", "self", "super", "torch"], "vit/modeling_vit.py:ViTForMaskedImageModeling": ["BaseModelOutputWithPooling", "BoolTensor", "Conv2d", "False", "Got", "MaskedImageModelingOutput", "Model", "ModelConfig", "ModelForMaskedImageModeling", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PixelShuffle", "Sequential", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "When", "__init__", "add_pooling_layer", "and", "as", "attentions", "auto_docstring", "batch_size", "be", "bool", "bool_masked_pos", "can_return_tuple", "class", "config", "contiguous", "decoder", "def", "dimensions", "encoder_stride", "ensure", "equal", "f", "floor", "forward", "functional", "has", "head_mask", "height", "hidden_size", "hidden_states", "if", "image", "image_size", "in_channels", "input", "interpolate_pos_encoding", "is", "kernel_size", "kwargs", "l1_loss", "last_hidden_state", "loss", "mask", "masked_im_loss", "math", "must", "nn", "none", "not", "num_channels", "out_channels", "outputs", "patch_size", "permute", "pixel_values", "post_init", "provided", "r", "raise", "reconstructed", "reconstructed_pixel_values", "reconstruction", "reconstruction_loss", "reduction", "repeat_interleave", "reshape", "return", "same", "self", "sequence_length", "sequence_output", "shape", "size", "sum", "super", "that", "the", "to", "torch", "unsqueeze", "use_mask_token", "width"], "vit/modeling_vit.py:ViTForImageClassification": ["BaseModelOutputWithPooling", "False", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "outputs", "pixel_values", "pooled_output", "post_init", "r", "return", "self", "sequence_output", "super", "torch"], "blenderbot/modeling_blenderbot.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "blenderbot/modeling_blenderbot.py:BlenderbotLearnedPositionalEmbedding": ["Embedding", "ModelLearnedPositionalEmbedding", "None", "Optional", "Size", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "embedding_dim", "forward", "if", "input_ids_shape", "int", "is", "long", "nn", "num_embeddings", "past_key_values_length", "position_ids", "return", "self", "seq_len", "super", "torch", "weight"], "blenderbot/modeling_blenderbot.py:BlenderbotScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "blenderbot/modeling_blenderbot.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "blenderbot/modeling_blenderbot.py:BlenderbotAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "blenderbot/modeling_blenderbot.py:BlenderbotEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "blenderbot/modeling_blenderbot.py:BlenderbotDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "blenderbot/modeling_blenderbot.py:BlenderbotPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "decoder_input_ids", "def", "device", "diagonal", "dim", "dtype", "dummy_inputs", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "ne", "nn", "normal_", "not", "npu", "ones", "pad_token", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "property", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tensor", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "blenderbot/modeling_blenderbot.py:BlenderbotEncoder": ["BaseModelOutput", "Embedding", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "The", "True", "ValueError", "You", "_", "__init__", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "both", "but", "cannot", "class", "config", "d_model", "def", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "blenderbot/modeling_blenderbot.py:BlenderbotDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "both", "but", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "blenderbot/modeling_blenderbot.py:BlenderbotModel": ["BaseModelOutput", "Cache", "FloatTensor", "FutureWarning", "In", "LongTensor", "Model", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSmallModel", "None", "Optional", "PathLike", "Seq2SeqModelOutput", "Tensor", "The", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "checkpoint", "class", "classmethod", "cls", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "deprecated", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "facebook", "forward", "from_pretrained", "future", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "identical", "if", "input_ids", "inputs_embeds", "instead", "is", "isinstance", "kwargs", "last_hidden_state", "len", "math", "model_args", "not", "os", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "please", "post_init", "pretrained_model_name_or_path", "r", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "small_Model", "sqrt", "str", "super", "the", "torch", "tuple", "use", "use_cache", "use_return_dict", "value", "vocab_size", "warn", "warnings", "weight", "with"], "blenderbot/modeling_blenderbot.py:BlenderbotForConditionalGeneration": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "In", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "ModelSmallForConditionalGeneration", "None", "Optional", "PathLike", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "checkpoint", "class", "classmethod", "cls", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "deprecated", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "facebook", "final_logits_bias", "forward", "from_pretrained", "future", "get_decoder", "get_encoder", "head_mask", "identical", "if", "input_ids", "inputs_embeds", "instead", "int", "is", "kwargs", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "model_args", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "os", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "please", "post_init", "pretrained_model_name_or_path", "provided", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "small_Model", "str", "super", "the", "to", "torch", "tuple", "use", "use_cache", "use_return_dict", "view", "vocab_size", "warn", "warning", "warnings", "weight", "with", "zeros"], "blenderbot/modeling_blenderbot.py:BlenderbotDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "blenderbot/modeling_blenderbot.py:BlenderbotForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "ernie/modeling_ernie.py:ErnieEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "shape", "size", "super", "task_type_embeddings", "task_type_ids", "task_type_vocab_size", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "use_task_id", "vocab_size", "word_embeddings", "zeros"], "ernie/modeling_ernie.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "ernie/modeling_ernie.py:ErnieSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "ernie/modeling_ernie.py:ErnieCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "ernie/modeling_ernie.py:ErnieSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "ernie/modeling_ernie.py:ErnieAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "ernie/modeling_ernie.py:ErnieIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "ernie/modeling_ernie.py:ErnieOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "ernie/modeling_ernie.py:ErnieLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "ernie/modeling_ernie.py:ErniePooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "ernie/modeling_ernie.py:ErniePredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "ernie/modeling_ernie.py:ErnieLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "ernie/modeling_ernie.py:ErnieEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "ernie/modeling_ernie.py:ErniePreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLMPredictionHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "ernie/modeling_ernie.py:ErnieModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Setting", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "at", "attention", "attention_mask", "auto_docstring", "batch_size", "be", "bool", "both", "cache_position", "cannot", "check_model_inputs", "checkpointing", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "heads", "heads_to_prune", "if", "in", "incompatible", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "not", "num_hidden_layers", "of", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "same", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "size", "specify", "super", "task_type_ids", "tgt_len", "the", "time", "to", "to_legacy_cache", "token_type_ids", "torch", "training", "tuple", "type", "use", "use_cache", "v4", "value", "warn_if_padding_and_no_attention_mask", "warning_once", "will", "with", "word_embeddings", "work"], "ernie/modeling_ernie.py:ErnieForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "ernie/modeling_ernie.py:ErniePreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "ernie/modeling_ernie.py:ErnieForPreTraining": ["CrossEntropyLoss", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "can_return_tuple", "class", "cls", "config", "decoder", "def", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "next_sentence_label", "next_sentence_loss", "not", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_output", "set_output_embeddings", "super", "task_type_ids", "token_type_ids", "torch", "total_loss", "tuple", "view", "vocab_size", "weight"], "ernie/modeling_ernie.py:ErnieOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "ernie/modeling_ernie.py:ErnieForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "If", "Model", "ModelForCausalLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "cls", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "list", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "task_type_ids", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "ernie/modeling_ernie.py:ErnieForMaskedLM": ["CrossEntropyLoss", "False", "If", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bi", "bias", "bool", "can_generate", "can_return_tuple", "cat", "class", "classmethod", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "task_type_ids", "to", "token", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "ernie/modeling_ernie.py:ErnieOnlyNSPHead": ["Linear", "ModelOnlyNSPHead", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "ernie/modeling_ernie.py:ErnieForNextSentencePrediction": ["CrossEntropyLoss", "FutureWarning", "Model", "ModelForNextSentencePrediction", "ModelModel", "ModelOnlyNSPHead", "ModelPreTrainedModel", "NextSentencePredictorOutput", "None", "Optional", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "a", "and", "argument", "attention_mask", "attentions", "auto_docstring", "be", "can_return_tuple", "class", "cls", "config", "def", "deprecated", "forward", "future", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "outputs", "pooled_output", "pop", "position_ids", "post_init", "r", "removed", "return", "return_dict", "self", "seq_relationship_scores", "super", "task_type_ids", "token_type_ids", "torch", "tuple", "use", "version", "view", "warn", "warnings", "will"], "ernie/modeling_ernie.py:ErnieForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "task_type_ids", "token_type_ids", "torch", "tuple", "view"], "ernie/modeling_ernie.py:ErnieForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "task_type_ids", "token_type_ids", "torch", "tuple", "view"], "ernie/modeling_ernie.py:ErnieForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "task_type_ids", "token_type_ids", "torch", "tuple", "view"], "ernie/modeling_ernie.py:ErnieForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "task_type_ids", "token_type_ids", "torch", "total_loss", "tuple"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrDecoderOutput": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "ModelDecoderOutput", "None", "Optional", "class", "intermediate_hidden_states", "r", "reference_points", "torch", "tuple"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrModelOutput": ["FloatTensor", "ModelModelOutput", "None", "Optional", "Seq2SeqModelOutput", "class", "intermediate_hidden_states", "r", "reference_points", "torch", "tuple"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrObjectDetectionOutput": ["FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrSegmentationOutput": ["FloatTensor", "ModelOutput", "ModelSegmentationOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "pred_masks", "r", "torch", "tuple"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "conditional_detr/modeling_conditional_detr.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrConvEncoder": ["Either", "False", "ModelConvEncoder", "Module", "None", "Tensor", "True", "ValueError", "__init__", "and", "append", "backbone", "backbone_config", "backbone_kwargs", "backbone_model_type", "be", "bool", "channels", "class", "config", "copy", "create_model", "def", "dilation", "elif", "else", "feature_info", "feature_map", "feature_maps", "features", "features_only", "float", "for", "forward", "functional", "get", "getattr", "if", "in", "in_chans", "intermediate_channel_sizes", "interpolate", "is", "kwargs", "layer2", "layer3", "layer4", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "num_channels", "or", "out", "out_indices", "output_stride", "parameter", "pixel_mask", "pixel_values", "pop", "pretrained", "provided", "raise", "replace_batch_norm", "requires_backends", "requires_grad_", "resnet", "return", "self", "shape", "should", "size", "stage", "super", "the", "timm", "to", "torch", "use_pretrained_backbone", "use_timm_backbone", "with"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrConvModel": ["ModelConvModel", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrSinePositionEmbedding": ["False", "ModelSinePositionEmbedding", "Module", "No", "None", "True", "ValueError", "__init__", "and", "arange", "be", "cat", "class", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "flatten", "float", "float32", "floor", "forward", "if", "int64", "is", "mask", "math", "nn", "normalize", "not", "passed", "permute", "pi", "pixel", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "provided", "raise", "return", "rounding_mode", "scale", "self", "should", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrLearnedPositionEmbedding": ["Embedding", "ModelLearnedPositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "conditional_detr/modeling_conditional_detr.py:build_position_encoding": ["ModelLearnedPositionEmbedding", "ModelSinePositionEmbedding", "Model_position_encoding", "Not", "True", "ValueError", "config", "d_model", "def", "elif", "else", "f", "if", "learned", "n_steps", "normalize", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "conditional_detr/modeling_conditional_detr.py:gen_sine_position_embeddings": ["Model_sine_position_embeddings", "None", "arange", "cat", "cos", "d_model", "def", "device", "dim", "dim_t", "div", "dtype", "flatten", "float32", "floor", "math", "pi", "pos", "pos_tensor", "pos_x", "pos_y", "return", "rounding_mode", "scale", "sin", "stack", "to", "torch", "x_embed", "y_embed"], "conditional_detr/modeling_conditional_detr.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "conditional_detr/modeling_conditional_detr.py:DetrAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "is_cross_attention", "k_proj", "key_states", "key_value_states", "key_value_states_original", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "object_queries", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "spatial_position_embeddings", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_qk_shape", "_v_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "if", "inf", "int", "is", "key_states", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "of", "out_dim", "out_proj", "output_attentions", "p", "proj_shape", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_head_dim", "v_proj_shape", "value_states", "view", "weights", "zeros_like"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrEncoderLayer": ["ACT2FN", "DetrAttention", "False", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderLayer", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "max", "min", "nn", "num_heads", "object_queries", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "_", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "batch_size", "bool", "ca_kcontent_proj", "ca_kpos_proj", "ca_qcontent_proj", "ca_qpos_proj", "ca_qpos_sine_proj", "ca_v_proj", "cat", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "dim", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "is_first", "k", "k_content", "k_pos", "key_states", "n_model", "nhead", "nn", "not", "num_heads", "num_queries", "object_queries", "out_dim", "output_attentions", "outputs", "p", "q", "q_content", "q_pos", "query_position_embeddings", "query_sine_embed", "residual", "return", "sa_kcontent_proj", "sa_kpos_proj", "sa_qcontent_proj", "sa_qpos_proj", "sa_v_proj", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "shape", "source_len", "super", "torch", "training", "v", "value_states", "view"], "conditional_detr/modeling_conditional_detr.py:MLP": ["Linear", "Model", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "Linear", "ModelConfig", "ModelConvEncoder", "ModelDecoderLayer", "ModelEncoderLayer", "ModelLearnedPositionEmbedding", "ModelMHAttentionMap", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "column_embeddings", "config", "data", "def", "elif", "gain", "if", "init", "init_std", "init_xavier_std", "is", "isinstance", "k_linear", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "q_linear", "r", "row_embeddings", "self", "std", "uniform_", "weight", "xavier_std", "xavier_uniform_", "zero_", "zeros_"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "True", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "for", "forward", "functional", "hidden_states", "i", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layers", "nn", "not", "object_queries", "output_attentions", "output_hidden_states", "p", "post_init", "r", "rand", "range", "return", "return_dict", "self", "super", "to_drop", "torch", "training", "tuple", "use_return_dict", "v"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrDecoder": ["False", "LayerNorm", "MLP", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "_prepare_4d_attention_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attention_mask", "attentions", "auxiliary_loss", "ca_qpos_proj", "class", "config", "continue", "cross_attentions", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gen_sine_position_embeddings", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "input_shape", "inputs_embeds", "intermediate", "intermediate_hidden_states", "is", "is_first", "last_hidden_state", "layer_id", "layer_outputs", "layerdrop", "layernorm", "layers", "nn", "not", "obj_center", "object_queries", "output_attentions", "output_hidden_states", "pos_transformation", "post_init", "query_position_embeddings", "query_scale", "query_sine_embed", "query_sine_embed_before_transformation", "r", "rand", "range", "ref_point_head", "reference_points", "reference_points_before_sigmoid", "return", "return_dict", "self", "sigmoid", "size", "stack", "super", "tgt_len", "torch", "training", "transpose", "tuple", "use_return_dict", "v"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrModel": ["Backbone", "BaseModelOutput", "Conv2d", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_size", "bool", "build_position_encoding", "class", "config", "conv_encoder", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "device", "does", "downsampled", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "feature_map", "features", "flatten", "flattened_features", "flattened_mask", "for", "forward", "freeze_backbone", "get_encoder", "height", "hidden_states", "if", "in", "input_projection", "inputs_embeds", "intermediate_channel_sizes", "intermediate_hidden_states", "is", "isinstance", "kernel_size", "last_hidden_state", "len", "mask", "model", "name", "named_parameters", "nn", "not", "num_channels", "num_queries", "object_queries", "object_queries_list", "ones", "output_attentions", "output_hidden_states", "param", "permute", "pixel", "pixel_mask", "pixel_values", "post_init", "projected_feature_map", "queries", "query_position_embeddings", "r", "raise", "reference_points", "repeat", "requires_grad_", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "unfreeze_backbone", "unsqueeze", "use_return_dict", "weight", "width", "zeros_like"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForObjectDetection", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_set_aux_loss", "a", "append", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "b", "bbox_predictor", "bool", "class", "class_labels_classifier", "config", "cross_attentions", "d_model", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "def", "device", "dict", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "hidden_dim", "hs", "if", "in", "input_dim", "inputs_embeds", "intermediate", "intermediate_hidden_states", "inverse_sigmoid", "is", "jit", "labels", "last_hidden_state", "list", "logits", "loss", "loss_dict", "loss_function", "lvl", "model", "nn", "not", "num_labels", "num_layers", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "outputs_class", "outputs_coord", "outputs_coords", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "r", "range", "reference", "reference_before_sigmoid", "reference_points", "return", "return_dict", "self", "sequence_output", "shape", "sigmoid", "stack", "super", "tmp", "torch", "transpose", "tuple", "unused", "use_return_dict", "zip"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrForSegmentation": ["BaseModelOutput", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelForObjectDetection", "ModelForSegmentation", "ModelMHAttentionMap", "ModelMaskHeadSmallConv", "ModelPreTrainedModel", "ModelSegmentationOutput", "None", "Optional", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "backbone", "batch_size", "bbox_attention", "bbox_mask", "bbox_predictor", "bool", "class", "class_labels_classifier", "config", "conv_encoder", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "device", "dict", "dropout", "elif", "else", "encoder", "encoder_attention_heads", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "feature_map", "features", "flatten", "flattened_features", "flattened_mask", "forward", "height", "hidden_size", "hidden_states", "if", "init_xavier_std", "input_projection", "inputs_embeds", "intermediate", "intermediate_channel_sizes", "intermediate_hidden_states", "is", "isinstance", "labels", "last_hidden_state", "len", "list", "logits", "loss", "loss_dict", "loss_function", "mask", "mask_head", "memory", "model", "not", "num_channels", "num_queries", "number_of_heads", "object_queries", "object_queries_list", "ones", "output", "output_attentions", "output_hidden_states", "outputs_class", "outputs_coord", "permute", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "pred_masks", "projected_feature_map", "queries", "query_position_embeddings", "r", "repeat", "return", "return_dict", "seg_masks", "self", "sequence_output", "shape", "sigmoid", "std", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "view", "weight", "width", "zeros_like"], "conditional_detr/modeling_conditional_detr.py:_expand": ["_expand", "def", "flatten", "int", "length", "repeat", "return", "tensor", "unsqueeze"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrMaskHeadSmallConv": ["Conv2d", "GroupNorm", "ModelMaskHeadSmallConv", "Module", "Tensor", "The", "ValueError", "__init__", "_expand", "a", "adapter1", "adapter2", "adapter3", "as", "attention", "bbox_mask", "be", "bias", "by", "cat", "class", "constant_", "context_dim", "cur_fpn", "def", "dim", "divisible", "flatten", "for", "forward", "fpn_dims", "fpns", "functional", "gn1", "gn2", "gn3", "gn4", "gn5", "groups", "heads", "hidden_size", "if", "in", "init", "inter_dims", "interpolate", "is", "isinstance", "kaiming_uniform_", "lay1", "lay2", "lay3", "lay4", "lay5", "list", "m", "min", "mode", "modules", "must", "nearest", "nn", "number", "of", "out_lay", "padding", "raise", "relu", "return", "self", "set", "shape", "size", "super", "the", "to", "torch", "weight", "x"], "conditional_detr/modeling_conditional_detr.py:ConditionalDetrMHAttentionMap": ["Dropout", "Linear", "ModelMHAttentionMap", "Module", "None", "Optional", "Tensor", "True", "__init__", "bias", "bnchw", "bqnc", "bqnhw", "class", "conv2d", "def", "dim", "dropout", "dtype", "einsum", "finfo", "flatten", "float", "forward", "functional", "hidden_dim", "if", "is", "k", "k_linear", "keys_per_head", "mask", "masked_fill", "min", "nn", "normalize_fact", "not", "num_heads", "q", "q_linear", "queries_per_head", "query_dim", "return", "self", "shape", "size", "softmax", "std", "super", "torch", "unsqueeze", "view", "weight", "weights"], "focalnet/modeling_focalnet.py:FocalNetEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "r", "reshaped_hidden_states", "torch", "tuple"], "focalnet/modeling_focalnet.py:FocalNetModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "reshaped_hidden_states", "torch", "tuple"], "focalnet/modeling_focalnet.py:FocalNetMaskedImageModelingOutput": ["FloatTensor", "ModelMaskedImageModelingOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "loss", "r", "reconstruction", "reshaped_hidden_states", "torch", "tuple"], "focalnet/modeling_focalnet.py:FocalNetImageClassifierOutput": ["FloatTensor", "ModelImageClassifierOutput", "ModelOutput", "None", "Optional", "class", "hidden_states", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "focalnet/modeling_focalnet.py:FocalNetEmbeddings": ["BoolTensor", "Dropout", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "True", "_", "__init__", "batch_size", "bool_masked_pos", "class", "config", "def", "dropout", "else", "embed_dim", "embeddings", "eps", "expand", "forward", "grid_size", "hidden_dropout_prob", "if", "image_size", "is", "is_stem", "layer_norm_eps", "mask", "mask_token", "mask_tokens", "nn", "norm", "not", "num_channels", "output_dimensions", "patch_embeddings", "patch_grid", "patch_size", "pixel_values", "return", "self", "seq_len", "size", "super", "torch", "tuple", "type_as", "unsqueeze", "use_conv_embed", "use_mask_token", "zeros"], "focalnet/modeling_focalnet.py:FocalNetPatchEmbeddings": ["Conv2d", "False", "FloatTensor", "Iterable", "LayerNorm", "Make", "ModelPatchEmbeddings", "Module", "None", "Optional", "Tensor", "ValueError", "_", "__init__", "abc", "add_norm", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embed_dim", "embeddings", "eps", "flatten", "forward", "functional", "grid_size", "height", "if", "image_size", "in", "int", "is", "is_stem", "isinstance", "kernel_size", "layer_norm_eps", "match", "maybe_pad", "nn", "norm", "not", "num_channels", "num_patches", "of", "one", "output_dimensions", "pad", "pad_values", "padding", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "tuple", "use_conv_embed", "values", "width", "with"], "focalnet/modeling_focalnet.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "focalnet/modeling_focalnet.py:FocalNetDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "focalnet/modeling_focalnet.py:FocalNetModulation": ["Conv2d", "Dropout", "False", "GELU", "LayerNorm", "Linear", "ModelModulation", "Model_factor", "Model_layers", "Model_level", "Model_levels", "Model_window", "Model_windows", "Module", "ModuleList", "Sequential", "True", "__init__", "activation", "append", "bias", "class", "config", "contiguous", "ctx", "ctx_all", "ctx_global", "def", "dim", "eps", "for", "forward", "gates", "groups", "hidden_state", "if", "in", "index", "k", "keepdim", "kernel_size", "kernel_sizes", "layer_norm_eps", "layernorm", "level", "mean", "modulator", "nn", "normalize_modulator", "num_channels", "padding", "permute", "projection_context", "projection_dropout", "projection_in", "projection_out", "q", "range", "return", "self", "shape", "split", "stride", "super", "torch", "use_post_layernorm_in_modulation", "x", "x_out"], "focalnet/modeling_focalnet.py:FocalNetMlp": ["ACT2FN", "Dropout", "Linear", "ModelMlp", "Module", "None", "__init__", "activation", "class", "config", "def", "drop", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "hidden_state", "in_features", "nn", "or", "out_features", "return", "self", "super"], "focalnet/modeling_focalnet.py:FocalNetLayer": ["Identity", "LayerNorm", "ModelDropPath", "ModelLayer", "ModelMlp", "ModelModulation", "Module", "Parameter", "True", "_", "__init__", "batch_size", "class", "config", "def", "dim", "drop", "drop_path", "else", "eps", "forward", "gamma_1", "gamma_2", "height", "hidden_dropout_prob", "hidden_features", "hidden_state", "if", "in_features", "index", "input_dimensions", "input_resolution", "int", "layer_norm_eps", "layerscale_value", "mlp", "mlp_hidden_dim", "mlp_ratio", "modulation", "nn", "norm1", "norm2", "not", "num_channels", "ones", "projection_dropout", "r", "requires_grad", "return", "self", "shape", "shortcut", "super", "torch", "use_layerscale", "use_post_layernorm", "view", "width"], "focalnet/modeling_focalnet.py:FocalNetStage": ["False", "GradientCheckpointingLayer", "ModelLayer", "ModelPatchEmbeddings", "ModelStage", "ModuleList", "None", "Tensor", "True", "__init__", "add_norm", "class", "config", "cpu", "def", "depths", "device", "dim", "downsample", "dpr", "drop_path", "drop_path_rate", "else", "embed_dim", "for", "forward", "height", "hidden_states", "hidden_states_before_downsampling", "i", "if", "image_size", "in", "index", "input_dimensions", "input_resolution", "int", "is", "is_stem", "isinstance", "item", "layer_module", "layers", "len", "linspace", "list", "nn", "not", "num_channels", "num_stages", "out_dim", "output_dimensions", "patch_size", "pointing", "range", "reshape", "return", "self", "shape", "stage_outputs", "sum", "super", "torch", "transpose", "tuple", "use_conv_embed", "width", "x"], "focalnet/modeling_focalnet.py:FocalNetEncoder": ["False", "ModelEncoder", "ModelEncoderOutput", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "and", "batch_size", "bool", "class", "config", "def", "depths", "elif", "else", "enumerate", "for", "forward", "gradient_checkpointing", "grid_size", "hidden_size", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "index", "input_dimensions", "input_resolution", "int", "is", "last_hidden_state", "len", "nn", "not", "num_stages", "output_dimensions", "output_hidden_states", "output_hidden_states_before_downsampling", "permute", "range", "reshaped_hidden_state", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "stage_module", "stage_outputs", "stages", "super", "torch", "tuple", "v", "view"], "focalnet/modeling_focalnet.py:FocalNetPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelStage", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "gamma_1", "gamma_2", "if", "initializer_range", "is", "isinstance", "layerscale_value", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "use_layerscale", "weight", "zero_"], "focalnet/modeling_focalnet.py:FocalNetModel": ["AdaptiveAvgPool1d", "BoolTensor", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "add_pooling_layer", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flatten", "forward", "get_input_embeddings", "have", "hidden_states", "if", "input_dimensions", "int", "is", "last_hidden_state", "layer_norm_eps", "layernorm", "len", "nn", "not", "num_features", "num_stages", "output", "output_hidden_states", "patch_embeddings", "patch_grid", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "r", "raise", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "tuple", "use_mask_token", "use_return_dict"], "focalnet/modeling_focalnet.py:FocalNetForMaskedImageModeling": ["BoolTensor", "Conv2d", "False", "FloatTensor", "Model", "ModelForMaskedImageModeling", "ModelMaskedImageModelingOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PixelShuffle", "Sequential", "True", "Union", "__init__", "add_pooling_layer", "auto_docstring", "batch_size", "bool", "bool_masked_pos", "class", "config", "contiguous", "decoder", "def", "depths", "else", "embed_dim", "encoder_stride", "floor", "forward", "functional", "height", "hidden_states", "if", "image_size", "in_channels", "int", "is", "kernel_size", "l1_loss", "len", "loss", "mask", "masked_im_loss", "math", "nn", "none", "not", "num_channels", "num_features", "num_stages", "out_channels", "output", "output_hidden_states", "outputs", "patch_size", "pixel_values", "post_init", "r", "reconstructed_pixel_values", "reconstruction", "reconstruction_loss", "reduction", "repeat_interleave", "reshape", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_length", "sequence_output", "shape", "size", "sum", "super", "torch", "transpose", "tuple", "unsqueeze", "use_mask_token", "use_return_dict", "width"], "focalnet/modeling_focalnet.py:FocalNetForImageClassification": ["FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelImageClassifierOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_features", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "post_init", "r", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "focalnet/modeling_focalnet.py:FocalNetBackbone": ["BackboneMixin", "BackboneOutput", "False", "Model", "ModelBackbone", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embed_dim", "enumerate", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_features", "out_features", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "reshaped_hidden_states", "return", "return_dict", "self", "stage", "stage_names", "super", "torch", "use_return_dict"], "mamba2/modeling_mamba2.py:pad_tensor_by_size": ["Model", "Model_shape", "Model_size", "Model_tensor_by_size", "Tensor", "constant", "def", "else", "functional", "if", "input_tensor", "int", "len", "mode", "nn", "return", "shape", "torch", "value"], "mamba2/modeling_mamba2.py:reshape_into_chunks": ["Model", "Model_into_chunks", "chunk_size", "def", "else", "if", "input_tensor", "len", "pad_size", "pad_tensor_by_size", "return", "shape"], "mamba2/modeling_mamba2.py:segment_sum": ["Model_sum", "None", "bool", "chunk_size", "cumsum", "def", "device", "diagonal", "dim", "dtype", "expand", "inf", "input_tensor", "mask", "masked_fill", "ones", "return", "size", "tensor_segsum", "torch", "tril"], "mamba2/modeling_mamba2.py:apply_mask_to_padding_states": ["Model_mask_to_padding_states", "None", "and", "attention_mask", "def", "dtype", "hidden_states", "if", "is", "not", "return", "shape", "to"], "mamba2/modeling_mamba2.py:Mamba2Cache": ["False", "ModelCache", "ModelConfig", "None", "Optional", "Tensor", "__init__", "batch_size", "bool", "cache_init", "class", "config", "conv_kernel", "conv_kernel_size", "conv_states", "def", "device", "dims", "dtype", "else", "expand", "float16", "head_dim", "hidden_size", "if", "int", "intermediate_size", "layer_idx", "n_groups", "new_conv_state", "new_ssm_state", "num_heads", "num_hidden_layers", "reset", "return", "roll", "self", "shifts", "ssm_states", "state_size", "str", "to", "torch", "update_conv_state", "update_ssm_state", "zero_", "zeros"], "mamba2/modeling_mamba2.py:MambaRMSNormGated": ["ModelRMSNormGated", "Module", "None", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "functional", "gate", "hidden_size", "hidden_states", "if", "input_dtype", "is", "keepdim", "mean", "nn", "not", "ones", "pow", "return", "rsqrt", "self", "silu", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "mamba2/modeling_mamba2.py:Mamba2Mixer": ["A", "ACT2FN", "AILab", "A_cumsum", "A_log", "B", "B_decay", "C", "C_reshaped", "C_times_states", "Conv1d", "D", "D_residual", "Dao", "Falling", "False", "G", "G_intermediate", "L", "Linear", "LongTensor", "M", "M_intermediate", "Model", "ModelCache", "ModelConfig", "ModelMixer", "ModelRMSNormGated", "Model_chunk_scan_combined", "Model_split_conv1d_scan_combined", "Module", "None", "Optional", "Parameter", "Tensor", "The", "To", "True", "Y_diag", "Y_off", "_", "__init__", "act", "activation", "and", "apply_mask_to_padding_states", "arange", "attention_mask", "available", "back", "batch_size", "because", "bias", "bmm", "cache_device", "cache_init", "cache_params", "cache_position", "cat", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk_size", "clamp", "class", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_dim", "conv_kernel", "conv_kernel_size", "conv_states", "cuda", "cuda_kernels_forward", "cumsum", "dA", "dB", "dBx", "d_mlp", "decay_chunk", "decay_states", "def", "device", "dim", "dt", "dt_bias", "dt_limit", "dt_limit_kwargs", "dt_softplus", "dtype", "else", "eps", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "groups_time_state_size", "head_dim", "headdim", "hidden_act", "hidden_size", "hidden_states", "hidden_states_B_C", "hidden_states_B_C_transposed", "hidden_states_reshaped", "https", "if", "implementation", "in", "in_channels", "in_proj", "inf", "install", "int", "intermediate_size", "is", "is_fast_path_available", "kernel_size", "layer_idx", "layer_norm_epsilon", "log", "logger", "n_groups", "naive", "new_conv_state", "new_ssm_state", "new_states", "ngroups", "nn", "norm", "norm_before_gate", "not", "num_heads", "of", "one", "ones", "out", "out_channels", "out_proj", "outproj_bias", "outproj_weight", "output_size", "pad", "pad_size", "pad_tensor_by_size", "padding", "path", "permute", "previous_states", "projected_states", "projection_size", "repeat_interleave", "reshape", "reshape_into_chunks", "return", "return_final_states", "rms_norm", "rmsnorm_eps", "rmsnorm_weight", "scan_output", "segment_sum", "selective_state_update", "self", "seq_idx", "seq_len", "shape", "silu", "softplus", "spaces", "split", "squeeze", "ssm_state", "ssm_state_size", "ssm_states", "ssm_states_reshaped", "state", "state_decay_out", "state_decay_out_permuted", "state_size", "states", "sum", "super", "swish", "t", "the", "time_step_limit", "time_step_max", "time_step_min", "time_step_rank", "to", "torch", "torch_forward", "training", "transpose", "type", "update_conv_state", "update_ssm_state", "use_bias", "use_conv_bias", "variance_epsilon", "view", "warning_once", "weight", "x", "y", "z", "zeros_like"], "mamba2/modeling_mamba2.py:Mamba2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "mamba2/modeling_mamba2.py:Mamba2Block": ["GradientCheckpointingLayer", "LongTensor", "ModelBlock", "ModelCache", "ModelMixer", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "cache_params", "cache_position", "class", "config", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "if", "layer_idx", "layer_norm_epsilon", "mixer", "norm", "residual", "residual_in_fp32", "return", "self", "super", "to", "torch", "weight"], "mamba2/modeling_mamba2.py:Mamba2PreTrainedModel": ["A", "A_log", "D", "Embedding", "False", "Linear", "ModelBlock", "ModelConfig", "ModelMixer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRMSNormGated", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_reinit", "_no_split_modules", "a", "arange", "backbone", "base_model_prefix", "bias", "clamp", "class", "config", "conv1d", "copy_", "data", "def", "dt", "dt_bias", "elif", "exp", "expm1", "fill_", "getattr", "if", "init", "initializer_range", "inv_dt", "is", "isinstance", "kaiming_uniform_", "log", "math", "min", "module", "nn", "normal_", "not", "num_heads", "num_hidden_layers", "out_proj", "p", "rand", "rescale_prenorm_residual", "self", "sqrt", "std", "supports_gradient_checkpointing", "time_step_floor", "time_step_max", "time_step_min", "torch", "weight", "zeros_"], "mamba2/modeling_mamba2.py:Mamba2Output": ["FloatTensor", "ModelCache", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "mamba2/modeling_mamba2.py:Mamba2CausalLMOutput": ["FloatTensor", "ModelCache", "ModelCausalLMOutput", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "mamba2/modeling_mamba2.py:Mamba2Model": ["Embedding", "False", "LongTensor", "ModelBlock", "ModelCache", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_register_load_state_dict_pre_hook", "a", "all_hidden_states", "and", "arange", "are", "args", "attention_mask", "auto_docstring", "automatically", "be", "because", "bool", "break", "cache_params", "cache_position", "case", "class", "config", "conv_kernel", "def", "device", "don", "dtype", "elif", "else", "embedding", "embeddings", "eps", "exactly", "for", "forward", "get_input_embeddings", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "initialized", "input_ids", "inputs_embeds", "is", "it", "k", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layers", "load_hook", "manually", "mixer_block", "must", "new_embeddings", "nn", "norm_f", "not", "num_hidden_layers", "of", "one", "or", "output_hidden_states", "pass", "passed", "pop", "post_init", "prefilling", "prefix", "r", "raise", "range", "replace", "return", "return_dict", "self", "set_input_embeddings", "size", "specify", "stage", "state_dict", "super", "t", "that", "the", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "vocab_size", "when", "will", "you"], "mamba2/modeling_mamba2.py:Mamba2ForCausalLM": ["False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCache", "ModelCausalLMOutput", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "arange", "attention_mask", "auto_docstring", "backbone", "bias", "bool", "cache_params", "cache_position", "class", "config", "contiguous", "conv_kernel", "def", "device", "dtype", "else", "float", "for", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "lm_head", "logits", "loss", "loss_function", "max_batch_size", "model_inputs", "new_embeddings", "nn", "not", "output", "output_hidden_states", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "set_input_embeddings", "size", "super", "to", "torch", "tuple", "unsqueeze", "update", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "mvp/modeling_mvp.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "mvp/modeling_mvp.py:MvpLearnedPositionalEmbedding": ["Embedding", "ModelLearnedPositionalEmbedding", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "else", "embedding_dim", "expand", "forward", "if", "input_ids", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "seq_len", "shape", "super", "torch", "unsqueeze", "weight"], "mvp/modeling_mvp.py:MvpAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_prompt", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "cat", "class", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "device", "dim", "divisible", "dropout", "else", "embed_dim", "expand", "f", "float", "for", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "mask", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "prompt_mask", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "should", "single", "size", "softmax", "src_len", "super", "tgt_len", "to", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view", "weights", "zeros"], "mvp/modeling_mvp.py:MvpEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_prompt", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "layer_head_mask", "max", "min", "nn", "num_heads", "or", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_prompt", "super", "torch", "training", "tuple"], "mvp/modeling_mvp.py:MvpDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_prompt", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_prompt", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_prompt", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "mvp/modeling_mvp.py:MvpClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "Tensor", "__init__", "class", "def", "dense", "dropout", "float", "forward", "hidden_states", "inner_dim", "input_dim", "int", "nn", "num_classes", "out_proj", "p", "pooler_dropout", "return", "self", "super", "tanh", "torch"], "mvp/modeling_mvp.py:MvpPrompt": ["Dropout", "Embedding", "GELU", "Linear", "ModelPrompt", "Module", "Sequential", "Tensor", "__init__", "class", "config", "d_model", "def", "dropout", "forward", "head_dim", "nn", "num_heads", "num_layers", "p", "permute", "prompt", "prompt_embedding", "prompt_ids", "prompt_length", "prompt_mid_dim", "prompt_trans", "return", "self", "split", "super", "torch", "tuple", "view"], "mvp/modeling_mvp.py:MvpPreTrainedModel": ["Embedding", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "attention_mask", "base_model_prefix", "bias", "class", "config", "data", "def", "device", "dummy_inputs", "elif", "if", "init_std", "input_ids", "is", "isinstance", "mean", "model", "module", "ne", "nn", "normal_", "not", "pad_token", "pad_token_id", "padding_idx", "property", "return", "self", "std", "supports_gradient_checkpointing", "tensor", "torch", "weight", "zero_"], "mvp/modeling_mvp.py:MvpEncoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelPrompt", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "and", "arange", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_heads", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "input", "input_ids", "input_shape", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "prompt_ids", "prompt_length", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "self_attn_prompt", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_prompt", "use_return_dict", "v", "view", "vocab_size"], "mvp/modeling_mvp.py:MvpDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelPrompt", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cross_attn_prompt", "d_model", "decoder_attention_heads", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "prompt_ids", "prompt_length", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attn_prompt", "shape", "should", "size", "specified", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_prompt", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "mvp/modeling_mvp.py:MvpModel": ["BaseModelOutput", "Cache", "Embedding", "False", "FloatTensor", "If", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Seq2SeqModelOutput", "Tensor", "True", "Union", "ValueError", "__init__", "_keys_to_ignore_on_load_unexpected", "_tied_weights_keys", "and", "are", "assert", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache_position", "cannot", "class", "config", "cross_attentions", "cross_attn_head_mask", "cross_attn_prompt", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "either", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "final_logits_bias", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "lightweight", "list", "make", "nn", "no", "not", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "pass", "passed", "past_key_values", "post_init", "r", "raise", "requires_grad_", "return", "return_dict", "self", "self_attn_prompt", "set_input_embeddings", "set_lightweight_tuning", "shared", "shift_tokens_right", "super", "sure", "that", "to", "torch", "tuning", "tuple", "use", "use_cache", "use_prompt", "use_return_dict", "value", "vocab_size", "want", "weight", "you"], "mvp/modeling_mvp.py:MvpForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_resize_final_logits_bias", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "list", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "register_buffer", "requires_grad_", "resize_token_embeddings", "return", "return_dict", "self", "set_lightweight_tuning", "shape", "shared", "shift_tokens_right", "since", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "mvp/modeling_mvp.py:MvpForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "ModelClassificationHead", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "class", "classification_head", "classifier_dropout", "config", "cross_attentions", "cross_attn_head_mask", "currently", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_states", "if", "input", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "len", "list", "logits", "long", "loss", "loss_fct", "model", "multi_label_classification", "must", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "requires_grad_", "return", "return_dict", "same", "self", "sentence_representation", "set_lightweight_tuning", "single_label_classification", "size", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "mvp/modeling_mvp.py:MvpForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqQuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "list", "logits", "loss", "loss_fct", "model", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "qa_outputs", "r", "requires_grad_", "return", "return_dict", "self", "sequence_output", "set_lightweight_tuning", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "weight"], "mvp/modeling_mvp.py:MvpDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "mvp/modeling_mvp.py:MvpForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "requires_grad_", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "set_lightweight_tuning", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "kosmos2/modeling_kosmos2.py:_expand_mask": ["None", "Optional", "Tensor", "_expand_mask", "bool", "bsz", "def", "dtype", "else", "expand", "expanded_mask", "finfo", "if", "int", "inverted_mask", "is", "mask", "masked_fill", "min", "not", "return", "size", "src_len", "tgt_len", "to", "torch"], "kosmos2/modeling_kosmos2.py:_make_causal_mask": ["None", "Size", "_make_causal_mask", "arange", "bsz", "cat", "def", "device", "dim", "dtype", "expand", "finfo", "full", "if", "input_ids_shape", "int", "mask", "mask_cond", "masked_fill_", "min", "past_key_values_length", "return", "size", "tgt_len", "to", "torch", "view", "zeros"], "kosmos2/modeling_kosmos2.py:Kosmos2ModelOutput": ["Any", "BaseModelOutputWithPooling", "Cache", "FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "def", "else", "for", "getattr", "hidden_states", "if", "image_embeds", "in", "k", "keys", "last_hidden_state", "not", "past_key_values", "projection_attentions", "r", "return", "self", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "kosmos2/modeling_kosmos2.py:Kosmos2ForConditionalGenerationModelOutput": ["Any", "BaseModelOutputWithPooling", "Cache", "FloatTensor", "ModelForConditionalGenerationModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "def", "else", "for", "getattr", "hidden_states", "if", "image_embeds", "in", "k", "keys", "logits", "loss", "not", "past_key_values", "projection_attentions", "r", "return", "self", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "kosmos2/modeling_kosmos2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "torch", "training", "transpose", "value"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoderLayer", "ModelVisionMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionTransformer": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "ValueError", "You", "__init__", "attentions", "bool", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layrnorm", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "kosmos2/modeling_kosmos2.py:Kosmos2TextSinusoidalPositionalEmbedding": ["False", "ModelTextSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "contiguous", "cos", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "detach", "device", "dim", "dtype", "else", "emb", "emb_weights", "embedding_dim", "exp", "expand", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sequence_length", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "kosmos2/modeling_kosmos2.py:KosmosTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "LayerNorm", "Linear", "ModelTextAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_attn_implementation", "add_inner_attn_layernorm", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "bool", "by", "cache_position", "class", "config", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "encoder_hidden_states", "eps", "f", "float", "forward", "get", "got", "head_dim", "hidden_states", "if", "inner_attn_ln", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "keys", "kwargs", "layer_head_mask", "layer_idx", "layer_norm_eps", "layers", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "past_key_value", "past_key_values", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "kosmos2/modeling_kosmos2.py:Kosmos2TextFFN": ["ACT2FN", "LayerNorm", "Linear", "ModelTextConfig", "ModelTextFFN", "Module", "__init__", "activation_dropout", "activation_fn", "activation_function", "class", "config", "def", "dropout", "embed_dim", "eps", "fc1", "fc2", "ffn_dim", "ffn_layernorm", "forward", "functional", "hidden_states", "layer_norm_eps", "nn", "p", "return", "self", "super", "training"], "kosmos2/modeling_kosmos2.py:Kosmos2TextBlock": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "LayerNorm", "ModelTextAttention", "ModelTextBlock", "ModelTextConfig", "ModelTextFFN", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "add_cross_attention", "add_inner_attn_layernorm", "are", "attention", "attention_dropout", "attention_heads", "attention_mask", "be", "bool", "by", "cache_position", "class", "config", "cross", "cross_attn_layer_head_mask", "cross_attn_weights", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "eps", "f", "ffn", "final_layer_norm", "forward", "functional", "has", "hasattr", "hidden_states", "if", "instantiated", "is", "is_decoder", "kwargs", "layer_head_mask", "layer_idx", "layer_norm_eps", "layers", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "passed", "past_key_value", "past_key_values", "raise", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "setting", "super", "to", "torch", "training", "tuple", "use_cache", "version", "with"], "kosmos2/modeling_kosmos2.py:Kosmos2TextTransformer": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "LayerNorm", "ModelTextBlock", "ModelTextConfig", "ModelTextSinusoidalPositionalEmbedding", "ModelTextTransformer", "Module", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_expand_mask", "_make_causal_mask", "_prepare_decoder_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "at", "attention_mask", "attentions", "attn_mask", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "combined_attention_mask", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "decoder_layer", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_dim", "embed_positions", "embed_scale", "embed_tokens", "embedding_dim", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "expanded_attn_mask", "f", "for", "forward", "forward_embedding", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "image_embeds", "image_embeds_position_mask", "img_input_mask", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "int", "is", "isinstance", "it", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "math", "max_position_embeddings", "nn", "not", "num_positions", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "positions", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "shape", "should", "size", "specified", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "kosmos2/modeling_kosmos2.py:Kosmos2PreTrainedModel": ["LayerNorm", "Linear", "ModelConfig", "ModelForConditionalGeneration", "ModelImageToTextProjection", "ModelModel", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextBlock", "ModelTextFFN", "ModelTextForCausalLM", "ModelTextModel", "ModelTextTransformer", "ModelVisionAttention", "ModelVisionEmbeddings", "ModelVisionEncoderLayer", "ModelVisionMLP", "ModelVisionModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "and", "bias", "class", "class_embedding", "config", "data", "def", "dense", "elif", "embed_dim", "embed_tokens", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "init_std", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "latent_query", "lm_head", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "padding_idx", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_config", "v_proj", "vision_config", "weight", "zero_"], "kosmos2/modeling_kosmos2.py:Kosmos2VisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "main_input_name", "model", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "return", "return_dict", "self", "super", "torch", "tuple"], "kosmos2/modeling_kosmos2.py:Kosmos2TextModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FlashAttentionKwargs", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "Unpack", "__init__", "attention_mask", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attn_head_mask", "def", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "head_mask", "image_embeds", "image_embeds_position_mask", "input_ids", "inputs_embeds", "kwargs", "model", "nn", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_cache"], "kosmos2/modeling_kosmos2.py:Kosmos2TextForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "Linear", "LongTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextForCausalLM", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "argument", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "def", "device", "dim", "dtype", "elif", "else", "embed_dim", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "in_features", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_function", "mask_len", "model", "model_inputs", "model_kwargs", "nn", "not", "out_features", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "provided", "r", "return", "return_dict", "self", "seq_len", "since", "size", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "warning", "weight", "zeros"], "kosmos2/modeling_kosmos2.py:Kosmos2ImageToTextProjection": ["False", "Linear", "ModelConfig", "ModelImageToTextProjection", "ModelTextAttention", "Module", "None", "Parameter", "__init__", "add_inner_attn_layernorm", "attention_dropout", "attention_heads", "attention_mask", "attn_weights", "cat", "class", "config", "def", "dense", "dim", "dropout", "embed_dim", "encoder_hidden_states", "expand", "features", "forward", "hidden_size", "hidden_states", "is_decoder", "key_value_states", "latent_query", "latent_query_num", "nn", "output_attentions", "past_key_values", "randn", "return", "self", "size", "super", "text_config", "torch", "unsqueeze", "vision_config", "x_attn"], "kosmos2/modeling_kosmos2.py:Kosmos2Model": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "ModelConfig", "ModelImageToTextProjection", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionModel", "Module", "None", "Optional", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "dim", "either", "else", "embed_tokens", "forward", "functional", "get_image_features", "get_input_embeddings", "have", "head_mask", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "image_to_text_projection", "input_ids", "inputs_embeds", "interpolate_pos_encoding", "is", "kwargs", "last_hidden_state", "main_input_name", "model", "nn", "normalize", "not", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "post_layernorm", "projection_attentions", "r", "raise", "return", "return_attentions", "return_dict", "self", "set_input_embeddings", "specify", "super", "text_config", "text_model", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_config", "vision_model", "vision_model_output"], "kosmos2/modeling_kosmos2.py:Kosmos2ForConditionalGeneration": ["Cache", "GenerationMixin", "LongTensor", "Make", "ModelConfig", "ModelForConditionalGeneration", "ModelForConditionalGenerationModelOutput", "ModelImageToTextProjection", "ModelPreTrainedModel", "ModelTextForCausalLM", "ModelVisionModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_tied_weights_keys", "allowed", "alongside", "and", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "dim", "either", "else", "embed_tokens", "f", "forward", "functional", "generate", "get_input_embeddings", "get_output_embeddings", "have", "head_mask", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "image_to_text_projection", "input_ids", "inputs", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_outputs", "logits", "loss", "main_input_name", "model", "new_embeddings", "nn", "no_grad", "normalize", "not", "or", "output", "output_attentions", "output_hidden_states", "pass", "passed", "past_key_values", "pixel_values", "pop", "position_ids", "post_init", "post_layernorm", "projection_attentions", "r", "raise", "return", "return_dict", "self", "set_input_embeddings", "set_output_embeddings", "specify", "super", "sure", "text_config", "text_model", "to", "torch", "tuple", "use_cache", "value", "vision_config", "vision_model", "vision_model_output", "weight", "were", "which"], "grounding_dino/modeling_grounding_dino.py:MultiScaleDeformableAttention": ["False", "ModelScaleDeformableAttention", "Module", "Tensor", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "class", "contiguous", "def", "dim", "enumerate", "flatten", "for", "forward", "functional", "grid_sample", "height", "hidden_dim", "im2col_step", "in", "int", "level_id", "level_start_index", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "self", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "value_spatial_shapes_list", "view", "width", "zeros"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "last_hidden_state_text", "last_hidden_state_vision", "r", "text_hidden_states", "torch", "tuple", "vision_hidden_states"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "decoder_attentions", "decoder_hidden_states", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "init_reference_points", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoObjectDetectionOutput": ["FloatTensor", "LongTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "decoder_attentions", "decoder_hidden_states", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "init_reference_points", "input_ids", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "grounding_dino/modeling_grounding_dino.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoConvEncoder": ["Either", "False", "ModelConvEncoder", "Module", "None", "Tensor", "True", "ValueError", "__init__", "and", "append", "backbone", "backbone_config", "backbone_kwargs", "backbone_model_type", "be", "bool", "channels", "class", "config", "create_model", "def", "elif", "else", "feature_info", "feature_map", "feature_maps", "features", "features_only", "float", "for", "forward", "functional", "if", "in", "intermediate_channel_sizes", "interpolate", "is", "layer2", "layer3", "layer4", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "or", "out", "parameter", "pixel_mask", "pixel_values", "pretrained", "provided", "raise", "replace_batch_norm", "requires_backends", "requires_grad_", "resnet", "return", "self", "shape", "should", "size", "stage", "super", "the", "timm", "to", "torch", "use_pretrained_backbone", "use_timm_backbone", "with"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoConvModel": ["ModelConvModel", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoSinePositionEmbedding": ["ModelSinePositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "config", "cos", "cumsum", "d_model", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "eps", "flatten", "float32", "floor", "forward", "math", "nn", "permute", "pi", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "positional_embedding_temperature", "return", "rounding_mode", "scale", "self", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoLearnedPositionEmbedding": ["Embedding", "ModelLearnedPositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "config", "d_model", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "grounding_dino/modeling_grounding_dino.py:build_position_encoding": ["ModelLearnedPositionEmbedding", "ModelSinePositionEmbedding", "Model_position_encoding", "Not", "ValueError", "config", "def", "elif", "else", "f", "if", "learned", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoMultiscaleDeformableAttention": ["CUDA", "F", "False", "Last", "Linear", "Make", "ModelConfig", "ModelMultiscaleDeformableAttention", "Module", "MultiScaleDeformableAttention", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "attn", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "config", "d", "d_model", "def", "dim", "dim_per_head", "dimension", "disable_custom_kernels", "divisible", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "forward", "got", "head", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "make", "masked_fill", "more", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_coordinates", "num_feature_levels", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "value", "value_proj", "view", "warn", "warnings", "which", "with", "with_pos_embed"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoTextEnhancerLayer": ["ACT2FN", "BoolTensor", "FloatTensor", "LayerNorm", "Linear", "ModelMultiheadAttention", "ModelTextEnhancerLayer", "Module", "None", "Optional", "Tensor", "True", "__init__", "activation", "activation_function", "and", "attention_mask", "attention_masks", "attention_output", "attention_weights", "class", "config", "d_model", "def", "dim", "dropout", "dtype", "else", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "finfo", "forward", "functional", "hidden_state", "hidden_states", "if", "is", "keys", "layer_norm_after", "layer_norm_before", "layer_norm_eps", "min", "nn", "num_attention_heads", "num_heads", "output_attentions", "p", "position_embeddings", "queries", "repeat", "residual", "return", "self", "self_attn", "shape", "super", "text_enhancer_dropout", "to", "torch", "training", "tuple", "values", "with_pos_embed"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoBiMultiHeadAttention": ["Attention", "BoolTensor", "F", "FloatTensor", "Linear", "ModelBiMultiHeadAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_reshape", "and", "attn_weights", "attn_weights_transposed", "batch_size", "be", "bmm", "but", "by", "clamp", "class", "config", "contiguous", "d_model", "def", "dim", "divisible", "dropout", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "f", "flatten", "float", "forward", "fusion_dropout", "got", "head_dim", "if", "inf", "int", "is", "keepdim", "masked_fill_", "max", "min", "must", "nn", "not", "num_heads", "of", "out_text_proj", "out_vision_proj", "p", "proj_shape", "raise", "repeat", "reshape", "return", "scale", "self", "seq_len", "should", "size", "softmax", "src_len", "super", "tensor", "text_attention_mask", "text_attn_output", "text_attn_probs", "text_attn_weights", "text_dim", "text_features", "text_key_states", "text_proj", "text_value_states", "tgt_len", "torch", "training", "transpose", "tuple", "values_text_proj", "values_vision_proj", "view", "vision_attention_mask", "vision_attn_output", "vision_attn_probs", "vision_attn_weights", "vision_dim", "vision_features", "vision_proj", "vision_query_states", "vision_value_states", "weights"], "grounding_dino/modeling_grounding_dino.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoFusionLayer": ["BoolTensor", "FloatTensor", "Identity", "LayerNorm", "ModelBiMultiHeadAttention", "ModelDropPath", "ModelFusionLayer", "Module", "None", "Optional", "Parameter", "True", "__init__", "attention_mask_text", "attention_mask_vision", "attn", "class", "config", "d_model", "def", "delta_t", "delta_v", "drop_path", "else", "forward", "fusion_droppath", "if", "init_values", "layer_norm_eps", "layer_norm_text", "layer_norm_vision", "nn", "ones", "requires_grad", "return", "self", "super", "text_attention_mask", "text_attn", "text_features", "text_param", "torch", "tuple", "vision_attention_mask", "vision_attn", "vision_features", "vision_param"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoDeformableLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDeformableLayer", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_attention_mask", "encoder_ffn_dim", "encoder_hidden_states", "encoder_n_points", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "layer_norm_eps", "level_start_index", "max", "min", "n_points", "nn", "num_heads", "or", "output_attentions", "p", "position_embeddings", "reference_points", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training"], "grounding_dino/modeling_grounding_dino.py:get_sine_pos_embed": ["Model_sine_pos_embed", "Tensor", "True", "arange", "bool", "cat", "cos", "def", "device", "dim", "dim_t", "div", "dtype", "exchange_xy", "flatten", "float32", "floor", "for", "if", "in", "int", "math", "num_pos_feats", "pi", "pos_tensor", "position_embeddings", "return", "rounding_mode", "scale", "shape", "sin", "sin_x", "sine_func", "split", "stack", "temperature", "torch", "x"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoEncoderLayer": ["False", "ModelDeformableLayer", "ModelEncoderLayer", "ModelFusionLayer", "ModelTextEnhancerLayer", "Module", "None", "Optional", "Tensor", "_", "__init__", "and", "arange", "attention_mask", "attention_mask_text", "attention_mask_vision", "attention_masks", "batch_size", "class", "config", "d_model", "def", "deformable_layer", "device", "else", "exchange_xy", "float", "forward", "fusion_layer", "get_sine_pos_embed", "get_text_position_embeddings", "hidden_states", "if", "int", "is", "key_padding_mask", "level_start_index", "list", "nn", "not", "num_pos_feats", "position_embeddings", "reference_points", "repeat", "return", "self", "seq_length", "shape", "spatial_shapes", "spatial_shapes_list", "super", "text_attention_mask", "text_enhanced_attn", "text_enhancer_layer", "text_features", "text_fused_attn", "text_position_embedding", "text_position_ids", "text_self_attention_masks", "torch", "tuple", "unsqueeze", "vision_deformable_attn", "vision_features", "vision_fused_attn", "vision_position_embedding"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoMultiheadAttention": ["Dropout", "False", "FloatTensor", "Linear", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_mask", "attention_probs", "attention_scores", "batch_size", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "functional", "hasattr", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "keys", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "out_proj", "output_attentions", "outputs", "permute", "queries", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "values", "view"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoDecoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "decoder_n_points", "def", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_attn_text", "encoder_attn_text_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "keys", "layer_norm_eps", "level_start_index", "n_points", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "queries", "reference_points", "residual", "return", "second_residual", "self", "self_attn", "self_attn_layer_norm", "self_attn_mask", "self_attn_weights", "spatial_shapes", "spatial_shapes_list", "super", "tensor", "text_cross_attn_weights", "text_encoder_attention_mask", "text_encoder_hidden_states", "third_residual", "torch", "training", "values", "vision_encoder_attention_mask", "vision_encoder_hidden_states", "with_pos_embed"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoContrastiveEmbedding": ["BoolTensor", "FloatTensor", "ModelContrastiveEmbedding", "Module", "None", "__init__", "class", "config", "def", "device", "float", "forward", "full", "inf", "masked_fill", "max_text_len", "new_output", "nn", "output", "return", "self", "shape", "super", "text_hidden_state", "text_token_mask", "torch", "transpose", "vision_hidden_state"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "False", "GroupNorm", "LayerNorm", "Linear", "ModelBiMultiHeadAttention", "ModelConfig", "ModelDecoder", "ModelFusionLayer", "ModelLearnedPositionEmbedding", "ModelMLPPredictionHead", "ModelMultiscaleDeformableAttention", "ModelPreTrainedModel", "None", "Parameter", "PreTrainedModel", "True", "_init_weights", "_set_gradient_checkpointing", "abs", "and", "arange", "attention_weights", "base_model_prefix", "bias", "class", "column_embeddings", "config", "constant_", "cos", "data", "def", "default_dtype", "dtype", "elif", "fill_", "for", "gain", "get_default_dtype", "gradient_checkpointing", "grid_init", "hasattr", "i", "if", "in", "init", "init_std", "int64", "is", "isinstance", "keepdim", "layers", "level_embed", "main_input_name", "math", "max", "mean", "model", "module", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "out_text_proj", "out_vision_proj", "output_proj", "padding_idx", "pi", "pixel_values", "range", "reference_points", "repeat", "row_embeddings", "sampling_offsets", "self", "sin", "stack", "std", "text_param", "text_proj", "thetas", "to", "torch", "two_stage", "uniform_", "value", "value_proj", "values_text_proj", "values_vision_proj", "view", "vision_param", "vision_proj", "weight", "with", "xavier_uniform_", "zero_"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoEncoder": ["ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelEncoderOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_attn_deformable", "all_attn_enhanced_text", "all_attn_fused_text", "all_attn_fused_vision", "all_attns", "append", "attentions", "cat", "class", "config", "def", "device", "dropout", "dtype", "else", "enc_outputs", "encoder_layer", "encoder_layers", "encoder_text_states", "encoder_vision_states", "enumerate", "float32", "for", "forward", "get_reference_points", "height", "i", "if", "ij", "in", "indexing", "int", "is", "key_padding_mask", "last_hidden_state_text", "last_hidden_state_vision", "layers", "level", "level_start_index", "linspace", "list", "meshgrid", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "r", "range", "ref", "ref_x", "ref_y", "reference_points", "reference_points_list", "reshape", "return", "return_dict", "self", "spatial_shapes", "spatial_shapes_list", "stack", "staticmethod", "super", "text_attention_mask", "text_features", "text_hidden_states", "text_position_embedding", "text_position_ids", "text_self_attention_masks", "torch", "tuple", "use_return_dict", "v", "valid_ratios", "vision_attention_mask", "vision_features", "vision_hidden_states", "vision_position_embedding", "width"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoDecoder": ["False", "Last", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelMLPPredictionHead", "ModelPreTrainedModel", "ModuleList", "None", "ValueError", "_", "__init__", "all_attns", "all_cross_attns_text", "all_cross_attns_vision", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "be", "but", "cat", "checkpoint", "class", "class_embed", "config", "create_custom_forward", "custom_forward", "d_model", "decoder_attention_heads", "decoder_layer", "decoder_layers", "def", "detach", "dim", "dropout", "dtype", "elif", "else", "enumerate", "eps", "f", "finfo", "for", "forward", "get_sine_pos_embed", "got", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs", "inputs_embeds", "intermediate", "intermediate_hidden_states", "intermediate_reference_points", "is", "last_hidden_state", "layer_norm", "layer_norm_eps", "layer_outputs", "layers", "level_start_index", "logit", "min", "module", "must", "new_reference_points", "nn", "not", "num_coordinates", "num_pos_feats", "num_queries", "of", "or", "output_attentions", "output_hidden_states", "position_embeddings", "post_init", "query_dim", "query_pos", "query_scale", "r", "raise", "range", "reference_points", "reference_points_head", "reference_points_input", "repeat", "return", "return_dict", "self", "self_attn_mask", "shape", "sigmoid", "spatial_shapes", "spatial_shapes_list", "special", "stack", "super", "text_encoder_attention_mask", "text_encoder_hidden_states", "tmp", "to", "torch", "training", "tuple", "use_return_dict", "utils", "v", "valid_ratios", "vision_encoder_attention_mask", "vision_encoder_hidden_states"], "grounding_dino/modeling_grounding_dino.py:generate_masks_with_special_tokens_and_transfer_map": ["LongTensor", "Model_masks_with_special_tokens_and_transfer_map", "SPECIAL_TOKENS", "Tensor", "True", "arange", "attention_mask", "batch_size", "bool", "col", "def", "device", "else", "eye", "for", "i", "idxs", "if", "in", "input_ids", "logical_or", "long", "nonzero", "num_token", "or", "position_ids", "previous_col", "range", "repeat", "return", "row", "shape", "special_token", "special_tokens_mask", "to", "torch", "tuple", "unsqueeze", "zeros"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoModel": ["AutoModel", "Conv2d", "Embedding", "False", "GroupNorm", "LayerNorm", "Linear", "ModelConfig", "ModelContrastiveEmbedding", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelEncoderOutput", "ModelMLPPredictionHead", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "Sequential", "Tensor", "True", "_", "__init__", "_len_sources", "add_pooling_layer", "all", "and", "append", "as_tensor", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_size", "bbox_embed", "bool", "build_position_encoding", "cat", "class", "config", "conv_encoder", "cumsum", "current_position", "d_model", "decoder", "decoder_attentions", "decoder_bbox_embed_share", "decoder_hidden_states", "decoder_outputs", "def", "delta_bbox", "detach", "device", "dim", "dtype", "elif", "else", "embedding_init_target", "enc_output", "enc_output_norm", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "encoder", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_output_bbox_embed", "encoder_output_class_embed", "encoder_outputs", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "enumerate", "expand", "feature_maps", "flatten", "float", "float32", "for", "forward", "freeze_backbone", "from_config", "functional", "gather", "generate_encoder_output_proposals", "generate_masks_with_special_tokens_and_transfer_map", "get_valid_ratio", "grid", "grid_x", "grid_y", "height", "hidden_dim", "hidden_size", "hidden_states", "i", "if", "ij", "in", "in_channels", "indexing", "inf", "init_reference_points", "input_dim", "input_ids", "input_proj_list", "input_proj_vision", "inputs_embeds", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_reference_points", "interpolate", "is", "isinstance", "keepdim", "kernel_size", "last_hidden_state", "last_hidden_state_text", "last_hidden_state_vision", "layer_norm_eps", "len", "level", "level_embed", "level_start_index", "linspace", "log", "long", "lvl_pos_embed", "lvl_pos_embed_flatten", "m", "mask", "mask_flatten", "mask_flatten_", "masked_fill", "masks", "max", "max_text_len", "meshgrid", "model", "name", "named_parameters", "new_zeros", "nn", "not", "num_backbone_outs", "num_channels", "num_feature_levels", "num_layers", "num_queries", "object_query", "object_query_embedding", "ones", "ones_like", "or", "output_attentions", "output_dim", "output_hidden_states", "output_proposals", "output_proposals_valid", "padding", "padding_mask", "param", "pixel_mask", "pixel_values", "pos_embed", "pos_l", "position_embedding", "position_embeddings", "position_embeddings_list", "position_ids", "post_init", "prod", "proposal", "proposals", "query_embeds", "query_position_embeddings", "r", "range", "reference_points", "repeat", "requires_grad_", "return", "return_dict", "scale", "self", "self_attn_mask", "shape", "sigmoid", "size", "source", "source_flatten", "spatial_shape", "spatial_shapes", "spatial_shapes_list", "stack", "stride", "sum", "super", "target", "text_attention_mask", "text_backbone", "text_config", "text_encoder_attention_mask", "text_encoder_hidden_states", "text_features", "text_hidden_states", "text_outputs", "text_position_embedding", "text_position_ids", "text_projection", "text_self_attention_masks", "text_token_mask", "to", "token_type_ids", "topk", "topk_coords_logits", "topk_logits", "topk_proposals", "torch", "transpose", "tuple", "tuple_outputs", "two_stage", "two_stage_bbox_embed_share", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_height", "valid_ratio", "valid_ratio_height", "valid_ratio_width", "valid_ratios", "valid_width", "value", "view", "vision_attention_mask", "vision_encoder_attention_mask", "vision_encoder_hidden_states", "vision_features", "vision_hidden_states", "vision_position_embedding", "weight", "width", "width_height", "zeros_like", "zip"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "grounding_dino/modeling_grounding_dino.py:build_label_maps": ["F", "FloatTensor", "LongTensor", "Model_label_maps", "None", "SPECIAL_TOKENS", "cumsum", "def", "delimiter_token_masks", "delimiter_tokens", "device", "dim", "for", "in", "input_ids", "int32", "isin", "label_group", "label_groups", "label_map", "label_maps", "logits", "max_seq_len", "num_labels", "pad", "repeat", "return", "shape", "tensor", "to", "torch", "tuple", "unique", "unique_labels", "unsqueeze", "value", "where"], "grounding_dino/modeling_grounding_dino.py:build_text_mask": ["Model_text_mask", "None", "attention_mask", "bool", "def", "device", "dtype", "logits", "return", "seq_len", "shape", "text_mask", "torch", "zeros_like"], "grounding_dino/modeling_grounding_dino.py:GroundingDinoForObjectDetection": ["BoolTensor", "FloatTensor", "LongTensor", "ModelConfig", "ModelContrastiveEmbedding", "ModelEncoderOutput", "ModelForObjectDetection", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Union", "ValueError", "_", "__init__", "_class_embed", "_tied_weights_keys", "append", "attention_mask", "auto_docstring", "auxiliary_outputs", "bbox_embed", "be", "bool", "build_label_maps", "build_text_mask", "but", "class", "class_embed", "config", "d", "d_model", "decoder", "decoder_attentions", "decoder_bbox_embed_share", "decoder_hidden_states", "decoder_layers", "def", "delta_bbox", "device", "dict", "dict_outputs", "elif", "else", "enc_outputs_class", "enc_outputs_coord_logits", "enc_text_hidden_state", "encoder_attentions", "encoder_last_hidden_state_text", "encoder_last_hidden_state_vision", "encoder_logits", "encoder_outputs", "encoder_pred_boxes", "encoder_text_hidden_states", "encoder_vision_hidden_states", "eps", "f", "for", "forward", "got", "hidden_dim", "hidden_states", "idx", "if", "in", "init_reference_points", "input_dim", "input_ids", "inter_references_points", "intermediate_hidden_states", "intermediate_reference_points", "is", "label_maps", "labels", "last_hidden_state", "level", "list", "logit", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_layers", "num_levels", "ones_like", "or", "out", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "outputs_class", "outputs_classes", "outputs_coord", "outputs_coord_logits", "outputs_coords", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "r", "raise", "range", "reference", "reference_coordinates", "return", "return_dict", "self", "shape", "shared_head", "should", "sigmoid", "special", "stack", "str", "super", "text_hidden_state", "text_mask", "text_token_mask", "token_type_ids", "torch", "tuple", "use_return_dict", "vision_hidden_state"], "bros/modeling_bros.py:BrosSpadeOutput": ["FloatTensor", "ModelOutput", "ModelSpadeOutput", "None", "Optional", "attentions", "class", "hidden_states", "initial_token_logits", "loss", "r", "subsequent_token_logits", "torch", "tuple"], "bros/modeling_bros.py:BrosPositionalEmbedding1D": ["ModelPositionalEmbedding1D", "Module", "Tensor", "__init__", "arange", "b1", "b2", "b3", "cat", "class", "config", "cos", "def", "dim", "dim_bbox_sinusoid_emb_1d", "forward", "inv_freq", "nn", "pos_emb", "pos_seq", "register_buffer", "return", "self", "seq_size", "sin", "sinusoid_inp", "size", "super", "torch", "view"], "bros/modeling_bros.py:BrosPositionalEmbedding2D": ["ModelPositionalEmbedding1D", "ModelPositionalEmbedding2D", "Module", "Tensor", "__init__", "append", "bbox", "bbox_pos_emb", "cat", "class", "config", "def", "dim", "dim_bbox", "else", "for", "forward", "i", "if", "in", "nn", "range", "return", "self", "stack", "super", "torch", "x_pos_emb", "y_pos_emb"], "bros/modeling_bros.py:BrosBboxEmbeddings": ["False", "Linear", "ModelBboxEmbeddings", "ModelPositionalEmbedding2D", "Module", "None", "Tensor", "__init__", "bbox", "bbox_pos", "bbox_pos_emb", "bbox_projection", "bbox_sinusoid_emb", "bbox_t", "bias", "class", "config", "def", "dim_bbox_projection", "dim_bbox_sinusoid_emb_2d", "forward", "nn", "return", "self", "super", "torch", "transpose"], "bros/modeling_bros.py:BrosTextEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "bros/modeling_bros.py:BrosSelfAttention": ["Dropout", "Embedding", "False", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Softmax", "Tensor", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bbox_pos_emb", "bbox_pos_scores", "bhld", "bhlr", "bhrd", "bijd", "bnid", "bnij", "class", "config", "context_layer", "contiguous", "d_head", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "long", "lrd", "math", "matmul", "max_position_embeddings", "multiple", "n_head", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "seq_length", "shape", "size", "sqrt", "super", "the", "to", "torch", "transpose", "tuple", "value", "value_layer", "view"], "bros/modeling_bros.py:BrosSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bros/modeling_bros.py:BrosAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bbox_pos_emb", "bool", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "bros/modeling_bros.py:BrosIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "bros/modeling_bros.py:BrosOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bros/modeling_bros.py:BrosLayer": ["Exception", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "bbox_pos_emb", "be", "bool", "by", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_output", "layers", "model", "not", "output", "output_attentions", "outputs", "passed", "raise", "return", "self", "self_attention_outputs", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "bros/modeling_bros.py:BrosEncoder": ["BaseModelOutputWithCrossAttentions", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bbox_pos_emb", "bool", "can_return_tuple", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "bros/modeling_bros.py:BrosPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "bros/modeling_bros.py:BrosRelationExtractor": ["Dropout", "Linear", "ModelRelationExtractor", "Module", "Parameter", "Tensor", "__init__", "axis", "backbone_hidden_size", "cat", "class", "classifier_dropout_prob", "config", "def", "drop", "dummy_node", "dummy_vec", "forward", "head_hidden_size", "hidden_size", "key", "key_layer", "matmul", "n_relations", "nn", "permute", "query", "query_layer", "relation_score", "repeat", "return", "self", "size", "super", "torch", "unsqueeze", "view", "zeros"], "bros/modeling_bros.py:BrosPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelRelationExtractor", "Module", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "dummy_node", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "bros/modeling_bros.py:BrosModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "ModelBboxEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "ModelTextEmbeddings", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bbox", "bbox_embeddings", "bbox_pos_emb", "bbox_position_embeddings", "bbox_scale", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "can_return_tuple", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "init_weights", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "prune_heads", "r", "raise", "return", "return_dict", "same", "scaled_bbox", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "word_embeddings", "zeros"], "bros/modeling_bros.py:BrosForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "True", "Union", "__init__", "_keys_to_ignore_on_load_unexpected", "attention_mask", "attentions", "auto_docstring", "bbox", "bbox_first_token_mask", "bool", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "hasattr", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pooler", "position_ids", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "bros/modeling_bros.py:BrosSpadeEEForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "ModelRelationExtractor", "ModelSpadeEEForTokenClassification", "ModelSpadeOutput", "None", "Optional", "Sequential", "Tensor", "True", "Union", "__init__", "_keys_to_ignore_on_load_unexpected", "and", "attention_mask", "attentions", "auto_docstring", "axis", "backbone_hidden_size", "batch_size", "bbox", "bbox_first_token_mask", "bool", "can_return_tuple", "cat", "class", "classifier_dropout", "config", "contiguous", "def", "device", "dtype", "else", "eye", "finfo", "forward", "hasattr", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_weights", "initial_token_classifier", "initial_token_labels", "initial_token_logits", "initial_token_loss", "input_ids", "inputs_embeds", "inv_attention_mask", "invalid_token_mask", "is", "last_hidden_states", "loss", "loss_fct", "masked_fill", "max_seq_length", "min", "n_relations", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pooler", "position_ids", "r", "return", "return_dict", "self", "self_token_mask", "shape", "squeeze", "subsequent_token_classifier", "subsequent_token_labels", "subsequent_token_logits", "subsequent_token_loss", "subsequent_token_mask", "super", "to", "token_type_ids", "torch", "transpose", "tuple", "use_return_dict", "view", "zeros"], "bros/modeling_bros.py:BrosSpadeELForTokenClassification": ["CrossEntropyLoss", "Model", "ModelModel", "ModelPreTrainedModel", "ModelRelationExtractor", "ModelSpadeELForTokenClassification", "None", "Optional", "Tensor", "TokenClassifierOutput", "True", "Union", "__init__", "_keys_to_ignore_on_load_unexpected", "attention_mask", "attentions", "auto_docstring", "axis", "backbone_hidden_size", "batch_size", "bbox", "bbox_first_token_mask", "bool", "can_return_tuple", "cat", "class", "classifier_dropout", "config", "contiguous", "def", "device", "dtype", "else", "entity_linker", "eye", "finfo", "forward", "hasattr", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_states", "logits", "loss", "loss_fct", "mask", "masked_fill", "max_seq_length", "min", "n_relations", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pooler", "position_ids", "r", "return", "return_dict", "self", "self_token_mask", "shape", "squeeze", "super", "to", "token_type_ids", "torch", "transpose", "tuple", "use_return_dict", "view", "zeros"], "qwen3/modeling_qwen3.py:Qwen3RMSNorm": ["ModelRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3/modeling_qwen3.py:Qwen3MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen3/modeling_qwen3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen3/modeling_qwen3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen3/modeling_qwen3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen3/modeling_qwen3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen3/modeling_qwen3.py:Qwen3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3/modeling_qwen3.py:Qwen3DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "qwen3/modeling_qwen3.py:Qwen3PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen3/modeling_qwen3.py:Qwen3RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3/modeling_qwen3.py:Qwen3Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_types", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "qwen3/modeling_qwen3.py:Qwen3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "qwen3/modeling_qwen3.py:Qwen3ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "qwen3/modeling_qwen3.py:Qwen3ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "qwen3/modeling_qwen3.py:Qwen3ForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "idefics/modeling_idefics.py:IdeficsBaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "idefics/modeling_idefics.py:IdeficsCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "idefics/modeling_idefics.py:expand_inputs_for_generation": ["False", "Model_inputs_for_generation", "Model_size", "Modeled_return_idx", "None", "arange", "attention_mask", "def", "device", "elif", "encoder_outputs", "get", "if", "image_attention_mask", "image_encoder_embeddings", "in", "index_select", "input_ids", "is", "is_encoder_decoder", "model_kwargs", "not", "perceiver_embeddings", "pixel_values", "repeat", "return", "shape", "to", "token_type_ids", "torch", "view"], "idefics/modeling_idefics.py:freeze_model": ["Embedding", "False", "LayerNorm", "Linear", "Model_model", "True", "and", "any", "def", "else", "for", "if", "in", "isinstance", "m", "mapping", "model", "module", "module_exceptions", "module_exceptions_mapped", "modules", "nn", "requires_grad_", "return", "t"], "idefics/modeling_idefics.py:IdeficsDecoupledEmbedding": ["Embedding", "F", "False", "Got", "ModelDecoupledEmbedding", "None", "Optional", "ValueError", "__init__", "additional_embedding", "additional_embeddings", "additional_vocab_indices", "and", "be", "bool", "class", "clone", "def", "device", "dtype", "embedding", "embedding_dim", "extra_repr", "f", "forward", "full_vector", "if", "input_ids", "input_ids_additional_vocab", "is", "kwargs", "must", "nn", "not", "num_additional_embeddings", "num_embeddings", "padding_idx", "partially_freeze", "raise", "requires_grad_", "return", "self", "str", "super", "torch", "weight", "where", "within"], "idefics/modeling_idefics.py:IdeficsDecoupledLinear": ["F", "False", "Linear", "ModelDecoupledLinear", "None", "Tensor", "True", "__init__", "additional_fc", "additional_features", "bias", "bool", "cat", "class", "def", "device", "dtype", "extra_repr", "f", "forward", "if", "in_features", "input", "int", "is", "linear", "nn", "not", "out_additional_features", "out_features", "output", "partially_freeze", "requires_grad_", "return", "self", "str", "super", "torch", "weight"], "idefics/modeling_idefics.py:IdeficsRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "extra_repr", "f", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "idefics/modeling_idefics.py:IdeficsEmbedding": ["False", "ModelEmbedding", "Module", "None", "__init__", "_set_cos_sin_cache", "arange", "base", "cat", "class", "cos", "cos_cached", "def", "device", "dim", "dtype", "einsum", "emb", "float", "forward", "freqs", "get_default_dtype", "i", "if", "ij", "int64", "inv_freq", "j", "max_position_embeddings", "max_seq_len_cached", "nn", "persistent", "register_buffer", "return", "self", "seq_len", "sin", "sin_cached", "super", "t", "to", "torch", "type_as", "x"], "idefics/modeling_idefics.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "idefics/modeling_idefics.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "idefics/modeling_idefics.py:IdeficsMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "int", "intermediate_size", "nn", "return", "self", "str", "super", "up_proj", "x"], "idefics/modeling_idefics.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "idefics/modeling_idefics.py:IdeficsAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelEmbedding", "ModelRMSNorm", "Module", "None", "Optional", "Please", "PretrainedConfig", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "_shape", "a", "and", "apply_rotary_pos_emb", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "eps", "errors", "f", "float", "forward", "functional", "got", "hasattr", "head_dim", "hidden_size", "hidden_states", "higher", "if", "int", "is", "is_causal", "is_cross_attention", "k_layer_norm", "k_proj", "key_states", "key_value_states", "kv_input_dim", "kv_len", "kv_seq_len", "kwargs", "layer_idx", "lead", "logger", "make", "max", "model", "must", "new_name", "nn", "not", "num_heads", "o_proj", "or", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "pytorch", "q_layer_norm", "q_len", "q_proj", "qk_layer_norms", "query_states", "raise", "recommended", "requires", "reshape", "return", "rms_norm_eps", "rotary_emb", "scaled_dot_product_attention", "scaling", "self", "seq_len", "shape", "sin", "size", "super", "sure", "tensor", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "version", "view", "vision_config", "warning_once", "when", "will", "without"], "idefics/modeling_idefics.py:IdeficsDecoderLayer": ["Cache", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "eps", "forward", "functional", "hidden_act", "hidden_size", "hidden_states", "input_layernorm", "int", "intermediate_size", "kwargs", "layer_idx", "mlp", "new_name", "nn", "num_attention_heads", "num_heads", "p", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "training", "version"], "idefics/modeling_idefics.py:IdeficsGatedCrossAttentionLayer": ["Alpha", "Cache", "FloatTensor", "GradientCheckpointingLayer", "Model", "ModelAttention", "ModelConfig", "ModelGatedCrossAttentionLayer", "ModelMLP", "ModelRMSNorm", "None", "NotImplementedError", "Optional", "Parameter", "Past", "Tanh", "Tensor", "TransformersKwargs", "True", "Unknown", "Unpack", "ValueError", "_", "__init__", "act_cross_attn", "act_dense", "alpha_cross_attn", "alpha_dense", "alpha_initializer", "alpha_type", "alphas_initializer_range", "and", "are", "attending", "attention", "attention_mask", "auto_docstring", "be", "class", "conditioned", "config", "correctly", "cross", "cross_attention_gate", "cross_attn", "def", "deprecate_kwarg", "dropout", "elif", "else", "eps", "f", "features", "float", "for", "forward", "functional", "gaussian", "hasattr", "hidden_act", "hidden_size", "hidden_states", "if", "image_attention_mask", "image_hidden_states", "images", "implemented", "in", "initialization", "initialized", "input_layernorm", "int", "intermediate_size", "is", "is_cross_attention", "key", "key_value_states", "kwargs", "layer_idx", "masked_fill", "mean", "mlp", "module", "new_name", "nn", "no", "normal", "not", "num_attention_heads", "num_heads", "on", "ones", "out", "p", "parameters", "past_key_value", "past_key_values", "post_attention_layernorm", "qk_layer_norms", "r", "raise", "random", "required", "residual", "return", "rms_norm_eps", "scheme", "self", "size", "states", "std", "super", "the", "to", "torch", "training", "value", "vector", "version", "visual", "which", "yet", "zero", "zeros"], "idefics/modeling_idefics.py:IdeficsPreTrainedModel": ["Conv2d", "Embedding", "False", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelGatedCrossAttentionLayer", "ModelPerceiverResampler", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionEmbeddings", "None", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "alpha_cross_attn", "alpha_dense", "alpha_initializer", "alphas_initializer_range", "attentions", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "fill_", "gaussian", "hidden_states", "if", "in", "index", "initializer_range", "is", "isinstance", "latents", "layer_name", "mean", "model", "module", "nn", "normal", "normal_", "not", "ones", "padding_idx", "random", "self", "self_attn", "std", "supports_gradient_checkpointing", "weight", "zero_", "zeros"], "idefics/modeling_idefics.py:IdeficsModel": ["Cache", "DynamicCache", "Exactly", "False", "FloatTensor", "If", "LongTensor", "ModelBaseModelOutputWithPast", "ModelConfig", "ModelDecoderLayer", "ModelDecoupledEmbedding", "ModelGatedCrossAttentionLayer", "ModelModel", "ModelPerceiverResampler", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionTransformer", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_attn_implementation", "additional_vocab_size", "and", "any", "arange", "are", "attention_mask", "auto_docstring", "batch_size", "be", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "contiguous", "create_causal_mask", "cross_attention_gate", "cross_attn_block", "cross_layer_interval", "cumsum", "decoder_layer", "def", "device", "dim", "dtype", "elif", "else", "embed_dim", "embed_tokens", "embedding_dim", "enumerate", "eps", "exactly", "for", "forward", "freeze_model", "freeze_relevant_params", "freeze_text_layers", "freeze_text_module_exceptions", "freeze_vision_layers", "freeze_vision_module_exceptions", "gated_cross_attn_layers", "get_seq_length", "gradient_checkpointing", "has", "hidden_size", "hidden_states", "i", "idx", "if", "image_attention_mask", "image_batch_size", "image_encoder_embeddings", "image_hidden_shape", "image_hidden_size", "image_hidden_states", "image_seq_len", "image_sequence_length", "image_size", "in", "input_embeds", "input_ids", "inputs_embeds", "interpolate_pos_encoding", "invert_attention_mask", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "long", "masked_fill_", "module", "module_exceptions", "must", "nn", "norm", "not", "num_additional_embeddings", "num_cross_layers", "num_embeddings", "num_hidden_layers", "num_images", "of", "one", "ones", "or", "pad_token_id", "padding_idx", "partially_freeze", "passed", "past_key_values", "past_key_values_length", "perceiver_config", "perceiver_embeddings", "perceiver_resampler", "pixel_values", "position_ids", "post_init", "r", "raise", "range", "repeat", "resampler_depth", "resampler_head_dim", "resampler_n_heads", "resampler_n_latents", "return", "rms_norm_eps", "self", "seq_length", "seq_length_with_past", "shape", "should", "size", "specify", "squeeze", "sum", "super", "text_seq_len", "to", "torch", "tuple", "unsqueeze", "use_cache", "use_resampler", "view", "vision_config", "vision_model", "vocab_size", "x"], "idefics/modeling_idefics.py:IdeficsForVisionText2Text": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "LongTensor", "ModelCausalLMOutputWithPast", "ModelDecoupledLinear", "ModelForVisionText2Text", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "_update_model_kwargs_for_generation", "additional_embedding", "additional_fc", "additional_vocab_size", "and", "assert", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "def", "dict", "dim", "else", "embed_tokens", "forward", "freeze_lm_head", "get", "get_input_embeddings", "get_output_embeddings", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "image_attention_mask", "image_encoder_embeddings", "image_hidden_states", "images_kwargs", "in", "in_features", "input_embeddings", "input_ids", "inputs_embeds", "interpolate_pos_encoding", "is", "is_encoder_decoder", "kwargs", "labels", "last_mask", "lm_head", "logits", "loss", "loss_function", "model", "model_inputs", "model_kwargs", "not", "num_additional_embeddings", "num_embeddings", "out_additional_features", "out_features", "output_embeddings", "outputs", "partially_freeze", "past_key_values", "perceiver_embeddings", "pixel_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "seq_length", "shape", "str", "super", "tie_weights", "tie_word_embeddings", "torch", "tuple", "unsqueeze", "use_cache", "use_resampler", "vision_model", "vocab_size", "weight"], "phimoe/modeling_phimoe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "phimoe/modeling_phimoe.py:PhimoeRotaryEmbedding": ["ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "ROPE_INIT_FUNCTIONS", "__init__", "and", "arange", "attention_scaling", "cat", "class", "config", "cos", "def", "default", "device", "dim", "dtype", "else", "emb", "float32", "forward", "freqs", "get", "if", "inv_freq", "is", "long_mscale", "mscale", "nn", "not", "original_max_position_embeddings", "outer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "seq_len", "short_mscale", "sin", "super", "t", "to", "torch", "type", "x"], "phimoe/modeling_phimoe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "phimoe/modeling_phimoe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "phimoe/modeling_phimoe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "phimoe/modeling_phimoe.py:PhimoeAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_shape", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "lead", "logger", "make", "math", "matmul", "max_position_embeddings", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "repeat_kv", "reshape", "return", "rope_theta", "self", "seq_len", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "tensor", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "phimoe/modeling_phimoe.py:PhimoeFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "The", "We", "_", "_flash_attention_forward", "_pre_quantization_dtype", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache_kwargs", "cache_position", "cast", "casted", "class", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "hasattr", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "k_proj", "key_states", "layer", "layer_idx", "layers", "logger", "might", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "related", "repeat_kv", "reshape", "return", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "target_dtype", "the", "this", "to", "torch", "training", "transpose", "tuple", "type", "upcasted", "update", "use_cache", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "you"], "phimoe/modeling_phimoe.py:PhimoeSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_size", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "phimoe/modeling_phimoe.py:PhimoeBlockSparseTop2MLP": ["ACT2FN", "False", "Linear", "ModelBlockSparseTop2MLP", "ModelConfig", "Module", "__init__", "act_fn", "bias", "class", "config", "current_hidden_states", "def", "ffn_dim", "forward", "hidden_act", "hidden_dim", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "w1", "w2", "w3"], "phimoe/modeling_phimoe.py:MultiplierProcessor": ["Function", "Model", "ModelProcessor", "None", "Tensor", "autograd", "backward", "class", "ctx", "def", "dim", "forward", "grad_at_output", "grad_at_scores_expanded", "index", "mask_for_one", "masked_gates", "mul", "return", "save_for_backward", "saved_tensors", "scatter_add_", "scores", "selected_experts", "src", "staticmethod", "torch"], "phimoe/modeling_phimoe.py:sparsemixer": ["Model", "MultiplierProcessor", "True", "ValueError", "abs", "add", "alpha", "apply", "be", "clamp", "concat", "def", "dim", "else", "empty_like", "equal", "exponential_", "factor", "float", "gather", "if", "index", "inf", "jitter_eps", "keepdim", "legacy_contiguous_format", "log", "logical_or", "mask_for_one", "mask_for_one_top2", "mask_logits_threshold", "masked_fill", "masked_gates", "masked_gates_top2", "masked_scores", "max", "max_ind", "max_scores", "memory_format", "min", "multiplier", "multiplier_o", "multiplier_top2", "multiplier_top2_o", "must", "no_grad", "raise", "rand_like", "return", "scatter", "scores", "selected_experts", "selected_experts_top2", "softmax", "to", "top_k", "torch", "training", "type_as", "uniform_", "unsqueeze", "with"], "phimoe/modeling_phimoe.py:PhimoeSparseMoeBlock": ["False", "Linear", "ModelBlockSparseTop2MLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "and", "batch_size", "bias", "class", "config", "continue", "current_hidden_states", "current_state", "def", "device", "dtype", "empty_like", "expert_idx", "expert_layer", "expert_mask", "experts", "ffn_dim", "final_hidden_states", "for", "forward", "functional", "gate", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "input_jitter_noise", "intermediate_size", "jitter_eps", "nn", "num_classes", "num_experts", "num_experts_per_tok", "num_local_experts", "one_hot", "permute", "range", "reshape", "return", "router_jitter_noise", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "sparsemixer", "super", "to", "top_k", "top_x", "torch", "training", "uniform_", "view", "where", "zeros"], "phimoe/modeling_phimoe.py:PhimoeDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelSparseMoeBlock", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "attention_mask", "block_sparse_moe", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "elementwise_affine", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "new_name", "nn", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "phimoe/modeling_phimoe.py:PhimoePreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "phimoe/modeling_phimoe.py:PhimoeModel": ["Attention", "AttentionMaskConverter", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "Flash", "FloatTensor", "LayerNorm", "LongTensor", "Make", "Model", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "arange", "are", "at", "attempting", "attention_mask", "attentions", "auto_docstring", "batch_size", "batched", "before", "behaviour", "bitwise_or_", "bool", "both", "cache_position", "call", "can_return_tuple", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal_attend_mask", "dim", "dtype", "either", "elementwise_affine", "else", "embed_tokens", "eps", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "generation", "get_max_cache_shape", "get_seq_length", "get_text_config", "getattr", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_padding_right", "is_sliding", "is_static_sliding_cache", "is_training", "isinstance", "item", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "lead", "left", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "may", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "padding_side", "past_key_values", "past_key_values_length", "past_seen_tokens", "perform", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "right", "rms_norm_eps", "rotary_emb", "router_logits", "same", "sdpa", "self", "seq_len", "sequence_length", "shape", "size", "sliding_attend_mask", "sliding_window", "specify", "staticmethod", "sum", "super", "sure", "target_length", "text_config", "the", "this", "time", "to", "tokenizer", "tokenizing", "torch", "training", "type", "unexpected", "unsqueeze", "use_cache", "use_sliding_window", "using_static_cache", "version", "vocab_size", "warning_once", "with", "xpu"], "phimoe/modeling_phimoe.py:PhimoeForCausalLM": ["Cache", "FloatTensor", "GenerationMixin", "If", "KV", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "after", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "aux_loss", "be", "bias", "bool", "cache", "cache_position", "can_return_tuple", "class", "config", "def", "device", "else", "encounter", "f", "forward", "generate", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "lm_head_bias", "load_balancing_loss_func", "logger", "logits", "logits_to_keep", "loss", "loss_function", "may", "method", "model", "model_inputs", "needs", "nn", "nonsensical", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "original_max_position_embeddings", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "past_length", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "recomputed", "return", "rope_scaling", "router_aux_loss_coef", "router_logits", "self", "shape", "slice", "slice_indices", "super", "th", "the", "to", "token", "torch", "use_cache", "using", "vocab_size", "warning", "weight", "you"], "phimoe/modeling_phimoe.py:PhimoeForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "pvt_v2/modeling_pvt_v2.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "pvt_v2/modeling_pvt_v2.py:PvtV2DropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "pvt_v2/modeling_pvt_v2.py:PvtV2OverlapPatchEmbeddings": ["Conv2d", "LayerNorm", "ModelConfig", "ModelOverlapPatchEmbeddings", "Module", "_", "__init__", "class", "config", "def", "else", "embeddings", "eps", "flatten", "forward", "height", "hidden_size", "hidden_sizes", "if", "int", "isinstance", "kernel_size", "layer_idx", "layer_norm", "layer_norm_eps", "nn", "num_channels", "padding", "patch_size", "patch_sizes", "pixel_values", "proj", "return", "self", "shape", "stride", "strides", "super", "transpose", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2DepthWiseConv": ["Conv2d", "ModelConfig", "ModelDepthWiseConv", "Module", "True", "__init__", "batch_size", "bias", "class", "config", "def", "dim", "dwconv", "flatten", "forward", "groups", "height", "hidden_states", "int", "nn", "num_channels", "return", "self", "seq_len", "shape", "super", "transpose", "view", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2SelfAttention": ["AdaptiveAvgPool2d", "Conv2d", "Dropout", "False", "GELU", "LayerNorm", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "Tensor", "The", "ValueError", "__init__", "a", "act", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "attn_drop", "batch_size", "bias", "bool", "class", "config", "context_layer", "def", "dim", "elif", "else", "eps", "f", "find_pruneable_heads_and_indices", "forward", "functional", "heads", "height", "hidden", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "index", "int", "is", "kernel_size", "key", "key_layer", "layer_norm", "layer_norm_eps", "len", "linear_attention", "math", "matmul", "multiple", "new_shape", "nn", "not", "num_attention_heads", "num_channels", "number", "of", "output_attentions", "outputs", "permute", "pool", "proj", "proj_drop", "prune_heads", "prune_linear_layer", "pruned_heads", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "self", "seq_len", "set", "shape", "size", "softmax", "spatial_reduction", "spatial_reduction_ratio", "sqrt", "stride", "super", "the", "torch", "transpose", "transpose_for_scores", "tuple", "union", "value", "value_layer", "view", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2ConvFeedForwardNetwork": ["ACT2FN", "Dropout", "Identity", "Linear", "ModelConfig", "ModelConvFeedForwardNetwork", "ModelDepthWiseConv", "Module", "None", "Optional", "ReLU", "Tensor", "__init__", "class", "config", "def", "dense1", "dense2", "dropout", "dwconv", "else", "forward", "height", "hidden_act", "hidden_dropout_prob", "hidden_features", "hidden_states", "if", "in_features", "int", "intermediate_act_fn", "is", "isinstance", "linear_attention", "nn", "not", "out_features", "relu", "return", "self", "str", "super", "torch", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2BlockLayer": ["False", "Identity", "LayerNorm", "ModelBlockLayer", "ModelConfig", "ModelConvFeedForwardNetwork", "ModelDropPath", "ModelSelfAttention", "Module", "Tensor", "__init__", "attention", "attention_output", "bool", "class", "config", "def", "drop_path", "else", "eps", "float", "forward", "height", "hidden_features", "hidden_size", "hidden_sizes", "hidden_states", "if", "in_features", "int", "layer_idx", "layer_norm_1", "layer_norm_2", "layer_norm_eps", "layer_output", "mlp", "mlp_hidden_size", "mlp_output", "mlp_ratio", "mlp_ratios", "nn", "num_attention_heads", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "spatial_reduction_ratio", "sr_ratios", "super", "torch", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2EncoderLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelBlockLayer", "ModelConfig", "ModelEncoderLayer", "ModelOverlapPatchEmbeddings", "ModuleList", "None", "__init__", "all_self_attentions", "append", "block", "block_idx", "block_layers", "blocks", "class", "config", "cpu", "def", "depths", "device", "drop_path", "drop_path_decays", "drop_path_rate", "else", "eps", "for", "forward", "height", "hidden_sizes", "hidden_states", "if", "in", "int", "layer_idx", "layer_norm", "layer_norm_eps", "layer_outputs", "linspace", "nn", "output_attentions", "outputs", "patch_embedding", "range", "return", "self", "sum", "super", "tolist", "torch", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2Encoder": ["BaseModelOutput", "False", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "batch_size", "bool", "class", "config", "contiguous", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "height", "hidden_states", "i", "idx", "if", "in", "is", "last_hidden_state", "layer", "layer_output", "layers", "nn", "not", "num_encoder_blocks", "output_attentions", "output_hidden_states", "outputs", "permute", "pixel_values", "range", "reshape", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "v", "width"], "pvt_v2/modeling_pvt_v2.py:PvtV2PreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "Union", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fan_out", "fill_", "groups", "if", "init", "initializer_range", "is", "isinstance", "kernel_size", "main_input_name", "math", "mean", "module", "nn", "normal_", "not", "out_channels", "pixel_values", "self", "sqrt", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight", "zero_"], "pvt_v2/modeling_pvt_v2.py:PvtV2Model": ["BaseModelOutput", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_prune_heads", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "encoder_outputs", "for", "forward", "heads", "heads_to_prune", "hidden_states", "if", "in", "is", "items", "last_hidden_state", "layer", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "prune_heads", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "pvt_v2/modeling_pvt_v2.py:PvtV2ForImageClassification": ["Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "config", "def", "dim", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "permute", "pixel_values", "post_init", "r", "reshape", "return", "return_dict", "self", "sequence_output", "shape", "super", "torch", "tuple", "use_return_dict"], "pvt_v2/modeling_pvt_v2.py:PvtV2Backbone": ["BackboneMixin", "BackboneOutput", "FloatTensor", "ModelBackbone", "ModelConfig", "ModelModel", "None", "Optional", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "enumerate", "feature_maps", "for", "forward", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "r", "return", "return_dict", "self", "stage", "stage_names", "super", "torch", "use_return_dict"], "llava_onevision/modeling_llava_onevision.py:LlavaOnevisionModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch", "video_hidden_states"], "llava_onevision/modeling_llava_onevision.py:LlavaOnevisionCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple", "video_hidden_states"], "llava_onevision/modeling_llava_onevision.py:LlavaOnevisionPreTrainedModel": ["Linear", "LlamaDecoderLayer", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "embed_std", "get_text_config", "getattr", "hidden_size", "if", "image_newline", "initializer_range", "is", "isinstance", "math", "mean", "module", "nn", "normal_", "not", "past_key_values", "self", "sqrt", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "llava_onevision/modeling_llava_onevision.py:LlavaOnevisionMultiModalProjector": ["ACT2FN", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "bias", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "image_features", "int", "isinstance", "len", "linear_1", "linear_2", "multimodal_projector_bias", "nn", "num_feature_layers", "projector_hidden_act", "return", "self", "super", "text_config", "vision_config", "vision_feature_layer"], "llava_onevision/modeling_llava_onevision.py:get_anyres_image_grid_shape": ["Model_anyres_image_grid_shape", "Tensor", "TypeError", "a", "be", "def", "either", "f", "grid_pinpoints", "height", "if", "image_size", "invalid", "isinstance", "list", "lists", "ndarray", "not", "np", "of", "or", "patch_size", "raise", "return", "select_best_resolution", "should", "tensor", "tolist", "torch", "tuple", "tuples", "type", "valid", "width"], "llava_onevision/modeling_llava_onevision.py:image_size_to_num_patches": ["Model_size", "Model_size_to_num_patches", "Tensor", "TypeError", "a", "be", "best_resolution", "def", "f", "for", "grid_pinpoints", "height", "i", "if", "in", "int", "invalid", "isinstance", "j", "list", "lists", "ndarray", "not", "np", "num_patches", "of", "or", "patch_size", "raise", "range", "return", "select_best_resolution", "should", "tolist", "torch", "tuple", "tuples", "type", "value", "width", "with"], "llava_onevision/modeling_llava_onevision.py:unpad_image": ["Model_image", "Modelded_tensor", "Tensor", "TypeError", "be", "current_aspect_ratio", "current_height", "current_width", "def", "either", "else", "f", "if", "image_size", "int", "invalid", "isinstance", "list", "ndarray", "new_height", "new_width", "not", "np", "or", "original_aspect_ratio", "original_height", "original_size", "original_width", "padding", "raise", "return", "round", "scale_factor", "shape", "should", "tensor", "tolist", "torch", "tuple", "type", "valid"], "llava_onevision/modeling_llava_onevision.py:LlavaOnevisionModel": ["AutoModel", "Cache", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "The", "True", "Union", "Unpack", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "_pixel_values_list", "all", "and", "anyres_max_", "anyres_max_9", "append", "apply_pooling", "attention_mask", "attentions", "auto_docstring", "base_image_feature", "batch_frames", "batch_num_images", "batch_size", "be", "bilinear", "bool", "cache_position", "can_return_tuple", "cat", "ceil", "channels", "class", "config", "consistent", "contiguous", "curr_height", "curr_width", "decoder", "def", "default", "device", "dim", "dimensions", "do", "dtype", "elif", "else", "embed_std", "enumerate", "exactly", "expand", "expand_as", "expect", "f", "feature_lens", "features", "flatten", "for", "forward", "frames", "from_config", "functional", "get_anyres_image_grid_shape", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_video_features", "grid_pinpoints", "height", "hidden_size", "hidden_states", "hs_pool", "if", "image", "image_feature", "image_features", "image_grid_pinpoints", "image_hidden_states", "image_idx", "image_newline", "image_num_patches", "image_size", "image_size_to_num_patches", "image_sizes", "image_sizes_videos", "image_token_id", "imsize", "in", "input_ids", "inputs_embeds", "int", "interpolate", "is", "isinstance", "kwargs", "language_model", "last_hidden_state", "layer_idx", "len", "list", "long", "masked_scatter", "match", "math", "max_num_patches", "mode", "model", "multi_modal_projector", "must", "n", "n_image_tokens", "n_video_tokens", "need_patching", "new_image_features", "nn", "not", "num_patch", "num_patch_height", "num_patch_width", "number", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pack_image_features", "pad_token_id", "past_key_values", "patch_size", "patches", "permute", "pix_val", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "r", "raise", "randn", "range", "ratio", "repeat", "reshape", "return", "return_dict", "scaled_shape", "selected_image_feature", "selected_video_feature", "self", "seq_len", "set_decoder", "set_input_embeddings", "shape", "should_patch", "size", "special_image_mask", "special_video_mask", "specify", "split", "sqrt", "str", "strip", "sum", "super", "tensor", "text_config", "the", "to", "tokens", "torch", "transpose", "tuple", "unpad_image", "unsqueeze", "use_cache", "use_return_dict", "value", "video_features", "video_hidden_states", "video_token_id", "view", "vision_aspect_ratio", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "width", "with", "zip"], "llava_onevision/modeling_llava_onevision.py:LlavaOnevisionForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_prepare_4d_causal_attention_mask_with_cache_position", "_tied_weights_keys", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_num_images", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "causal_mask", "class", "clone", "config", "decoder", "def", "device", "diagonal", "dim", "dtype", "else", "expand", "fill_value", "finfo", "forward", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_video_features", "hidden_size", "hidden_states", "if", "image_features", "image_hidden_states", "image_newline", "image_sizes", "image_sizes_videos", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "mask_length", "masked_fill", "min", "min_dtype", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "pack_image_features", "padding_mask", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "reshape", "return", "return_dict", "self", "sequence_length", "set_decoder", "set_input_embeddings", "shape", "slice", "slice_indices", "staticmethod", "str", "super", "target_length", "text_config", "to", "torch", "triu", "tuple", "use_cache", "use_return_dict", "value", "video_hidden_states", "vision_aspect_ratio", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "vipllava/modeling_vipllava.py:VipLlavaModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "vipllava/modeling_vipllava.py:VipLlavaCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "vipllava/modeling_vipllava.py:VipLlavaMultiModalProjector": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "True", "__init__", "act", "bias", "class", "config", "def", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "int", "isinstance", "len", "linear_1", "linear_2", "nn", "num_feature_layers", "projector_hidden_act", "projector_layernorm", "projector_layernorm_eps", "return", "self", "super", "text_config", "vision_config", "vision_feature_layers"], "vipllava/modeling_vipllava.py:VipLlavaPreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "past_key_values", "supports_gradient_checkpointing"], "vipllava/modeling_vipllava.py:VipLlavaModel": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "cat", "class", "config", "decoder", "def", "device", "dim", "do", "dtype", "else", "exactly", "expand_as", "f", "features", "for", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "in", "index", "input_ids", "inputs_embeds", "int", "is", "isinstance", "language_model", "last_hidden_state", "list", "lm_kwargs", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "r", "raise", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "to_tuple", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_config", "vision_feature_layers", "vision_tower"], "vipllava/modeling_vipllava.py:VipLlavaForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "lm_kwargs", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_feature_layers", "vision_tower", "vocab_size", "weight"], "pix2struct/modeling_pix2struct.py:Pix2StructLayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "pix2struct/modeling_pix2struct.py:Pix2StructVisionEmbeddings": ["Dropout", "Embedding", "Linear", "ModelConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "__init__", "class", "col_embeddings", "col_indices", "column_embedder", "config", "def", "dropout", "dropout_rate", "embeddings", "flattened_patches", "forward", "hidden_size", "long", "nn", "patch_embed_hidden_size", "patch_projection", "r", "return", "row_embedder", "row_embeddings", "row_indices", "self", "seq_len", "super", "torch"], "pix2struct/modeling_pix2struct.py:Pix2StructVisionAttention": ["False", "Linear", "ModelVisionAttention", "Module", "None", "True", "__init__", "and", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "class", "config", "contiguous", "d_kv", "def", "device", "dim", "dropout", "dtype", "elif", "finfo", "float32", "forward", "functional", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "inner_dim", "is", "is_torchdynamo_compiling", "key", "key_states", "key_value_proj_dim", "layer_head_mask", "masked_fill", "matmul", "max", "min", "n_heads", "nn", "not", "num_attention_heads", "ones", "output", "output_attentions", "outputs", "p", "position_bias", "position_bias_masked", "query", "query_states", "requires_grad", "return", "scores", "self", "seq_length", "shape", "softmax", "states", "super", "tensor", "to", "to_projection_shape", "torch", "training", "transpose", "type_as", "value", "value_states", "view", "zeros"], "pix2struct/modeling_pix2struct.py:Pix2StructVisionMlp": ["ACT2FN", "Dropout", "False", "Linear", "ModelVisionConfig", "ModelVisionMlp", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_size", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "pix2struct/modeling_pix2struct.py:Pix2StructVisionLayer": ["False", "GradientCheckpointingLayer", "ModelConfig", "ModelLayerNorm", "ModelVisionAttention", "ModelVisionLayer", "ModelVisionMlp", "None", "Optional", "Tensor", "Union", "__init__", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "layer_head_mask", "layer_norm_eps", "layer_output", "mlp", "output_attentions", "outputs", "pre_attention_layer_norm", "pre_mlp_layer_norm", "residual", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "pix2struct/modeling_pix2struct.py:Pix2StructVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "pix2struct/modeling_pix2struct.py:Pix2StructPreTrainedModel": ["Conv2d", "DUMMY_INPUTS", "DUMMY_MASK", "Embedding", "False", "In", "Linear", "Model", "ModelConfig", "ModelLayerNorm", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextDenseGatedActDense", "ModelTextModel", "None", "PreTrainedModel", "See", "ValueError", "_can_compile_fullgraph", "_init_weights", "_shift_right", "and", "be", "bias", "cat", "class", "clone", "config", "d_ff", "d_kv", "data", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "def", "defined", "dim", "docs", "dtype", "dummy_inputs", "elif", "else", "factor", "fill_", "float32", "for", "full", "has", "has_relative_attention_bias", "hasattr", "hidden_size", "if", "information", "init", "initializer_factor", "initializer_range", "input_ids", "input_mask", "is", "is_torch_fx_proxy", "isinstance", "it", "key", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "nn", "normal_", "not", "num_heads", "output", "pad_token_id", "padding_idx", "property", "query", "raise", "relative_attention_bias", "return", "self", "set", "shape", "shifted_input_ids", "std", "tensor", "text_config", "the", "to", "torch", "trunc_normal_", "usually", "value", "weight", "wi_0", "wi_1", "wo", "zero_"], "pix2struct/modeling_pix2struct.py:Pix2StructVisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "ModelLayerNorm", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionLayer", "ModelVisionModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_no_split_modules", "_prune_heads", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "dict", "dim", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flattened_patches", "float", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "head_outputs", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "main_input_name", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_projection", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "self", "sequence_output", "specify", "sum", "super", "supports_gradient_checkpointing", "to", "torch", "tuple", "use_return_dict"], "pix2struct/modeling_pix2struct.py:Pix2StructTextDenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelTextConfig", "ModelTextDenseGatedActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_size", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "pix2struct/modeling_pix2struct.py:Pix2StructTextLayerFF": ["DenseReluDense", "Dropout", "ModelLayerNorm", "ModelTextConfig", "ModelTextDenseGatedActDense", "ModelTextLayerFF", "Module", "__init__", "class", "config", "def", "dropout", "dropout_rate", "eps", "forward", "forwarded_states", "hidden_size", "hidden_states", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "pix2struct/modeling_pix2struct.py:Pix2StructTextAttention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "ModelTextAttention", "ModelTextConfig", "Module", "None", "Optional", "Please", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "hidden_size", "hidden_states", "if", "inner_dim", "int", "is", "is_cross_attention", "is_small", "is_updated", "isinstance", "key", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "list", "log", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "ones", "output", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "pruned_heads", "query", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "unsqueeze", "update", "use_cache", "used", "value", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "pix2struct/modeling_pix2struct.py:Pix2StructTextLayerSelfAttention": ["Dropout", "False", "ModelLayerNorm", "ModelTextAttention", "ModelTextLayerSelfAttention", "Module", "None", "Optional", "__init__", "attention", "attention_mask", "attention_output", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_size", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "pix2struct/modeling_pix2struct.py:Pix2StructTextLayerCrossAttention": ["Dropout", "False", "ModelLayerNorm", "ModelTextAttention", "ModelTextLayerCrossAttention", "Module", "None", "Optional", "__init__", "attention", "attention_mask", "attention_output", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_size", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "pix2struct/modeling_pix2struct.py:Pix2StructTextBlock": ["False", "GradientCheckpointingLayer", "ModelTextBlock", "ModelTextLayerCrossAttention", "ModelTextLayerFF", "ModelTextLayerSelfAttention", "None", "Optional", "True", "__init__", "and", "any", "attention_mask", "attention_outputs", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "encoder_attention_mask", "encoder_decoder_attention", "encoder_decoder_position_bias", "encoder_hidden_states", "finfo", "float16", "forward", "has_relative_attention_bias", "hidden_states", "if", "int", "is", "isinf", "key_value_states", "layer_head_mask", "layer_idx", "max", "min", "mlp", "new_name", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "return_dict", "self", "self_attention", "self_attention_outputs", "super", "torch", "use_cache", "version"], "pix2struct/modeling_pix2struct.py:Pix2StructTextModel": ["AttentionMaskConverter", "BlockMask", "Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "Linear", "LongTensor", "ModelLayerNorm", "ModelPreTrainedModel", "ModelTextBlock", "ModelTextConfig", "ModelTextModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_no_split_modules", "_prepare_4d_causal_attention_mask_with_cache_position", "_tied_weights_keys", "_unmask_unattended", "_update_causal_mask", "all_attentions", "all_cross_attentions", "all_hidden_states", "and", "any", "arange", "assert", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "both", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "diagonal", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder_attention_mask", "encoder_batch_size", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "expand", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_relative_attention_bias", "have", "head_mask", "hidden_size", "hidden_states", "i", "if", "ignore_index", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_training", "isinstance", "kwargs", "labels", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "lm_head", "logger", "logits", "loss", "loss_fct", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "mean", "min", "min_dtype", "model", "new_embeddings", "nn", "not", "npu", "num_layers", "ones", "or", "output_attentions", "output_hidden_states", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_bias", "post_init", "r", "raise", "range", "reduction", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "supports_gradient_checkpointing", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "using_compilable_cache", "v", "valid", "view", "vocab_size", "warning", "weight", "with", "xpu"], "pix2struct/modeling_pix2struct.py:Pix2StructForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "Cache", "FloatTensor", "GenerationMixin", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionModel", "Module", "None", "Optional", "Seq2SeqLMOutput", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_shift_right", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "flattened_patches", "float", "forward", "get_encoder", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_vqa", "isinstance", "labels", "last_hidden_state", "len", "lm_head", "logits", "loss", "main_input_name", "ne", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "pad_token_id", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "set_output_embeddings", "super", "text_config", "torch", "tuple", "use_cache", "use_return_dict", "vision_config", "weight"], "mobilevitv2/modeling_mobilevitv2.py:make_divisible": ["Model_divisible", "None", "Optional", "def", "divisor", "if", "int", "is", "max", "min_value", "new_value", "return", "value"], "mobilevitv2/modeling_mobilevitv2.py:clip": ["Model", "def", "float", "inf", "max", "max_val", "min", "min_val", "return", "value"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2ConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Input", "ModelConfig", "ModelConvLayer", "Module", "None", "Output", "Tensor", "True", "Union", "ValueError", "__init__", "activation", "affine", "are", "bias", "bool", "by", "channels", "class", "config", "convolution", "def", "dilation", "divisible", "elif", "else", "eps", "f", "features", "forward", "groups", "hidden_act", "if", "in_channels", "int", "is", "isinstance", "kernel_size", "momentum", "nn", "normalization", "not", "num_features", "out_channels", "padding", "padding_mode", "raise", "return", "self", "str", "stride", "super", "torch", "track_running_stats", "use_activation", "use_normalization", "zeros"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2InvertedResidual": ["False", "Invalid", "ModelConfig", "ModelConvLayer", "ModelInvertedResidual", "Module", "None", "Tensor", "ValueError", "__init__", "and", "class", "config", "conv_3x3", "def", "dilation", "else", "expand_1x1", "expand_ratio", "expanded_channels", "f", "features", "forward", "groups", "if", "in", "in_channels", "int", "kernel_size", "make_divisible", "nn", "not", "out_channels", "raise", "reduce_1x1", "residual", "return", "round", "self", "stride", "super", "torch", "use_activation", "use_residual"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2MobileNetLayer": ["ModelConfig", "ModelInvertedResidual", "ModelModelNetLayer", "Module", "ModuleList", "None", "Tensor", "__init__", "append", "class", "config", "def", "else", "features", "for", "forward", "i", "if", "in", "in_channels", "int", "layer", "layer_module", "nn", "num_stages", "out_channels", "range", "return", "self", "stride", "super", "torch"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2LinearSelfAttention": ["Dropout", "False", "ModelConfig", "ModelConvLayer", "ModelLinearSelfAttention", "Module", "None", "Tensor", "True", "__init__", "attn_dropout", "bias", "class", "config", "context_scores", "context_vector", "def", "dim", "embed_dim", "expand_as", "forward", "functional", "hidden_states", "in_channels", "int", "keepdim", "kernel_size", "key", "nn", "out", "out_channels", "out_proj", "p", "qkv", "qkv_proj", "query", "relu", "return", "self", "softmax", "split", "split_size_or_sections", "sum", "super", "torch", "use_activation", "use_normalization", "value"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2FFN": ["Dropout", "False", "ModelConfig", "ModelConvLayer", "ModelFFN", "Module", "None", "Tensor", "True", "__init__", "bias", "class", "config", "conv1", "conv2", "def", "dropout1", "dropout2", "embed_dim", "ffn_dropout", "ffn_latent_dim", "float", "forward", "hidden_states", "in_channels", "int", "kernel_size", "nn", "out_channels", "return", "self", "stride", "super", "torch", "use_activation", "use_normalization"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2TransformerLayer": ["Dropout", "GroupNorm", "ModelConfig", "ModelFFN", "ModelLinearSelfAttention", "ModelTransformerLayer", "Module", "None", "Tensor", "__init__", "attention", "attention_output", "class", "config", "def", "dropout", "dropout1", "embed_dim", "eps", "ffn", "ffn_dropout", "ffn_latent_dim", "float", "forward", "hidden_states", "int", "layer_norm_eps", "layer_output", "layernorm_1_out", "layernorm_after", "layernorm_before", "nn", "num_channels", "num_groups", "p", "return", "self", "super", "torch"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2Transformer": ["ModelConfig", "ModelTransformer", "ModelTransformerLayer", "Module", "ModuleList", "None", "Tensor", "__init__", "append", "block_idx", "class", "config", "d", "d_model", "def", "embed_dim", "ffn_dims", "ffn_latent_dim", "ffn_multiplier", "for", "forward", "hidden_states", "in", "int", "layer", "layer_module", "n_layers", "nn", "range", "return", "self", "super", "torch", "transformer_layer"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2Layer": ["False", "GradientCheckpointingLayer", "GroupNorm", "ModelConfig", "ModelConvLayer", "ModelInvertedResidual", "ModelLayer", "ModelTransformer", "None", "Tensor", "True", "__init__", "attn_unit_dim", "batch_size", "class", "cnn_out_dim", "config", "conv_1x1", "conv_kernel_size", "conv_kxk", "conv_projection", "d_model", "def", "dilation", "downsampling_layer", "else", "eps", "feature_map", "features", "fold", "folding", "forward", "functional", "groups", "if", "img_height", "img_width", "in_channels", "in_dim", "int", "kernel_size", "layer_norm_eps", "layernorm", "n_attn_blocks", "n_layers", "n_patches", "nn", "num_channels", "num_groups", "out_channels", "output_size", "patch_height", "patch_size", "patch_width", "patches", "reshape", "return", "self", "shape", "stride", "super", "torch", "transformer", "tuple", "unfold", "unfolding", "use_activation", "use_normalization"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2Encoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "ModelModelNetLayer", "Module", "ModuleList", "None", "Tensor", "True", "Union", "__init__", "all_hidden_states", "append", "attn_unit_dim", "base_attn_unit_dims", "bool", "class", "clip", "config", "def", "dilate_layer_4", "dilate_layer_5", "dilation", "divisor", "elif", "else", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "in_channels", "is", "last_hidden_state", "layer", "layer_0_dim", "layer_1", "layer_1_dim", "layer_2", "layer_2_dim", "layer_3", "layer_3_dim", "layer_4", "layer_4_dim", "layer_5", "layer_5_dim", "layer_module", "make_divisible", "max_val", "min_val", "min_value", "n_attn_blocks", "nn", "not", "num_stages", "out_channels", "output_hidden_states", "output_stride", "return", "return_dict", "self", "stride", "super", "torch", "tuple", "v", "value", "width_multiplier"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2PreTrainedModel": ["BatchNorm2d", "Conv2d", "GroupNorm", "Linear", "Model", "ModelConfig", "ModelLayer", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2Model": ["BaseModelOutputWithPoolingAndNoAttention", "False", "ModelConfig", "ModelConvLayer", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPreTrainedModel", "Model_layer", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "bool", "class", "clip", "config", "conv_stem", "def", "dim", "divisor", "else", "embedding_output", "encoder", "encoder_outputs", "expand_output", "for", "forward", "have", "heads", "heads_to_prune", "hidden_states", "if", "in", "in_channels", "is", "isinstance", "items", "keepdim", "kernel_size", "last_hidden_state", "layer", "layer_0_dim", "layer_index", "make_divisible", "max_val", "mean", "min_val", "min_value", "not", "num_channels", "out_channels", "output", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "self", "specify", "stride", "super", "to", "torch", "transformer", "transformer_layer", "tuple", "use_activation", "use_normalization", "use_return_dict", "value", "width_multiplier"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2ForImageClassification": ["Identity", "ImageClassifierOutputWithNoAttention", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "divisor", "else", "forward", "hidden_states", "if", "in_features", "is", "labels", "logits", "loss", "loss_function", "make_divisible", "nn", "not", "num_labels", "out_channels", "out_features", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "width_multiplier"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2ASPPPooling": ["AdaptiveAvgPool2d", "False", "ModelASPPPooling", "ModelConfig", "ModelConvLayer", "Module", "None", "Tensor", "True", "__init__", "align_corners", "bilinear", "class", "config", "conv_1x1", "def", "features", "forward", "functional", "global_pool", "in_channels", "int", "interpolate", "kernel_size", "mode", "nn", "out_channels", "output_size", "relu", "return", "self", "shape", "size", "spatial_size", "stride", "super", "torch", "use_activation", "use_normalization"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2ASPP": ["Dropout", "Expected", "ModelASPP", "ModelASPPPooling", "ModelConfig", "ModelConvLayer", "Module", "ModuleList", "None", "Tensor", "ValueError", "__init__", "append", "aspp_dropout_prob", "aspp_out_channels", "atrous_rates", "cat", "class", "config", "conv", "convs", "def", "dilation", "dim", "divisor", "dropout", "encoder_out_channels", "extend", "features", "for", "forward", "if", "in", "in_channels", "in_projection", "kernel_size", "len", "make_divisible", "nn", "out_channels", "p", "pool_layer", "pooled_features", "project", "pyramid", "raise", "rate", "relu", "return", "self", "super", "torch", "use_activation", "values", "width_multiplier"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2DeepLabV3": ["Dropout2d", "False", "ModelASPP", "ModelConfig", "ModelConvLayer", "ModelDeepLabV3", "Module", "None", "Tensor", "True", "__init__", "aspp", "aspp_out_channels", "bias", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "features", "forward", "hidden_states", "in_channels", "kernel_size", "nn", "num_labels", "out_channels", "return", "self", "super", "torch", "use_activation", "use_normalization"], "mobilevitv2/modeling_mobilevitv2.py:MobileViTV2ForSemanticSegmentation": ["CrossEntropyLoss", "False", "Model", "ModelConfig", "ModelDeepLabV3", "ModelForSemanticSegmentation", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SemanticSegmenterOutput", "Tensor", "The", "True", "Union", "ValueError", "__init__", "align_corners", "and", "attentions", "auto_docstring", "be", "bilinear", "bool", "class", "config", "def", "else", "encoder_hidden_states", "expand_output", "forward", "functional", "greater", "hidden_states", "if", "ignore_index", "interpolate", "is", "labels", "logits", "loss", "loss_fct", "mode", "nn", "not", "num_labels", "number", "of", "one", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "segmentation_head", "self", "semantic_loss_ignore_index", "shape", "should", "size", "super", "than", "torch", "tuple", "upsampled_logits", "use_return_dict"], "deformable_detr/modeling_deformable_detr.py:MultiScaleDeformableAttention": ["False", "ModelScaleDeformableAttention", "Module", "Tensor", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "class", "contiguous", "def", "dim", "enumerate", "flatten", "for", "forward", "functional", "grid_sample", "height", "hidden_dim", "im2col_step", "in", "int", "level_id", "level_start_index", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "self", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "value_spatial_shapes_list", "view", "width", "zeros"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrObjectDetectionOutput": ["Any", "FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "deformable_detr/modeling_deformable_detr.py:_get_clones": ["ModuleList", "N", "_get_clones", "copy", "deepcopy", "def", "for", "i", "in", "module", "nn", "range", "return"], "deformable_detr/modeling_deformable_detr.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "deformable_detr/modeling_deformable_detr.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrConvEncoder": ["Either", "False", "ModelConvEncoder", "Module", "None", "Tensor", "True", "ValueError", "__init__", "and", "append", "backbone", "backbone_config", "backbone_kwargs", "backbone_model_type", "be", "bool", "channels", "class", "config", "copy", "create_model", "def", "dilation", "elif", "else", "feature_info", "feature_map", "feature_maps", "features", "features_only", "float", "for", "forward", "functional", "get", "getattr", "if", "in", "in_chans", "intermediate_channel_sizes", "interpolate", "is", "kwargs", "layer2", "layer3", "layer4", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "num_channels", "num_feature_levels", "or", "out", "out_indices", "output_stride", "parameter", "pixel_mask", "pixel_values", "pop", "pretrained", "provided", "raise", "replace_batch_norm", "requires_backends", "requires_grad_", "resnet", "return", "self", "shape", "should", "size", "stage", "super", "the", "timm", "to", "torch", "use_pretrained_backbone", "use_timm_backbone", "with"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrConvModel": ["ModelConvModel", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrSinePositionEmbedding": ["False", "ModelSinePositionEmbedding", "Module", "No", "None", "True", "ValueError", "__init__", "and", "arange", "be", "cat", "class", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "eps", "flatten", "floor", "forward", "if", "is", "mask", "math", "nn", "normalize", "not", "passed", "permute", "pi", "pixel", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "provided", "raise", "return", "rounding_mode", "scale", "self", "should", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrLearnedPositionEmbedding": ["Embedding", "ModelLearnedPositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "deformable_detr/modeling_deformable_detr.py:build_position_encoding": ["ModelLearnedPositionEmbedding", "ModelSinePositionEmbedding", "Model_position_encoding", "Not", "True", "ValueError", "config", "d_model", "def", "elif", "else", "f", "if", "learned", "n_steps", "normalize", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrMultiscaleDeformableAttention": ["CUDA", "F", "False", "Last", "Linear", "Make", "ModelConfig", "ModelMultiscaleModelAttention", "Module", "MultiScaleModelAttention", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "attn", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "config", "d", "d_model", "def", "dim", "dim_per_head", "dimension", "disable_custom_kernels", "divisible", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "for", "forward", "got", "head", "height", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "make", "masked_fill", "more", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_coordinates", "num_feature_levels", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "total_elements", "value", "value_proj", "view", "warn", "warnings", "which", "width", "with", "with_pos_embed"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrMultiheadAttention": ["Attention", "False", "Linear", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_prepare_4d_attention_mask", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "k_proj", "key_states", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderLayer", "ModelMultiscaleModelAttention", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_attention_mask", "encoder_ffn_dim", "encoder_hidden_states", "encoder_n_points", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "level_start_index", "max", "min", "n_points", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelMultiheadAttention", "ModelMultiscaleModelAttention", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "decoder_n_points", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "level_start_index", "n_points", "nn", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "second_residual", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "Linear", "ModelConfig", "ModelConvEncoder", "ModelDecoderLayer", "ModelEncoderLayer", "ModelLearnedPositionEmbedding", "ModelMultiscaleModelAttention", "ModelPreTrainedModel", "None", "Parameter", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "abs", "and", "arange", "attention_weights", "base_model_prefix", "bias", "class", "column_embeddings", "config", "constant_", "cos", "data", "def", "default_dtype", "dtype", "elif", "for", "gain", "get_default_dtype", "grid_init", "hasattr", "i", "if", "in", "init", "init_std", "int64", "is", "isinstance", "keepdim", "level_embed", "main_input_name", "math", "max", "mean", "model", "module", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "output_proj", "padding_idx", "pi", "pixel_values", "r", "range", "reference_points", "repeat", "row_embeddings", "sampling_offsets", "self", "sin", "stack", "std", "supports_gradient_checkpointing", "thetas", "to", "torch", "two_stage", "uniform_", "value_proj", "view", "weight", "with", "xavier_uniform_", "zero_"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "all_attentions", "append", "attention_mask", "attentions", "cat", "class", "config", "def", "device", "dropout", "dtype", "else", "encoder_layer", "encoder_layers", "encoder_states", "enumerate", "for", "forward", "functional", "get_reference_points", "gradient_checkpointing", "height", "hidden_states", "i", "if", "ij", "in", "indexing", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "level", "level_start_index", "linspace", "meshgrid", "nn", "not", "output_attentions", "output_hidden_states", "p", "position_embeddings", "post_init", "r", "range", "ref", "ref_x", "ref_y", "reference_points", "reference_points_list", "reshape", "return", "return_dict", "self", "spatial_shapes", "spatial_shapes_list", "spatial_shapes_tuple", "stack", "staticmethod", "super", "torch", "training", "tuple", "use_return_dict", "v", "valid_ratios", "width"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrDecoder": ["False", "Last", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelPreTrainedModel", "ModuleList", "None", "Reference", "ValueError", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "be", "but", "cat", "class", "class_embed", "config", "cross_attentions", "decoder_layer", "decoder_layers", "def", "detach", "dim", "dimension", "dropout", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "got", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "intermediate", "intermediate_hidden_states", "intermediate_reference_points", "inverse_sigmoid", "is", "last", "last_hidden_state", "layer_outputs", "layers", "level_start_index", "must", "new_reference_points", "nn", "not", "num_coordinates", "of", "or", "output_attentions", "output_hidden_states", "points", "position_embeddings", "post_init", "r", "raise", "range", "reference_points", "reference_points_input", "return", "return_dict", "self", "shape", "sigmoid", "size", "spatial_shapes", "spatial_shapes_list", "stack", "super", "tmp", "torch", "tuple", "use_return_dict", "v", "valid_ratios"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrModel": ["BaseModelOutput", "Conv2d", "Embedding", "False", "FloatTensor", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModuleList", "No", "None", "Optional", "Parameter", "Sequential", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_cur", "_len_sources", "all", "and", "append", "arange", "as_tensor", "attention", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_size", "bbox_embed", "bool", "build_position_encoding", "cat", "class", "class_embed", "config", "conv_encoder", "cos", "cross_attentions", "cumsum", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "delta_bbox", "detach", "device", "dim", "dim_t", "div", "dtype", "elif", "else", "enc_output", "enc_output_norm", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "enumerate", "expand", "features", "flatten", "float", "float32", "floor", "for", "forward", "freeze_backbone", "functional", "gather", "gen_encoder_output_proposals", "get_encoder", "get_proposal_pos_embed", "get_valid_ratio", "grid", "grid_x", "grid_y", "height", "hidden_states", "if", "ij", "in", "in_channels", "indexing", "inf", "init_reference_points", "input_proj", "input_proj_list", "inputs_embeds", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_reference_points", "interpolate", "is", "isinstance", "keepdim", "kernel_size", "last_hidden_state", "len", "level", "level_embed", "level_start_index", "linspace", "log", "long", "lvl_pos_embed", "lvl_pos_embed_flatten", "m", "mask", "mask_flatten", "mask_flatten_", "masked_fill", "masks", "math", "meshgrid", "model", "name", "named_parameters", "new_zeros", "nn", "not", "num_backbone_outs", "num_channels", "num_feature_levels", "num_pos_feats", "num_queries", "object_query", "object_query_embedding", "ones", "ones_like", "output_attentions", "output_hidden_states", "output_proposals", "output_proposals_valid", "padding", "padding_mask", "param", "pi", "pixel_mask", "pixel_values", "pos", "pos_embed", "pos_l", "pos_trans", "pos_trans_norm", "pos_trans_out", "position_embedding", "position_embeddings", "position_embeddings_list", "post_init", "prod", "proposal", "proposals", "provided", "query_embed", "query_embeds", "query_position_embeddings", "r", "raise", "range", "reference_points", "repeat", "requires_grad_", "return", "return_dict", "rounding_mode", "scale", "self", "shape", "sigmoid", "sin", "size", "source", "source_flatten", "sources", "spatial_shape", "spatial_shapes", "spatial_shapes_list", "split", "stack", "stride", "sum", "super", "target", "temperature", "to", "topk", "topk_coords_logits", "topk_proposals", "torch", "transpose", "tuple", "tuple_outputs", "two_stage", "two_stage_num_proposals", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_height", "valid_ratio", "valid_ratio_height", "valid_ratio_width", "valid_ratios", "valid_width", "value", "view", "was", "weight", "width", "width_height", "zip"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "deformable_detr/modeling_deformable_detr.py:DeformableDetrForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForObjectDetection", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Union", "ValueError", "_", "__init__", "_get_clones", "_no_split_modules", "_tied_weights_keys", "append", "auto_docstring", "auxiliary_outputs", "bbox_embed", "be", "bool", "but", "class", "class_embed", "config", "cross_attentions", "d", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_layers", "def", "delta_bbox", "device", "dict", "dict_outputs", "elif", "else", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "for", "forward", "got", "hidden_dim", "hidden_states", "if", "in", "init_reference", "init_reference_points", "input_dim", "inputs_embeds", "inter_references", "intermediate_hidden_states", "intermediate_reference_points", "inverse_sigmoid", "is", "labels", "last_hidden_state", "level", "list", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_labels", "num_layers", "num_pred", "or", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "outputs_class", "outputs_classes", "outputs_coord", "outputs_coord_logits", "outputs_coords", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "r", "raise", "range", "reference", "return", "return_dict", "self", "shape", "should", "sigmoid", "stack", "super", "torch", "tuple", "tuple_outputs", "two_stage", "use_return_dict", "with_box_refine"], "encoder_decoder/modeling_encoder_decoder.py:shift_tokens_right": ["Make", "Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "attribute", "clone", "configuration", "decoder_start_token_id", "def", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "of", "pad_token_id", "raise", "return", "s", "set", "shape", "sure", "the", "to", "torch"], "encoder_decoder/modeling_encoder_decoder.py:EncoderDecoderModel": ["AutoConfig", "AutoModel", "AutoModelForCausalLM", "BaseModelOutput", "BoolTensor", "Cache", "Config", "Cross", "CrossEntropyLoss", "DEPRECATION_WARNING", "Decoder", "Either", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "GitHub", "Got", "Head", "If", "In", "Initializing", "LM", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "Model_", "Model_attention_mask", "Model_attentions", "Model_config", "Model_hidden_states", "Model_last_hidden_state", "Model_model", "Model_outputs", "Model_pretrained_model_name_or_path", "None", "NotImplementedError", "Optional", "Please", "PreTrainedModel", "PretrainedConfig", "Resizing", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "ValueError", "__class__", "__init__", "_attn_implementation", "_dynamic_tied_weights_keys", "_init_weights", "_modules", "_supports_flash_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "_tie_Model_weights", "a", "add_cross_attention", "added", "allows", "an", "and", "architecture", "are", "args", "argument", "as", "attention", "attention_mask", "attentions", "attributes", "auto_docstring", "base_model_prefix", "be", "bool", "by", "cache_position", "can_return_tuple", "causal", "class", "classmethod", "cls", "com", "config", "config_class", "configuration", "cross", "cross_attention_hidden_size", "cross_attentions", "decoder", "decoder_", "decoder_attention_mask", "decoder_attentions", "decoder_base_model_prefix", "decoder_config", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_model", "decoder_outputs", "decoder_pretrained_model_name_or_path", "decoder_signature", "decoder_start_token_id", "def", "defined", "del", "directly", "disabled", "discussion", "do", "elif", "else", "embedding", "enc_to_dec_proj", "equal", "f", "following", "for", "forward", "from", "from_Model_configs", "from_Model_pretrained", "from_config", "from_pretrained", "get_Model", "get_input_embeddings", "get_output_embeddings", "github", "has", "have", "hidden", "hidden_size", "hidden_states", "https", "huggingface", "if", "in", "info", "initialize", "initialized", "input_ids", "inputs_embeds", "inspect", "is", "is_decoder", "isinstance", "issues", "it", "items", "key", "keys", "kwargs", "kwargs_Model", "kwargs_decoder", "kwargs_shared", "labels", "last_hidden_state", "layers", "len", "logger", "logits", "loss", "loss_fct", "main_input_name", "make", "mask", "methods", "model", "model_args", "module", "modules", "new_embeddings", "new_tensor", "nn", "not", "num_items_in_batch", "objects", "of", "on", "or", "order", "output_attentions", "output_hidden_states", "overwritten", "pad_token_id", "parameters", "pass", "passed", "past_key_values", "pop", "prepare_decoder_input_ids_from_labels", "prepared", "provided", "r", "raise", "randomly", "reshape", "resize_token_embeddings", "respective", "return", "return_dict", "return_unused_kwargs", "s", "see", "selected", "self", "set", "set_output_embeddings", "shared", "shift_tokens_right", "should", "signature", "specified", "startswith", "states", "str", "super", "supported", "supports_gradient_checkpointing", "sure", "that", "the", "tie_Model", "tie_weights", "tied_weights", "to", "to_dict", "torch", "transformers", "tuple", "type", "use", "use_cache", "value", "via", "view", "vocab_size", "warn", "warning", "warnings", "without", "wrapped"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapanesePreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelLayer", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "base_model_prefix", "bias", "class", "config", "data", "def", "dense_bias", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "weight", "zero_"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapaneseAttention": ["Cache", "Dropout", "False", "FloatTensor", "Instantiating", "Linear", "LongTensor", "Model", "ModelRotaryEmbedding", "Module", "None", "Optional", "Parameter", "Please", "Tensor", "__class__", "__init__", "__name__", "_attn", "_merge_heads", "_split_heads", "a", "alpha", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attention_scores", "attn_head_size", "attn_output", "attn_scores", "attn_weights", "baddbmm", "batch_size", "beta", "bias", "bool", "cache_kwargs", "cache_position", "caching", "call", "cat", "causal_mask", "class", "classmethod", "cls", "config", "contiguous", "cos", "creating", "def", "dense", "dense_bias", "device", "dim", "dtype", "during", "else", "errors", "f", "forward", "functional", "head_mask", "head_size", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_length", "key_pass", "key_rot", "layer_idx", "layer_past", "lead", "logger", "make", "math", "matmul", "new_qkv_shape", "new_shape", "nn", "norm_factor", "not", "num_attention_heads", "output_attentions", "partial_rotation_size", "passing", "permute", "position_embeddings", "position_ids", "provide", "qkv", "query", "query_key_value", "query_length", "query_pass", "query_rot", "recommended", "return", "rope_theta", "rotary_emb", "rotary_emb_base", "rotary_ndims", "rotary_pct", "self", "shape", "sin", "size", "softmax", "sqrt", "super", "sure", "tensor", "the", "this", "to", "torch", "transpose", "tuple", "update", "use_bias", "use_cache", "used", "value", "view", "warning_once", "when", "will", "without", "zeros"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapaneseRotaryEmbedding": ["False", "Model", "ModelConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:bias_dropout_add": ["Model", "Model_dropout_add", "None", "Optional", "Tensor", "bool", "def", "dropout", "float", "functional", "if", "is", "nn", "not", "out", "p", "prob", "residual", "return", "torch", "training", "x"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapaneseMLP": ["ACT2FN", "False", "Linear", "Model", "Module", "__init__", "act", "bias", "class", "config", "def", "dense_4h_to_h", "dense_h_to_4h", "forward", "hidden_act", "hidden_size", "hidden_states", "int", "intermediate", "intermediate_multiple_size", "intermediate_size", "nn", "output", "return", "self", "super"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapaneseLayer": ["Cache", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelAttention", "ModelMLP", "Module", "None", "Optional", "Tensor", "__init__", "attention", "attention_mask", "attn_bias", "attn_output", "attn_weights", "bias", "bias_dropout_add", "bool", "cache_position", "class", "config", "def", "else", "eps", "expand_as", "forward", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "is", "layer_idx", "layer_norm_eps", "layer_number", "layer_past", "ln_out", "mlp", "mlp_output", "nn", "not", "num_hidden_layers", "output_attentions", "position_embeddings", "position_ids", "post_attention_layernorm", "prob", "residual", "return", "self", "super", "torch", "training", "tuple", "use_bias", "use_cache"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapaneseModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelLayer", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_attentions", "all_hidden_states", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "dtype", "else", "embed_in", "enumerate", "eps", "exactly", "expand", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layer_number", "layer_past", "layers", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "nn", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "reshape", "return", "return_dict", "rotary_emb", "sdpa", "self", "seq_length", "sequence_length", "set_input_embeddings", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "value", "vocab_size", "xpu"], "gpt_neox_japanese/modeling_gpt_neox_japanese.py:GPTNeoXJapaneseForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "def", "device", "else", "embed_out", "forward", "get_output_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_logits", "lm_loss", "logits", "loss", "loss_function", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "videomae/modeling_videomae.py:VideoMAEDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "r", "torch", "tuple"], "videomae/modeling_videomae.py:VideoMAEForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "videomae/modeling_videomae.py:get_sinusoid_encoding_table": ["FloatTensor", "Model_position_angle_vec", "Model_sinusoid_encoding_table", "array", "cos", "d_hid", "def", "for", "hid_j", "in", "n_position", "np", "pos_i", "position", "power", "range", "return", "sin", "sinusoid_table", "torch", "unsqueeze"], "videomae/modeling_videomae.py:VideoMAEEmbeddings": ["ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "True", "_", "__init__", "batch_size", "bool_masked_pos", "class", "config", "copy", "def", "detach", "device", "embeddings", "forward", "get_sinusoid_encoding_table", "hidden_size", "if", "is", "nn", "not", "num_channels", "num_patches", "patch_embeddings", "pixel_values", "position_embeddings", "reshape", "return", "self", "shape", "super", "to", "type_as"], "videomae/modeling_videomae.py:VideoMAEPatchEmbeddings": ["Conv3d", "Input", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "flatten", "forward", "height", "hidden_size", "if", "image", "image_size", "in", "in_channels", "int", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_frames", "num_patches", "of", "one", "or", "out_channels", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "transpose", "tubelet_size", "values", "width", "with"], "videomae/modeling_videomae.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "videomae/modeling_videomae.py:VideoMAESelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "The", "ValueError", "_", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "functional", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "input", "int", "is", "is_causal", "k_bias", "key", "key_layer", "keys", "linear", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "q_bias", "qkv_bias", "queries", "query", "query_layer", "raise", "requires_grad", "reshape", "return", "scaling", "self", "seq_length", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "v_bias", "value", "value_layer", "values", "view", "weight", "zeros", "zeros_like"], "videomae/modeling_videomae.py:VideoMAESelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "videomae/modeling_videomae.py:VideoMAEAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "videomae/modeling_videomae.py:VideoMAEIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "videomae/modeling_videomae.py:VideoMAEOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "videomae/modeling_videomae.py:VideoMAELayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "videomae/modeling_videomae.py:VideoMAEEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "videomae/modeling_videomae.py:VideoMAEPreTrainedModel": ["Conv3d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "videomae/modeling_videomae.py:VideoMAEModel": ["BaseModelOutput", "BoolTensor", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_prune_heads", "attention", "auto_docstring", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "nn", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "post_init", "prune_heads", "r", "return", "self", "sequence_output", "super", "torch", "use_mean_pooling"], "videomae/modeling_videomae.py:VideoMAEDecoder": ["False", "Identity", "LayerNorm", "Linear", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelLayer", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "class", "config", "decoder_config", "decoder_hidden_size", "decoder_intermediate_size", "decoder_layers", "decoder_num_attention_heads", "decoder_num_hidden_layers", "decoder_num_labels", "deepcopy", "def", "else", "for", "forward", "gradient_checkpointing", "head", "head_mask", "hidden_size", "hidden_states", "if", "in", "int", "intermediate_size", "layer_module", "logits", "nn", "norm", "num_attention_heads", "num_channels", "num_hidden_layers", "patch_size", "range", "return", "return_token_num", "self", "super", "torch", "tubelet_size"], "videomae/modeling_videomae.py:VideoMAEForPreTraining": ["BaseModelOutput", "BoolTensor", "Can", "Consider", "False", "FloatTensor", "IMAGENET_DEFAULT_MEAN", "IMAGENET_DEFAULT_STD", "Linear", "MSELoss", "Model", "ModelDecoder", "ModelDecoderOutput", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "Models_patch", "None", "One", "Optional", "Parameter", "RGB", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "as_tensor", "attentions", "auto_docstring", "batch_size", "bias", "bool_masked_pos", "boolean", "can_return_tuple", "cat", "class", "config", "contiguous", "copy", "decoder", "decoder_hidden_size", "decoder_outputs", "def", "detach", "device", "dim", "dtype", "else", "embeddings", "encoder_to_decoder", "expand", "expanded_position_embeddings", "forward", "frames", "frames_norm", "get_sinusoid_encoding_table", "head_mask", "height", "hidden_size", "hidden_states", "if", "images", "is", "keepdim", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_fct", "mask", "mask_token", "mean", "must", "nn", "no_grad", "non", "norm_pix_loss", "num_channels", "num_patches", "outputs", "patch_size", "permute", "pixel_values", "pos_emb_mask", "pos_emb_visible", "position_embeddings", "post_init", "provided", "r", "raise", "reshape", "return", "self", "sequence_output", "setting", "shape", "sqrt", "std", "super", "t", "time", "to", "torch", "tubelet_size", "type_as", "unbiased", "unnormalize", "var", "view", "width", "with", "x_full", "zeros"], "videomae/modeling_videomae.py:VideoMAEForVideoClassification": ["BaseModelOutput", "Identity", "ImageClassifierOutput", "LayerNorm", "Linear", "Model", "ModelForModelClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "else", "fc_norm", "forward", "head_mask", "hidden_size", "hidden_states", "if", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "output", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch", "use_mean_pooling"], "regnet/modeling_regnet.py:RegNetConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Identity", "ModelConvLayer", "Module", "None", "Optional", "__init__", "activation", "bias", "class", "convolution", "def", "else", "forward", "groups", "hidden_state", "if", "in_channels", "int", "is", "kernel_size", "nn", "normalization", "not", "out_channels", "padding", "relu", "return", "self", "str", "stride", "super"], "regnet/modeling_regnet.py:RegNetEmbeddings": ["Make", "ModelConfig", "ModelConvLayer", "ModelEmbeddings", "Module", "ValueError", "__init__", "activation", "channel", "class", "config", "configuration", "def", "dimension", "embedder", "embedding_size", "forward", "hidden_act", "hidden_state", "if", "in", "kernel_size", "match", "nn", "num_channels", "of", "one", "pixel", "pixel_values", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "values", "with"], "regnet/modeling_regnet.py:RegNetShortCut": ["BatchNorm2d", "Conv2d", "False", "ModelShortCut", "Module", "Tensor", "__init__", "bias", "class", "convolution", "def", "forward", "hidden_state", "in_channels", "input", "int", "kernel_size", "nn", "normalization", "out_channels", "return", "self", "stride", "super"], "regnet/modeling_regnet.py:RegNetSELayer": ["AdaptiveAvgPool2d", "Conv2d", "ModelSELayer", "Module", "ReLU", "Sequential", "Sigmoid", "__init__", "attention", "class", "def", "forward", "hidden_state", "in_channels", "int", "kernel_size", "nn", "pooled", "pooler", "reduced_channels", "return", "self", "super"], "regnet/modeling_regnet.py:RegNetXLayer": ["ACT2FN", "Identity", "ModelConfig", "ModelConvLayer", "ModelShortCut", "ModelXLayer", "Module", "None", "Sequential", "__init__", "activation", "class", "config", "def", "else", "forward", "groups", "groups_width", "hidden_act", "hidden_state", "if", "in_channels", "int", "kernel_size", "layer", "max", "nn", "or", "out_channels", "residual", "return", "self", "shortcut", "should_apply_shortcut", "stride", "super"], "regnet/modeling_regnet.py:RegNetYLayer": ["ACT2FN", "Identity", "ModelConfig", "ModelConvLayer", "ModelSELayer", "ModelShortCut", "ModelYLayer", "Module", "None", "Sequential", "__init__", "activation", "class", "config", "def", "else", "forward", "groups", "groups_width", "hidden_act", "hidden_state", "if", "in_channels", "int", "kernel_size", "layer", "max", "nn", "or", "out_channels", "reduced_channels", "residual", "return", "round", "self", "shortcut", "should_apply_shortcut", "stride", "super"], "regnet/modeling_regnet.py:RegNetStage": ["ModelConfig", "ModelStage", "ModelXLayer", "ModelYLayer", "Module", "Sequential", "_", "__init__", "class", "config", "def", "depth", "else", "for", "forward", "hidden_state", "if", "in", "in_channels", "int", "layer", "layer_type", "layers", "nn", "out_channels", "range", "return", "self", "stride", "super", "x"], "regnet/modeling_regnet.py:RegNetEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "append", "bool", "class", "config", "def", "depth", "depths", "downsample_in_first_stage", "else", "embedding_size", "for", "forward", "hidden_sizes", "hidden_state", "hidden_states", "if", "in", "in_channels", "in_out_channels", "is", "last_hidden_state", "nn", "not", "out_channels", "output_hidden_states", "return", "return_dict", "self", "stage_module", "stages", "stride", "super", "tuple", "v", "zip"], "regnet/modeling_regnet.py:RegNetPreTrainedModel": ["BatchNorm2d", "Conv2d", "GroupNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelYLayer", "None", "PreTrainedModel", "_", "_calculate_fan_in_and_fan_out", "_init_weights", "_no_split_modules", "a", "base_model_prefix", "bias", "bound", "class", "config", "constant_", "def", "elif", "else", "fan_in", "fan_out", "if", "init", "is", "isinstance", "kaiming_normal_", "kaiming_uniform_", "main_input_name", "math", "mode", "module", "nn", "nonlinearity", "not", "pixel_values", "relu", "self", "sqrt", "uniform_", "weight"], "regnet/modeling_regnet.py:RegNetModel": ["AdaptiveAvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embedder", "embedding_output", "encoder", "encoder_outputs", "forward", "hidden_states", "if", "is", "last_hidden_state", "nn", "not", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "return", "return_dict", "self", "super", "use_return_dict"], "regnet/modeling_regnet.py:RegNetForImageClassification": ["Flatten", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Sequential", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "use_return_dict"], "luke/modeling_luke.py:BaseLukeModelOutputWithPooling": ["FloatTensor", "ModelModelModelOutputWithPooling", "ModelModelOutputWithPooling", "None", "Optional", "class", "entity_hidden_states", "entity_last_hidden_state", "r", "torch", "tuple"], "luke/modeling_luke.py:BaseLukeModelOutput": ["FloatTensor", "ModelModelModelOutput", "ModelModelOutput", "None", "Optional", "class", "entity_hidden_states", "entity_last_hidden_state", "r", "torch", "tuple"], "luke/modeling_luke.py:LukeMaskedLMOutput": ["FloatTensor", "ModelMaskedLMOutput", "ModelOutput", "None", "Optional", "attentions", "class", "entity_hidden_states", "entity_logits", "hidden_states", "logits", "loss", "mep_loss", "mlm_loss", "r", "torch", "tuple"], "luke/modeling_luke.py:EntityClassificationOutput": ["FloatTensor", "ModelClassificationOutput", "ModelOutput", "Model_hidden_states", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "luke/modeling_luke.py:EntityPairClassificationOutput": ["FloatTensor", "ModelOutput", "ModelPairClassificationOutput", "Model_hidden_states", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "luke/modeling_luke.py:EntitySpanClassificationOutput": ["FloatTensor", "ModelOutput", "ModelSpanClassificationOutput", "Model_hidden_states", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "luke/modeling_luke.py:LukeSequenceClassifierOutput": ["FloatTensor", "ModelOutput", "ModelSequenceClassifierOutput", "None", "Optional", "attentions", "class", "entity_hidden_states", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "luke/modeling_luke.py:LukeTokenClassifierOutput": ["FloatTensor", "ModelOutput", "ModelTokenClassifierOutput", "None", "Optional", "attentions", "class", "entity_hidden_states", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "luke/modeling_luke.py:LukeQuestionAnsweringModelOutput": ["FloatTensor", "ModelOutput", "ModelQuestionAnsweringModelOutput", "None", "Optional", "attentions", "class", "end_logits", "entity_hidden_states", "hidden_states", "loss", "r", "start_logits", "torch", "tuple"], "luke/modeling_luke.py:LukeMultipleChoiceModelOutput": ["FloatTensor", "ModelMultipleChoiceModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "entity_hidden_states", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "luke/modeling_luke.py:LukeEmbeddings": ["Dropout", "Embedding", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "arange", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "position_embeddings", "position_ids", "return", "self", "sequence_length", "size", "super", "to", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "luke/modeling_luke.py:LukeEntityEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelEntityEmbeddings", "Module", "None", "Optional", "__init__", "bias", "clamp", "class", "config", "def", "dim", "dropout", "embeddings", "entity_emb_size", "entity_embedding_dense", "entity_embeddings", "entity_ids", "entity_vocab_size", "eps", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "layer_norm_eps", "max_position_embeddings", "min", "nn", "padding_idx", "position_embedding_mask", "position_embeddings", "position_ids", "return", "self", "sum", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "zeros_like"], "luke/modeling_luke.py:LukeSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "ValueError", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "cat", "class", "concat_hidden_states", "config", "context_layer", "contiguous", "def", "dim", "dropout", "e2e_attention_scores", "e2e_key_layer", "e2e_query", "e2e_query_layer", "e2w_attention_scores", "e2w_key_layer", "e2w_query", "e2w_query_layer", "else", "embedding_size", "entity_attention_scores", "entity_hidden_states", "f", "forward", "functional", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "output_entity_hidden_states", "output_word_hidden_states", "outputs", "permute", "query", "query_layer", "raise", "return", "self", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "transpose_for_scores", "use_entity_aware_attention", "value", "value_layer", "view", "w2e_attention_scores", "w2e_key_layer", "w2e_query", "w2e_query_layer", "w2w_attention_scores", "w2w_key_layer", "w2w_query_layer", "word_attention_scores", "word_hidden_states", "word_size", "x"], "luke/modeling_luke.py:LukeSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "luke/modeling_luke.py:LukeAttention": ["False", "Model", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "NotImplementedError", "__init__", "attention", "attention_mask", "attention_output", "cat", "class", "concat_hidden_states", "concat_self_outputs", "config", "def", "dim", "does", "else", "entity_attention_output", "entity_hidden_states", "forward", "head_mask", "heads", "if", "is", "nn", "not", "of", "output", "output_attentions", "outputs", "prune_heads", "pruned_heads", "pruning", "raise", "return", "self", "self_outputs", "set", "size", "super", "support", "the", "torch", "word_attention_output", "word_hidden_states", "word_size"], "luke/modeling_luke.py:LukeIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "luke/modeling_luke.py:LukeOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "luke/modeling_luke.py:LukeLayer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "cat", "chunk_size_feed_forward", "class", "concat_attention_output", "config", "def", "dim", "else", "entity_hidden_states", "entity_layer_output", "feed_forward_chunk", "forward", "head_mask", "if", "intermediate", "intermediate_output", "is", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "size", "super", "torch", "word_hidden_states", "word_layer_output", "word_size"], "luke/modeling_luke.py:LukeEncoder": ["BaseModelModelOutput", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_entity_hidden_states", "all_self_attentions", "all_word_hidden_states", "attention_mask", "attentions", "class", "config", "def", "else", "entity_hidden_states", "entity_last_hidden_state", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v", "word_hidden_states"], "luke/modeling_luke.py:LukePooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "luke/modeling_luke.py:EntityPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Model_emb_size", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "luke/modeling_luke.py:EntityPredictionHead": ["False", "Linear", "ModelPredictionHead", "ModelPredictionHeadTransform", "Model_emb_size", "Model_vocab_size", "Module", "Parameter", "__init__", "bias", "class", "config", "decoder", "def", "forward", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "zeros"], "luke/modeling_luke.py:LukePreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEntityEmbeddings", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "embedding_dim", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "luke/modeling_luke.py:LukeModel": ["BaseModelModelOutputWithPooling", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelEntityEmbeddings", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "True", "Union", "ValueError", "Wrong", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cannot", "cat", "class", "config", "def", "device", "dim", "does", "dtype", "either", "elif", "else", "embeddings", "encoder", "encoder_outputs", "entity_attention_mask", "entity_embedding_output", "entity_embeddings", "entity_hidden_states", "entity_ids", "entity_last_hidden_state", "entity_position_ids", "entity_seq_length", "entity_token_type_ids", "extended_attention_mask", "f", "finfo", "for", "forward", "get_entity_embeddings", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "long", "min", "not", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "pruning", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_entity_embeddings", "set_input_embeddings", "shape", "size", "specify", "super", "support", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_attention_mask", "word_embedding_output", "word_embeddings", "zeros"], "luke/modeling_luke.py:create_position_ids_from_input_ids": ["Model_position_ids_from_input_ids", "cumsum", "def", "dim", "incremental_indices", "input_ids", "int", "long", "mask", "ne", "padding_idx", "return", "torch", "type_as"], "luke/modeling_luke.py:LukeLMHead": ["LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "luke/modeling_luke.py:LukeForMaskedLM": ["CrossEntropyLoss", "EntityPredictionHead", "FloatTensor", "LongTensor", "Model", "ModelForMaskedLM", "ModelLMHead", "ModelMaskedLMOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "_tie_or_clone_weights", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "decoder", "def", "device", "else", "entity_attention_mask", "entity_embeddings", "entity_hidden_states", "entity_ids", "entity_labels", "entity_last_hidden_state", "entity_logits", "entity_position_ids", "entity_predictions", "entity_token_type_ids", "entity_vocab_size", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "lm_head", "logits", "loss", "loss_fn", "mep_loss", "mlm_loss", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "tie_weights", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "v", "view", "vocab_size", "weight"], "luke/modeling_luke.py:LukeForEntityClassification": ["Dropout", "EntityClassificationOutput", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForEntityClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "binary_cross_entropy_with_logits", "bool", "class", "classifier", "config", "cross_entropy", "def", "device", "dropout", "else", "entity_attention_mask", "entity_hidden_states", "entity_ids", "entity_last_hidden_state", "entity_position_ids", "entity_token_type_ids", "feature_vector", "for", "forward", "functional", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "ndim", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "tuple", "type_as", "use_return_dict", "v", "view"], "luke/modeling_luke.py:LukeForEntityPairClassification": ["Dropout", "EntityPairClassificationOutput", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForEntityPairClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "binary_cross_entropy_with_logits", "bool", "cat", "class", "classifier", "config", "cross_entropy", "def", "device", "dim", "dropout", "else", "entity_attention_mask", "entity_hidden_states", "entity_ids", "entity_last_hidden_state", "entity_position_ids", "entity_token_type_ids", "feature_vector", "for", "forward", "functional", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "ndim", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "tuple", "type_as", "use_return_dict", "v", "view"], "luke/modeling_luke.py:LukeForEntitySpanClassification": ["Dropout", "EntitySpanClassificationOutput", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForEntitySpanClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "binary_cross_entropy_with_logits", "bool", "cat", "class", "classifier", "config", "cross_entropy", "def", "device", "dim", "dropout", "else", "end_states", "entity_attention_mask", "entity_end_positions", "entity_hidden_states", "entity_ids", "entity_last_hidden_state", "entity_position_ids", "entity_start_positions", "entity_token_type_ids", "expand", "feature_vector", "for", "forward", "functional", "gather", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "logits", "loss", "ndim", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "size", "start_states", "super", "to", "token_type_ids", "torch", "tuple", "type_as", "unsqueeze", "use_return_dict", "v", "view"], "luke/modeling_luke.py:LukeForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "ModelSequenceClassifierOutput", "None", "Optional", "True", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "dtype", "elif", "else", "entity_attention_mask", "entity_hidden_states", "entity_ids", "entity_position_ids", "entity_token_type_ids", "for", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "pooler_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "v", "view"], "luke/modeling_luke.py:LukeForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "ModelTokenClassifierOutput", "None", "Optional", "True", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "entity_attention_mask", "entity_hidden_states", "entity_ids", "entity_position_ids", "entity_token_type_ids", "for", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "last_hidden_state", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "v", "view"], "luke/modeling_luke.py:LukeForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "ModelQuestionAnsweringModelOutput", "None", "Optional", "True", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp_", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "entity_attention_mask", "entity_hidden_states", "entity_ids", "entity_position_ids", "entity_token_type_ids", "for", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "in", "input_ids", "inputs_embeds", "is", "last_hidden_state", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "v"], "luke/modeling_luke.py:LukeForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelMultipleChoiceModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "entity_attention_mask", "entity_hidden_states", "entity_ids", "entity_position_ids", "entity_token_type_ids", "for", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "pooler_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "v", "view"], "perception_lm/modeling_perception_lm.py:PerceptionLMAdaptiveAvgPooling": ["F", "ModelAdaptiveAvgPooling", "Module", "ValueError", "__init__", "a", "adaptive_avg_pool2d", "b", "be", "c", "class", "def", "expected", "f", "flatten", "forward", "h", "hidden_states", "if", "int", "is", "math", "nn", "num_tokens", "number", "permute", "pooling_ratio", "raise", "reshape", "return", "self", "shape", "sqrt", "square", "super", "to", "transpose"], "perception_lm/modeling_perception_lm.py:PerceptionLMMultiModalProjector": ["GELU", "Identity", "Linear", "ModelAdaptiveAvgPooling", "ModelConfig", "ModelMultiModalProjector", "Module", "True", "__init__", "bias", "class", "config", "def", "else", "embed_dim", "features", "forward", "gelu", "hidden_size", "if", "in_features", "input_size", "linear_1", "linear_2", "model_args", "nn", "out_features", "output_size", "permute", "pooling", "projector_pooling_ratio", "return", "self", "super", "text_config", "vision_config"], "perception_lm/modeling_perception_lm.py:PerceptionLMPreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "perception_lm/modeling_perception_lm.py:PerceptionLMModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch", "video_hidden_states"], "perception_lm/modeling_perception_lm.py:PerceptionLMCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple", "video_hidden_states"], "perception_lm/modeling_perception_lm.py:PerceptionLMModel": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cache_position", "can_return_tuple", "cannot", "class", "config", "decoder", "def", "device", "do", "dtype", "either", "else", "exactly", "expand_as", "f", "features", "flatten", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "input_ids", "inputs_embeds", "int", "is", "kwargs", "language_model", "last_hidden_state", "lm_kwargs", "logits_to_keep", "long", "masked_scatter", "match", "multi_modal_projector", "must", "n_image_tokens", "n_video_tokens", "not", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "raise", "return", "return_dict", "same", "self", "set_decoder", "set_input_embeddings", "size", "special_image_mask", "special_video_mask", "specify", "sum", "super", "tensor", "text_config", "the", "time", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "value", "video_features", "video_hidden_states", "video_token_id", "vision_config", "vision_tower", "vision_use_cls_token"], "perception_lm/modeling_perception_lm.py:PerceptionLMForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "lm_kwargs", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "value", "video_hidden_states", "vocab_size", "weight"], "segformer/modeling_segformer.py:SegFormerImageClassifierOutput": ["FloatTensor", "ImageClassifierOutput", "ModelImageClassifierOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "torch", "tuple"], "segformer/modeling_segformer.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "segformer/modeling_segformer.py:SegformerDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "segformer/modeling_segformer.py:SegformerOverlapPatchEmbeddings": ["Conv2d", "LayerNorm", "ModelOverlapPatchEmbeddings", "Module", "_", "__init__", "class", "def", "embeddings", "flatten", "forward", "height", "hidden_size", "kernel_size", "layer_norm", "nn", "num_channels", "padding", "patch_size", "pixel_values", "proj", "return", "self", "shape", "stride", "super", "transpose", "width"], "segformer/modeling_segformer.py:SegformerEfficientSelfAttention": ["Conv2d", "Dropout", "False", "LayerNorm", "Linear", "ModelEfficientSelfAttention", "Module", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "f", "forward", "functional", "heads", "height", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "kernel_size", "key", "key_layer", "layer_norm", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_channels", "number", "of", "output_attentions", "outputs", "permute", "query", "query_layer", "raise", "reshape", "return", "self", "seq_len", "seq_length", "sequence_reduction_ratio", "shape", "size", "softmax", "sqrt", "sr", "sr_ratio", "stride", "super", "the", "torch", "transpose", "value", "value_layer", "view", "width"], "segformer/modeling_segformer.py:SegformerSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super"], "segformer/modeling_segformer.py:SegformerAttention": ["False", "ModelAttention", "ModelEfficientSelfAttention", "ModelSelfOutput", "Module", "__init__", "all_head_size", "attention_head_size", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "height", "hidden_size", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "sequence_reduction_ratio", "set", "super", "union", "value", "width"], "segformer/modeling_segformer.py:SegformerDWConv": ["Conv2d", "ModelDWConv", "Module", "True", "__init__", "batch_size", "bias", "class", "def", "dim", "dwconv", "flatten", "forward", "groups", "height", "hidden_states", "nn", "num_channels", "return", "self", "seq_len", "shape", "super", "transpose", "view", "width"], "segformer/modeling_segformer.py:SegformerMixFFN": ["ACT2FN", "Dropout", "Linear", "ModelDWConv", "ModelMixFFN", "Module", "None", "__init__", "class", "config", "def", "dense1", "dense2", "dropout", "dwconv", "else", "forward", "height", "hidden_act", "hidden_dropout_prob", "hidden_features", "hidden_states", "if", "in_features", "intermediate_act_fn", "isinstance", "nn", "or", "out_features", "return", "self", "str", "super", "width"], "segformer/modeling_segformer.py:SegformerLayer": ["False", "Identity", "LayerNorm", "ModelAttention", "ModelDropPath", "ModelLayer", "ModelMixFFN", "Module", "__init__", "attention", "attention_output", "class", "config", "def", "drop_path", "else", "forward", "height", "hidden_features", "hidden_size", "hidden_states", "if", "in_features", "int", "layer_norm_1", "layer_norm_2", "layer_output", "mlp", "mlp_hidden_size", "mlp_output", "mlp_ratio", "nn", "num_attention_heads", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "sequence_reduction_ratio", "super", "width"], "segformer/modeling_segformer.py:SegformerEncoder": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelLayer", "ModelOverlapPatchEmbeddings", "Module", "ModuleList", "None", "Optional", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "and", "append", "attentions", "batch_size", "blk", "block", "block_layer", "blocks", "bool", "class", "config", "contiguous", "cpu", "cur", "def", "depths", "device", "drop_path", "drop_path_decays", "drop_path_rate", "else", "embedding_layer", "embeddings", "enumerate", "for", "forward", "height", "hidden_size", "hidden_sizes", "hidden_states", "i", "idx", "if", "in", "is", "item", "j", "last_hidden_state", "layer_norm", "layer_outputs", "layers", "len", "linspace", "mlp_ratio", "mlp_ratios", "nn", "norm_layer", "not", "num_attention_heads", "num_channels", "num_encoder_blocks", "or", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_size", "patch_sizes", "permute", "pixel_values", "range", "reshape", "reshape_last_stage", "return", "return_dict", "self", "sequence_reduction_ratio", "shape", "sr_ratios", "stride", "strides", "sum", "super", "torch", "tuple", "v", "width", "x", "zip"], "segformer/modeling_segformer.py:SegformerPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "self", "std", "weight", "zero_"], "segformer/modeling_segformer.py:SegformerModel": ["BaseModelOutput", "FloatTensor", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_prune_heads", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "encoder_outputs", "for", "forward", "heads", "heads_to_prune", "hidden_states", "if", "in", "is", "items", "last_hidden_state", "layer", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "prune_heads", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "segformer/modeling_segformer.py:SegformerForImageClassification": ["FloatTensor", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelImageClassifierOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "config", "def", "dim", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "permute", "pixel_values", "post_init", "r", "reshape", "reshape_last_stage", "return", "return_dict", "self", "sequence_output", "shape", "super", "torch", "tuple", "use_return_dict"], "segformer/modeling_segformer.py:SegformerMLP": ["Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "__init__", "class", "config", "decoder_hidden_size", "def", "flatten", "forward", "hidden_states", "input_dim", "nn", "proj", "return", "self", "super", "torch", "transpose"], "segformer/modeling_segformer.py:SegformerDecodeHead": ["BatchNorm2d", "Conv2d", "Dropout", "False", "FloatTensor", "ModelDecodeHead", "ModelMLP", "ModelPreTrainedModel", "ModuleList", "ReLU", "Tensor", "__init__", "activation", "align_corners", "all_hidden_states", "and", "append", "batch_norm", "batch_size", "bias", "bilinear", "cat", "class", "classifier", "classifier_dropout_prob", "config", "contiguous", "decoder_hidden_size", "def", "dim", "dropout", "encoder_hidden_state", "encoder_hidden_states", "for", "forward", "functional", "height", "hidden_sizes", "hidden_states", "i", "if", "in", "in_channels", "input_dim", "int", "interpolate", "is", "kernel_size", "linear_c", "linear_fuse", "logits", "math", "mlp", "mlps", "mode", "ndim", "nn", "num_encoder_blocks", "num_labels", "out_channels", "permute", "range", "reshape", "reshape_last_stage", "return", "self", "shape", "size", "sqrt", "super", "torch", "width", "zip"], "segformer/modeling_segformer.py:SegformerForSemanticSegmentation": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "Model", "ModelDecodeHead", "ModelForSemanticSegmentation", "ModelModel", "ModelPreTrainedModel", "None", "Number", "Optional", "SemanticSegmenterOutput", "True", "Union", "ValueError", "__init__", "align_corners", "and", "attentions", "auto_docstring", "be", "bilinear", "bool", "class", "config", "decode_head", "def", "elif", "else", "encoder_hidden_states", "f", "float", "forward", "functional", "hidden_states", "if", "ignore_index", "interpolate", "is", "labels", "logits", "loss", "loss_fct", "mean", "mode", "nn", "none", "not", "num_labels", "of", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "reduction", "return", "return_dict", "self", "semantic_loss_ignore_index", "shape", "should", "size", "squeeze", "super", "torch", "tuple", "upsampled_logits", "use_return_dict", "valid_mask"], "wavlm/modeling_wavlm.py:WavLMSamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "wavlm/modeling_wavlm.py:WavLMPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "wavlm/modeling_wavlm.py:WavLMFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "wavlm/modeling_wavlm.py:WavLMAttention": ["BoolTensor", "Embedding", "F", "False", "FloatTensor", "Linear", "LongTensor", "ModelAttention", "Module", "None", "Optional", "Parameter", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_relative_positions_bucket", "abs", "add_zero_attn", "and", "arange", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bias_k", "bias_v", "bool", "broadcast_to", "bsz", "by", "cat", "chunk", "class", "compute_bias", "context_position", "def", "device", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "empty", "f", "float", "forward", "full_like", "gate_a", "gate_b", "gate_output", "gated_hidden_states", "gated_position_bias", "got", "gru_rel_pos_const", "gru_rel_pos_linear", "has_relative_position_bias", "head_dim", "hidden_states", "if", "index", "int", "is", "is_small", "k_proj", "k_proj_weight", "key", "key_length", "key_padding_mask", "log", "long", "math", "max_distance", "max_exact", "memory_position", "min", "multi_head_attention_forward", "must", "ne", "nn", "not", "num_buckets", "num_heads", "ones", "out_proj", "output_attentions", "permute", "position_bias", "q_proj", "q_proj_weight", "query", "query_length", "raise", "rel_attn_embed", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "relative_position_proj", "relative_positions", "relative_positions_if_large", "repeat", "return", "scaling", "self", "shape", "sigmoid", "size", "sum", "super", "tgt_len", "to", "torch", "torch_multi_head_self_attention", "training", "transpose", "tuple", "unsqueeze", "use_separate_proj_weight", "v_proj", "v_proj_weight", "value", "values", "view", "weight", "where"], "wavlm/modeling_wavlm.py:WavLMFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "wavlm/modeling_wavlm.py:WavLMEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelFeedForward", "None", "True", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "bool", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "has_relative_position_bias", "hidden_dropout", "hidden_size", "hidden_states", "if", "index", "layer_norm", "layer_norm_eps", "max_bucket_distance", "max_distance", "nn", "num_attention_heads", "num_buckets", "num_heads", "output_attentions", "outputs", "position_bias", "return", "self", "super"], "wavlm/modeling_wavlm.py:WavLMEncoderLayerStableLayerNorm": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayerStableLayerNorm", "ModelFeedForward", "None", "True", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "bool", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "has_relative_position_bias", "hidden_dropout", "hidden_size", "hidden_states", "if", "layer_norm", "layer_norm_eps", "max_bucket_distance", "max_distance", "nn", "num_attention_heads", "num_buckets", "num_heads", "output_attentions", "outputs", "position_bias", "return", "self", "super"], "wavlm/modeling_wavlm.py:WavLMEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalConvEmbedding", "Module", "ModuleList", "None", "True", "__init__", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "else", "enumerate", "eps", "expand_attention_mask", "for", "forward", "gradient_checkpointing", "has_relative_position_bias", "hidden_dropout", "hidden_size", "hidden_states", "i", "if", "in", "index", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_bias", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "self", "shape", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "unsqueeze", "v"], "wavlm/modeling_wavlm.py:WavLMEncoderStableLayerNorm": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoderLayerStableLayerNorm", "ModelEncoderStableLayerNorm", "ModelPositionalConvEmbedding", "Module", "ModuleList", "None", "True", "__init__", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "else", "enumerate", "eps", "expand_attention_mask", "for", "forward", "gradient_checkpointing", "has_relative_position_bias", "hidden_dropout", "hidden_size", "hidden_states", "i", "if", "in", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_bias", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "self", "shape", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "unsqueeze", "v"], "wavlm/modeling_wavlm.py:WavLMGumbelVectorQuantizer": ["FloatTensor", "Linear", "ModelGumbelVectorQuantizer", "Module", "Parameter", "True", "ValueError", "__init__", "_compute_perplexity", "argmax", "batch_size", "be", "by", "class", "codevector_dim", "codevector_idx", "codevector_probs", "codevector_soft_dist", "codevectors", "codevectors_per_group", "concatenation", "config", "conv_dim", "def", "dim", "divisible", "else", "exp", "f", "float", "for", "forward", "functional", "gumbel_softmax", "hard", "hidden_size", "hidden_states", "if", "log", "marginal_probs", "mean", "must", "new_zeros", "nn", "num_codevector_groups", "num_codevectors_per_group", "num_groups", "num_vars", "perplexity", "probs", "raise", "return", "scatter_", "self", "sequence_length", "shape", "softmax", "staticmethod", "sum", "super", "tau", "temperature", "torch", "training", "type_as", "unsqueeze", "view", "weight_proj"], "wavlm/modeling_wavlm.py:WavLMPreTrainedModel": ["Conv1d", "False", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelFeatureProjection", "ModelGumbelVectorQuantizer", "ModelPositionalConvEmbedding", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "True", "Union", "_", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "adapter_stride", "add_adapter", "arange", "attention_mask", "b", "base_model_prefix", "batch_size", "bias", "bool", "class", "codevectors", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "data", "def", "device", "dim", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "groups", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "module", "nn", "non_padded_lengths", "normal_", "not", "num_adapter_layers", "output_lengths", "projection", "range", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "supports_gradient_checkpointing", "to", "torch", "uniform_", "weight", "weight_proj", "zero_", "zeros", "zip"], "wavlm/modeling_wavlm.py:WavLMNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "wavlm/modeling_wavlm.py:WavLMLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "wavlm/modeling_wavlm.py:WavLMGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "wavlm/modeling_wavlm.py:WavLMFeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "wavlm/modeling_wavlm.py:WavLMAdapterLayer": ["Conv1d", "ModelAdapterLayer", "Module", "__init__", "adapter_kernel_size", "adapter_stride", "class", "config", "conv", "def", "dim", "forward", "functional", "glu", "hidden_states", "nn", "output_hidden_size", "padding", "return", "self", "stride", "super"], "wavlm/modeling_wavlm.py:WavLMAdapter": ["LayerNorm", "Linear", "ModelAdapter", "ModelAdapterLayer", "Module", "ModuleList", "None", "_", "__init__", "and", "class", "config", "def", "else", "for", "forward", "hidden_size", "hidden_states", "if", "in", "is", "layer", "layerdrop", "layerdrop_prob", "layers", "nn", "not", "np", "num_adapter_layers", "or", "output_hidden_size", "proj", "proj_layer_norm", "random", "range", "return", "self", "super", "training", "transpose"], "wavlm/modeling_wavlm.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "wavlm/modeling_wavlm.py:WavLMModel": ["False", "FloatTensor", "FutureWarning", "LongTensor", "ModelAdapter", "ModelBaseModelOutput", "ModelConfig", "ModelEncoder", "ModelEncoderStableLayerNorm", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Tensor", "The", "Transformers", "True", "Union", "__init__", "_compute_mask_indices", "_freeze_parameters", "_get_feature_vector_attention_mask", "_mask_hidden_states", "adapter", "add_adapter", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "class", "config", "def", "deprecated", "device", "do_stable_layer_norm", "dtype", "elif", "else", "encoder", "encoder_outputs", "equivalent", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "freeze_feature_encoder", "freeze_feature_extractor", "getattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "method", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "removed", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "the", "to", "torch", "training", "transpose", "tuple", "uniform_", "use", "use_return_dict", "v5", "warn", "warnings", "will"], "wavlm/modeling_wavlm.py:WavLMForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "wavlm/modeling_wavlm.py:WavLMForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "wavlm/modeling_wavlm.py:WavLMForAudioFrameClassification": ["Audio", "CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForAudioFrameClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Tensor", "The", "TokenClassifierOutput", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "adapters", "add_adapter", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "axis", "be", "bool", "class", "classification", "classifier", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "feature_extractor", "for", "forward", "frame", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "init_weights", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "param", "parameters", "r", "raise", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "wavlm/modeling_wavlm.py:AMSoftmaxLoss": ["CrossEntropyLoss", "Model", "Module", "Parameter", "True", "__init__", "bool", "class", "cos_theta", "def", "dim", "flatten", "forward", "functional", "hidden_states", "input_dim", "labels", "logits", "loss", "margin", "mm", "nn", "normalize", "num_labels", "one_hot", "onehot", "psi", "randn", "requires_grad", "return", "scale", "self", "super", "torch", "weight", "where"], "wavlm/modeling_wavlm.py:TDNNLayer": ["Detected", "Linear", "LoRA", "LoraLayer", "Model", "Module", "ReLU", "Tensor", "You", "__init__", "activation", "applied", "be", "bias", "class", "config", "conv1d", "def", "dilation", "due", "else", "exclude", "forward", "from", "functional", "hidden_states", "if", "in_conv_dim", "is_peft_available", "isinstance", "kernel", "kernel_size", "layer_id", "modules", "nn", "on", "optimization", "out_conv_dim", "return", "s", "self", "should", "super", "t", "target", "tdnn_dilation", "tdnn_dim", "tdnn_kernel", "to", "torch", "transpose", "view", "warn", "warnings", "weight", "weights", "won"], "wavlm/modeling_wavlm.py:WavLMForXVector": ["AMSoftmaxLoss", "False", "FutureWarning", "Linear", "LongTensor", "Model", "ModelForXVector", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "Please", "TDNNLayer", "Tensor", "The", "Transformers", "True", "Union", "XVectorOutput", "_HIDDEN_STATES_START_POSITION", "__init__", "_conv_out_length", "_freeze_parameters", "_get_feat_extract_output_lengths", "_get_tdnn_output_lengths", "and", "append", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cat", "class", "classifier", "config", "def", "deprecated", "dim", "else", "embeddings", "enumerate", "equivalent", "feat_extract_output_lengths", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hidden_size", "hidden_states", "i", "if", "in", "init_weights", "input_length", "input_lengths", "input_values", "instead", "int", "is", "kernel_size", "labels", "layer_weights", "len", "length", "logits", "loss", "mean", "mean_features", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "objective", "ones", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "parameters", "projector", "r", "range", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "statistic_pooling", "std", "std_features", "stride", "sum", "super", "tdnn", "tdnn_dim", "tdnn_kernel", "tdnn_layer", "tdnn_layers", "tdnn_output_lengths", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will", "xvector_output_dim"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoePreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:_get_feat_extract_output_lengths": ["_get_feat_extract_output_lengths", "def", "feat_lengths", "input_lengths", "input_lengths_leave", "output_lengths", "return"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoePreTrainedModelForConditionalGeneration": ["False", "LongTensor", "ModelPreTrainedModel", "ModelPreTrainedModelForConditionalGeneration", "None", "Optional", "Tensor", "True", "_", "_get_feat_extract_output_lengths", "_iter", "_llm_pos_ids", "_prepare_4d_causal_attention_mask_with_cache_position", "and", "append", "arange", "argwhere", "attention_mask", "audio_data_index", "audio_idx", "audio_len", "audio_llm_pos_ids", "audio_nums", "audio_seqlens", "audio_start_token_id", "audio_token_id", "batch_size", "bool", "bos_len", "cache_position", "cat", "causal_mask", "class", "clone", "config", "cpu", "cumsum", "current_chunk", "def", "device", "diagonal", "dim", "dtype", "ed_audio_start", "ed_vision_start", "elif", "else", "enumerate", "eos_len", "expand", "fill_value", "flatten", "float", "for", "full", "get_chunked_index", "get_llm_pos_ids_for_vision", "get_rope_index", "grid_hs", "grid_t", "grid_ws", "h_index", "i", "if", "image_grid_thw", "image_idx", "image_len", "image_nums", "image_token_id", "in", "index", "input_ids", "input_tokens", "int", "is", "item", "keepdim", "len", "list", "llm_grid_h", "llm_grid_w", "llm_pos_ids", "llm_pos_ids_list", "llm_positions", "mask_length", "masked_fill", "masked_fill_", "max", "max_position_ids", "min", "min_dtype", "min_ed", "mrope_position_deltas", "multimodal_nums", "not", "or", "padding_mask", "position_id_per_seconds", "position_ids", "prod", "range", "remain_audios", "remain_images", "remain_videos", "remove_index", "reshape", "return", "second_per_grids", "self", "sequence_length", "shape", "spatial_merge_size", "squeeze", "st", "st_idx", "stack", "start_idx", "sum", "t_index", "target_length", "tensor", "text_len", "to", "token_indices", "tokens_per_chunk", "tolist", "torch", "total_input_ids", "triu", "tuple", "unsqueeze", "use_audio_in_video", "video_data_index", "video_grid_thw", "video_idx", "video_len", "video_llm_pos_ids", "video_nums", "video_token_id", "view", "vision_idx", "vision_start_indices", "vision_start_token_id", "vision_tokens", "w_index", "while", "yield", "zeros"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeAudioAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAudioAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "be", "bias", "by", "class", "config", "contiguous", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "d_model", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "encoder_attention_heads", "f", "forward", "got", "head_dim", "hidden_states", "if", "is_causal", "is_decoder", "k_proj", "key_states", "kwargs", "max", "max_length_k", "max_length_q", "max_seqlen", "must", "nn", "not", "num_heads", "num_key_value_groups", "out_proj", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_length", "size", "super", "torch", "training", "transpose", "tuple", "unsqueeze", "v_proj", "value_states"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeAudioEncoderLayer": ["ACT2FN", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAudioAttention", "ModelAudioEncoderConfig", "ModelAudioEncoderLayer", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_mask", "clamp", "clamp_value", "class", "config", "cu_seqlens", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "hidden_states", "if", "kwargs", "max", "min", "nn", "outputs", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:SinusoidsPositionEmbedding": ["False", "ModelPositionEmbedding", "Module", "ValueError", "__init__", "arange", "cat", "channels", "class", "cos", "def", "dim", "even", "exp", "float", "forward", "if", "input", "int", "inv_timescales", "length", "log", "log_timescale_increment", "max_timescale", "needs", "newaxis", "nn", "np", "persistent", "positional_embedding", "raise", "register_buffer", "return", "scaled_time", "self", "seqlen", "sin", "super", "torch"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeAudioEncoder": ["ACT2FN", "BaseModelOutput", "Conv2d", "F", "False", "LayerNorm", "Linear", "LongTensor", "ModelAudioEncoder", "ModelAudioEncoderConfig", "ModelAudioEncoderLayer", "ModelPreTrainedModel", "Module", "ModuleList", "None", "SinusoidsPositionEmbedding", "T", "Tensor", "True", "_", "__init__", "_attn_implementation", "_freeze_parameters", "_get_feat_extract_output_lengths", "_no_split_modules", "_prepare_attention_mask", "_requires_grad", "_supports_sdpa", "act", "activation_function", "aftercnn_lens", "append", "attention_mask", "auto_docstring", "b", "batch_first", "batch_mask", "batch_mask_after_cnn", "bias", "bool", "c", "cat", "ceil", "chunk", "chunk_lengths", "chunk_list", "chunk_num", "class", "cnn_len", "config", "contiguous", "conv1", "conv2d1", "conv2d2", "conv2d3", "conv_chunksize", "conv_out", "cu_chunk_lens", "cu_seqlens", "cumsum", "d_model", "def", "device", "dim", "downsample_hidden_size", "dropout", "dtype", "else", "embed_dim", "embed_scale", "encoder_layer", "encoder_layers", "enumerate", "f", "feature_lens", "feature_lens_after_cnn", "fill_value", "finfo", "flash_attention_2", "for", "forward", "full", "gelu", "get_input_embeddings", "gradient_checkpointing", "hidden_states", "i", "if", "in", "input_features", "input_lengths", "inputs_tensor", "int32", "last_hidden_state", "layer_outputs", "layers", "len", "length", "ln_post", "long", "main_input_name", "math", "max", "max_len", "max_len_after_cnn", "max_source_positions", "min", "n_window", "n_window_infer", "nn", "num_mel_bins", "ones", "output_dim", "output_lengths", "pad", "pad_sequence", "padded_and_mask_function", "padded_embed", "padded_embeds", "padded_feature", "padded_mask_after_cnn", "padded_tensor", "padding", "padding_side", "padding_value", "param", "parameters", "permute", "positional_embedding", "post_init", "proj1", "proj2", "r", "range", "remainder", "requires_grad", "return", "right", "rnn", "scale_embedding", "self", "seq_length", "set_input_embeddings", "shape", "size", "split", "sqrt", "sum", "super", "t", "tail_chunk_index", "tensor", "tensor_len", "tensor_list", "to", "tolist", "torch", "transpose", "unsqueeze", "utils", "value", "view", "window_aftercnn", "zeros"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionEncoderConfig", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionPatchMerger": ["False", "GELU", "LayerNorm", "Linear", "ModelVisionEncoderConfig", "ModelVisionPatchMerger", "Module", "ModuleList", "None", "Tensor", "__init__", "class", "config", "def", "else", "eps", "for", "forward", "hidden", "hidden_size", "if", "in", "layer", "ln_q", "mlp", "nn", "out_hidden_size", "return", "self", "spatial_merge_size", "super", "torch", "use_postshuffle_norm", "view"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "True", "__init__", "act_fn", "bias", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "linear_fc1", "linear_fc2", "nn", "return", "self", "super"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionPatchEmbed": ["Conv3d", "ModelVisionPatchEmbed", "Module", "None", "Tensor", "True", "__init__", "bias", "class", "config", "def", "dtype", "embed_dim", "forward", "hidden_size", "hidden_states", "in_channels", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionRotaryEmbedding": ["False", "ModelVisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionBlock": ["GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionBlock", "ModelVisionMLP", "None", "Optional", "Tensor", "__init__", "attn", "attn_implementation", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "nn", "norm1", "norm2", "position_embeddings", "return", "rotary_pos_emb", "sdpa", "self", "str", "super", "torch", "tuple"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeVisionEncoder": ["Embedding", "F", "False", "ModelPreTrainedModel", "ModelVisionBlock", "ModelVisionEncoder", "ModelVisionEncoderConfig", "ModelVisionPatchEmbed", "ModelVisionPatchMerger", "ModelVisionRotaryEmbedding", "ModuleList", "None", "T", "Tensor", "True", "_", "__init__", "_no_split_modules", "append", "arange", "base_h", "base_h_ceil", "blk", "block_cols", "block_rows", "blocks", "cat", "class", "clip", "col_idx", "config", "coords", "cos", "cu_seqlens", "cumsum", "deepstack_feature", "deepstack_feature_lists", "deepstack_merger_list", "deepstack_visual_indexes", "def", "depth", "device", "dh", "dim", "dtype", "dw", "else", "emb", "embeddings", "empty", "enumerate", "expand", "extend", "fast_pos_embed_interpolate", "flatten", "for", "forward", "freq_table", "gradient_checkpointing", "grid_hs", "grid_thw", "grid_ts", "grid_ws", "h", "h_idxs", "h_idxs_ceil", "h_idxs_floor", "head_dim", "height", "hidden_size", "hidden_states", "i", "idx_list", "idx_tensor", "if", "in", "index", "indices", "inputs", "int", "int32", "intra_col", "intra_row", "is_tracing", "item", "jit", "kwargs", "layer_num", "len", "linspace", "long", "max", "max_hw", "merge_size", "merged_h", "merged_w", "merger", "merger_list", "nn", "num_frames", "num_grid_per_side", "num_heads", "num_position_embeddings", "num_tokens", "offset", "pad", "patch_embed", "patch_pos_embeds", "patch_pos_embeds_permute", "patch_size", "permute", "pos_embed", "pos_embeds", "pos_ids", "position_embeddings", "prod", "property", "range", "repeat", "repeat_interleave", "reshape", "return", "rot_pos_emb", "rotary_pos_emb", "row_idx", "self", "seq_len", "shape", "sin", "size", "spatial_merge_size", "spatial_merge_unit", "split", "stack", "sum", "super", "t", "tensor", "tolist", "torch", "total_tokens", "use_postshuffle_norm", "value", "view", "w", "w_idxs", "w_idxs_ceil", "w_idxs_floor", "weight", "weight_list", "weight_tensor", "weights", "width", "zip"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelThinkerTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "apply_interleaved_mrope", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "enumerate", "expand", "float", "for", "forward", "freqs", "freqs_t", "get", "hasattr", "idx", "if", "in", "inv_freq", "inv_freq_expanded", "is", "isinstance", "length", "max_position_embeddings", "max_seq_len_cached", "mps", "mrope_section", "ndim", "nn", "no_grad", "not", "offset", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "slice", "start", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextMLP": ["ACT2FN", "False", "Linear", "ModelThinkerTextMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "not", "return", "self", "super", "up_proj", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextSparseMoeBlock": ["F", "False", "Linear", "ModelThinkerTextMLP", "ModelThinkerTextSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "keepdim", "moe_intermediate_size", "nn", "nonzero", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextRMSNorm": ["ModelThinkerTextRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelThinkerTextAttention", "ModelThinkerTextRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextDecoderLayer": ["Cache", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelThinkerTextAttention", "ModelThinkerTextDecoderLayer", "ModelThinkerTextMLP", "ModelThinkerTextRMSNorm", "ModelThinkerTextSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "and", "attention_mask", "cache_position", "class", "config", "decoder_sparse_step", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "intermediate_size", "isinstance", "kwargs", "layer_idx", "mlp", "mlp_only_layers", "new_name", "not", "num_experts", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextPreTrainedModel": ["False", "ModelTextConfig", "ModelThinkerTextAttention", "ModelThinkerTextDecoderLayer", "ModelThinkerTextPreTrainedModel", "ModelThinkerTextSparseMoeBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "config_class", "hidden_states", "index", "model", "past_key_values", "router_logits", "supports_gradient_checkpointing"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTextRMSNorm": ["ModelTextRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextRMSNorm", "ModelThinkerTextAttention", "ModelThinkerTextDecoderLayer", "ModelThinkerTextModel", "ModelThinkerTextRotaryEmbedding", "ModelThinkerTextSparseMoeBlock", "ModuleList", "None", "Optional", "OutputRecorder", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "_deepstack_process", "_no_split_modules", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "check_model_inputs", "class", "clone", "config", "config_class", "create_causal_mask", "decoder_layer", "deepstack_visual_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "enumerate", "eps", "exactly", "expand", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "index", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "len", "list", "local_this", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "router_logits", "self", "shape", "specify", "super", "text_position_ids", "to", "torch", "tuple", "use_cache", "view", "visual_embeds", "visual_pos_masks", "vocab_size"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerCausalLMOutputWithPast": ["LongTensor", "ModelThinkerCausalLMOutputWithPast", "MoeCausalLMOutputWithPast", "None", "Optional", "class", "r", "rope_deltas", "torch"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeThinkerForConditionalGeneration": ["False", "FloatTensor", "GenerationMixin", "Image", "Linear", "LongTensor", "ModelAudioEncoder", "ModelAudioEncoderLayer", "ModelPreTrainedModelForConditionalGeneration", "ModelThinkerCausalLMOutputWithPast", "ModelThinkerConfig", "ModelThinkerForConditionalGeneration", "ModelThinkerTextAttention", "ModelThinkerTextDecoderLayer", "ModelThinkerTextModel", "ModelThinkerTextSparseMoeBlock", "ModelVisionEncoder", "None", "Optional", "OutputRecorder", "True", "Union", "ValueError", "Videos", "_", "__init__", "_can_record_outputs", "_from_config", "_no_split_modules", "_tied_weights_keys", "add", "all", "and", "arange", "attention_mask", "attentions", "audio_config", "audio_feature_lengths", "audio_features", "audio_mask", "audio_outputs", "audio_token_id", "audio_tower", "auto_docstring", "aux_loss", "base_model_prefix", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "deepstack_visual_embeds", "def", "delta", "delta0", "device", "dim", "do", "dtype", "else", "embed_joint", "embed_tokens", "expand", "expand_as", "f", "feature_attention_mask", "feature_lens", "features", "for", "forward", "get_audio_features", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_text_config", "get_video_features", "grid_thw", "hidden_size", "hidden_states", "if", "image", "image_embeds", "image_embeds_multiscale", "image_features", "image_grid_thw", "image_mask", "image_mask_joint", "image_token_id", "img_embed", "in", "index", "input_features", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "long", "loss", "loss_function", "masked_scatter", "match", "model", "model_inputs", "n_image_tokens", "n_video_tokens", "new_zeros", "nn", "not", "num_experts", "num_experts_per_tok", "numel", "or", "output_router_logits", "outputs", "pad_token_id", "past_key_values", "permute", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "raise", "return", "rope_deltas", "router_aux_loss_coef", "router_logits", "self", "seq_length", "set_input_embeddings", "shape", "spatial_merge_size", "special_audio_mask", "special_image_mask", "special_video_mask", "sum", "super", "tensor", "text_config", "thinker", "to", "tokens", "torch", "tuple", "type", "unsqueeze", "use_audio_in_video", "use_cache", "value", "vid_embed", "video_embeds", "video_embeds_multiscale", "video_features", "video_grid_thw", "video_mask", "video_mask_joint", "video_second_per_grid", "video_token_id", "view", "vision_config", "visual", "visual_embeds_multiscale", "visual_embeds_multiscale_joint", "visual_pos_masks", "vocab_size", "weight", "zip"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerResizeMLP": ["ACT2FN", "Linear", "ModelTalkerConfig", "ModelTalkerResizeMLP", "Module", "True", "__init__", "act_fn", "bias", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "linear_fc1", "linear_fc2", "nn", "return", "self", "super", "text_config", "thinker_hidden_size"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerCodePredictorOutputWithPast": ["CausalLMOutputWithPast", "ModelTalkerCodePredictorOutputWithPast", "None", "Optional", "class", "generation_steps", "int", "r"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeRMSNorm": ["ModelRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerCodePredictorAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelConfig", "ModelRMSNorm", "ModelTalkerCodePredictorAttention", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerCodePredictorDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelMLP", "ModelRMSNorm", "ModelTalkerCodePredictorAttention", "ModelTalkerCodePredictorDecoderLayer", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerCodePredictorModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTalkerCodePredictorAttention", "ModelTalkerCodePredictorConfig", "ModelTalkerCodePredictorDecoderLayer", "ModelTalkerCodePredictorModel", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "_", "__init__", "_can_record_outputs", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "base_model_prefix", "be", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "code_predictor", "codec_embedding", "config", "config_class", "create_causal_mask", "decoder_layer", "def", "device", "dict", "else", "eps", "expected", "for", "forward", "full_attention", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_types", "layers", "mask_kwargs", "model", "nn", "norm", "not", "num_code_groups", "num_hidden_layers", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "super", "talker", "to", "torch", "unsqueeze", "use_cache", "vocab_size"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration": ["BaseModelOutputWithPast", "CausalLMOutputWithPast", "False", "GenerationMixin", "Linear", "ModelPreTrainedModel", "ModelTalkerCodePredictorAttention", "ModelTalkerCodePredictorConfig", "ModelTalkerCodePredictorDecoderLayer", "ModelTalkerCodePredictorModel", "ModelTalkerCodePredictorModelForConditionalGeneration", "ModelTalkerCodePredictorOutputWithPast", "ModuleList", "None", "_", "__init__", "_can_record_outputs", "_from_config", "_pp_plan", "_tied_weights_keys", "_tp_plan", "_update_model_kwargs_for_generation", "and", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "cache_position", "can_return_tuple", "class", "code_predictor", "colwise_rep", "config", "config_class", "def", "else", "for", "forward", "generation_steps", "get_input_embeddings", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "is_encoder_decoder", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "loss", "loss_function", "model", "model_kwargs", "nn", "not", "num_code_groups", "num_new_tokens", "outputs", "past_key_values", "position_ids", "post_init", "r", "range", "return", "self", "shape", "super", "talker", "use_cache", "vocab_size", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerOutputWithPast": ["ModelTalkerOutputWithPast", "MoeCausalLMOutputWithPast", "None", "Optional", "class", "generation_step", "int", "r"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerRotaryEmbedding": ["ModelTalkerRotaryEmbedding", "ModelThinkerTextRotaryEmbedding", "class", "pass"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerTextMLP": ["ACT2FN", "False", "Linear", "ModelTalkerTextMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "not", "return", "self", "super", "up_proj", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerTextSparseMoeBlock": ["F", "False", "Linear", "ModelTalkerTextMLP", "ModelTalkerTextSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "keepdim", "moe_intermediate_size", "nn", "nonzero", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "shared_expert", "shared_expert_gate", "shared_expert_intermediate_size", "shared_expert_output", "sigmoid", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerDecoderLayer": ["Cache", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelTalkerDecoderLayer", "ModelTalkerTextSparseMoeBlock", "ModelThinkerTextAttention", "ModelThinkerTextMLP", "ModelThinkerTextRMSNorm", "ModelThinkerTextSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "and", "attention_mask", "cache_position", "class", "config", "decoder_sparse_step", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "intermediate_size", "isinstance", "kwargs", "layer_idx", "mlp", "mlp_only_layers", "new_name", "not", "num_experts", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelTalkerDecoderLayer", "ModelTalkerModel", "ModelTalkerRotaryEmbedding", "ModelTalkerTextConfig", "ModelTalkerTextSparseMoeBlock", "ModelTextConfig", "ModelTextRMSNorm", "ModelThinkerTextAttention", "ModuleList", "None", "Optional", "OutputRecorder", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "_deepstack_process", "_no_split_modules", "and", "arange", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bool", "cache_position", "check_model_inputs", "class", "clone", "codec_embedding", "config", "config_class", "create_causal_mask", "decoder_layer", "deepstack_visual_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "enumerate", "eps", "exactly", "expand", "for", "forward", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "index", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "len", "list", "local_this", "model", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "router_logits", "self", "shape", "specify", "super", "talker", "text_position_ids", "to", "torch", "tuple", "use_cache", "view", "visual_embeds", "visual_pos_masks", "vocab_size"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeTalkerForConditionalGeneration": ["False", "GenerationMixin", "Linear", "LongTensor", "ModelPreTrainedModelForConditionalGeneration", "ModelTalkerCodePredictorModelForConditionalGeneration", "ModelTalkerConfig", "ModelTalkerForConditionalGeneration", "ModelTalkerModel", "ModelTalkerOutputWithPast", "ModelTalkerResizeMLP", "ModelTalkerTextSparseMoeBlock", "ModelThinkerTextAttention", "ModelThinkerTextPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "OutputRecorder", "Tensor", "True", "__init__", "_can_record_outputs", "_from_config", "_no_split_modules", "_pp_plan", "_tied_weights_keys", "_tp_plan", "_update_model_kwargs_for_generation", "add", "and", "arange", "attention_mask", "attentions", "audio_feature_lengths", "audio_seqlens", "auto_docstring", "aux_loss", "base_model_prefix", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "code_predictor", "code_predictor_config", "codec_head", "codec_hiddens", "colwise_rep", "config", "config_class", "def", "delta", "delta0", "device", "dim", "do_sample", "else", "expand", "for", "forward", "generate", "generation_step", "get", "get_input_embeddings", "get_llm_pos_ids_for_vision", "get_rope_index", "grid_hs", "grid_ws", "hid", "hidden_projection", "hidden_size", "hidden_states", "if", "image_grid_thw", "in", "index", "input_ids", "inputs", "inputs_embeds", "int", "is", "is_encoder_decoder", "keepdim", "kwargs", "labels", "last_hidden_state", "last_id_hidden", "last_residual_hidden", "list", "lm_head", "load_balancing_loss_func", "logits", "loss", "loss_function", "max_new_tokens", "mid_residual_hiddens", "model", "model_kwargs", "nn", "not", "num_code_groups", "num_experts", "num_experts_per_tok", "num_new_tokens", "or", "output_hidden_states", "output_router_logits", "outputs", "past_hidden", "past_key_values", "pop", "position_ids", "post_init", "predictor_result", "prepare_inputs_for_generation", "r", "residual_codes", "return", "return_dict_in_generate", "rope_deltas", "router_aux_loss_coef", "router_logits", "second_per_grids", "self", "seq_length", "sequences", "shape", "spatial_merge_size", "start_idx", "sum", "super", "t_index", "talker", "talker_input_ids", "text_config", "text_projection", "to", "top_k", "top_p", "torch", "trailing_text_hidden", "tts_pad_embed", "tuple", "unsqueeze", "use_audio_in_video", "use_cache", "video_grid_thw", "video_second_per_grid", "view", "vision_idx", "vocab_size", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCausalConvNet": ["Conv1d", "F", "ModelCausalConvNet", "Module", "Tensor", "__init__", "_get_extra_padding_for_conv1d", "ceil", "class", "constant", "contiguous", "conv", "def", "dilation", "extra_padding", "forward", "groups", "hidden_state", "ideal_length", "in_channels", "int", "kernel_size", "length", "math", "mode", "n_frames", "nn", "out_channels", "pad", "padding", "return", "self", "shape", "stride", "super", "torch", "value"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCausalTransConvNet": ["ConvTranspose1d", "ModelCausalTransConvNet", "Module", "__init__", "ceil", "class", "contiguous", "conv", "def", "forward", "hidden_state", "in_channels", "kernel_size", "left_pad", "math", "nn", "out_channels", "pad", "return", "right_pad", "self", "shape", "stride", "super"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeConvNeXtBlock": ["GELU", "LayerNorm", "Linear", "ModelCausalConvNet", "ModelConvNeXtBlock", "Module", "Parameter", "__init__", "act", "class", "def", "dilation", "dim", "dwconv", "eps", "forward", "gamma", "groups", "hidden_states", "input", "int", "kernel_size", "nn", "norm", "ones", "permute", "pwconv1", "pwconv2", "return", "self", "super", "torch"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavRotatoryEmbedding": ["False", "ModelCode2WavRotatoryEmbedding", "ModelConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Identity", "Linear", "LongTensor", "ModelCode2WavAttention", "ModelCode2WavConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavMlp": ["ACT2FN", "False", "Linear", "ModelCode2WavMlp", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavRMSNorm": ["ModelCode2WavRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavLayerScale": ["ModelCode2WavLayerScale", "Module", "Parameter", "Tensor", "True", "__init__", "channels", "class", "config", "def", "forward", "full", "hidden_size", "initial_scale", "layer_scale_initial_scale", "nn", "requires_grad", "return", "scale", "self", "super", "torch", "x"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavTransformerLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelCode2WavAttention", "ModelCode2WavConfig", "ModelCode2WavLayerScale", "ModelCode2WavMlp", "ModelCode2WavRMSNorm", "ModelCode2WavTransformerLayer", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "forward", "hidden_size", "hidden_states", "input_layernorm", "kwargs", "layer_idx", "mlp", "mlp_layer_scale", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_layer_scale", "sliding_attention", "super", "torch", "tuple", "use_cache"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavTransformerModel": ["BaseModelOutputWithPast", "DynamicCache", "False", "ModelCode2WavAttention", "ModelCode2WavConfig", "ModelCode2WavTransformerLayer", "ModelCode2WavTransformerModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "ValueError", "You", "__init__", "_can_record_outputs", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "expected", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_types", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "window_size"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:SnakeBeta": ["ModelBeta", "Module", "Parameter", "__init__", "alpha", "beta", "class", "def", "exp", "forward", "hidden_states", "in_features", "nn", "no_div_by_zero", "pow", "return", "self", "sin", "super", "torch", "unsqueeze", "zeros"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavDecoderResidualUnit": ["ModelCausalConvNet", "ModelCode2WavDecoderResidualUnit", "Module", "SnakeBeta", "__init__", "act1", "act2", "class", "conv1", "conv2", "def", "dilation", "dim", "forward", "hidden_state", "int", "kernel_size", "nn", "residual", "return", "self", "super"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2WavDecoderBlock": ["ModelCausalTransConvNet", "ModelCode2WavConfig", "ModelCode2WavDecoderBlock", "ModelCode2WavDecoderResidualUnit", "ModelPreTrainedModel", "ModuleList", "SnakeBeta", "__init__", "append", "block", "class", "config", "decoder_dim", "def", "dilation", "for", "forward", "hidden", "in", "in_dim", "layer_idx", "nn", "out_dim", "return", "self", "super", "upsample_rate", "upsample_rates"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeCode2Wav": ["Embedding", "Expected", "False", "ModelCausalConvNet", "ModelCausalTransConvNet", "ModelCode2Wav", "ModelCode2WavConfig", "ModelCode2WavDecoderBlock", "ModelCode2WavTransformerModel", "ModelConvNeXtBlock", "ModelPreTrainedModel", "ModuleList", "SnakeBeta", "ValueError", "__init__", "_from_config", "append", "arange", "block", "blocks", "cat", "chunk_size", "chunked_decode", "clamp", "class", "code_embedding", "code_offset", "codebook_size", "codes", "codes_chunk", "config", "context_size", "decoder", "decoder_dim", "def", "dim", "else", "end_index", "f", "factor", "for", "forward", "got", "hidden", "hidden_size", "i", "if", "in", "inputs_embeds", "last_hidden_state", "layer", "left_context_size", "len", "max", "mean", "min", "nn", "np", "num_quantizers", "of", "output_dim", "permute", "persistent", "post_init", "pre_transformer", "prod", "raise", "range", "register_buffer", "return", "self", "shape", "start_index", "super", "torch", "total_upsample", "upsample", "upsample_rates", "upsampling_ratios", "view", "wav", "wav_chunk", "wavs", "while"], "qwen3_omni_moe/modeling_qwen3_omni_moe.py:Qwen3OmniMoeForConditionalGeneration": ["AssertionError", "Cannot", "Ethan", "Expect", "False", "GenerationMixin", "Model", "ModelCode2Wav", "ModelConfig", "ModelForConditionalGeneration", "ModelPreTrainedModel", "ModelTalkerForConditionalGeneration", "ModelThinkerForConditionalGeneration", "None", "NotImplementedError", "Omni", "Optional", "Speaker", "Tensor", "True", "Use", "ValueError", "__init__", "_from_config", "_get_talker_assistant_parts", "_get_talker_user_parts", "accept_hidden_layer", "after", "and", "any", "append", "assistant", "assistant_codec_hidden", "assistant_hidden", "assistant_text_hidden", "assistant_token_id", "attention_mask", "audio", "audio_feature_lengths", "audio_token_id", "batched", "bool", "cat", "chunk", "chunk_size", "chunked_decode", "class", "code2wav", "code2wav_config", "codec_bos_id", "codec_eos_token_id", "codec_nothink_id", "codec_pad_id", "codec_special_tokens", "codec_think_bos_id", "codec_think_eos_id", "config", "config_class", "continue", "currently", "def", "del", "device", "dim", "disable_talker", "do_sample", "does", "dtype", "elif", "else", "embed", "empty", "enable", "enable_audio_output", "enable_talker", "eos_token_id", "expand", "f", "feature_attention_mask", "fill_value", "float", "for", "full", "generate", "generate_audio", "get", "get_input_embeddings", "has_talker", "hasattr", "hid", "hidden_projection", "hidden_size", "hidden_states", "i", "id", "if", "im_start", "im_start_index", "im_start_indexes", "im_start_token_id", "image_grid_thw", "image_token_id", "implemented", "in", "inference", "initialized", "input_embeds", "input_features", "input_ids", "inputs_embeds", "int", "is", "items", "key", "kwargs", "left_context_size", "len", "long", "lower", "max_new_tokens", "method", "mm_hidden", "module", "multimodal_mask", "no_grad", "nonzero", "not", "or", "output", "output_hidden_states", "post_init", "raise", "range", "repetition_penalty", "return", "return_audio", "return_dict_in_generate", "role", "role_token", "segment_end_index", "self", "sequences", "set", "shape", "shared_kwargs", "speaker", "speaker_id", "squeeze", "stack", "startswith", "str", "sum", "super", "support", "suppress_tokens", "system", "system_token_id", "talker", "talker_", "talker_assistant_embeds", "talker_assistant_ids", "talker_codes", "talker_config", "talker_do_sample", "talker_input_embed", "talker_input_embeds", "talker_input_id", "talker_input_ids", "talker_kwargs", "talker_max_new_tokens", "talker_repetition_penalty", "talker_result", "talker_special_tokens", "talker_supppressed_tokens", "talker_temperature", "talker_top_k", "talker_top_p", "talker_user_part", "talker_wavs", "temperature", "tensor", "text_config", "text_projection", "thinker", "thinker_", "thinker_config", "thinker_embed", "thinker_eos_token_id", "thinker_hidden", "thinker_kwargs", "thinker_max_new_tokens", "thinker_result", "to", "token2wav_", "token2wav_kwargs", "top_k", "top_p", "torch", "trailing_text_hidden", "transpose", "tts_bos_embed", "tts_bos_token_id", "tts_eos_embed", "tts_eos_token_id", "tts_pad_embed", "tts_pad_token_id", "use", "use_audio_in_video", "user", "user_mm_mask", "user_talker_part", "user_text_hidden", "user_thinker_embed", "user_thinker_hidden_mm", "user_token_id", "value", "video_grid_thw", "video_second_per_grid", "video_token_id", "vocab_size", "when", "with", "zeros"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormAttention": ["False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "hidden_states_pre_layer_norm", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "layer_norm_eps", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormIntermediate": ["ACT2FN", "LayerNorm", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormOutput": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelEmbeddings", "ModelLMHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "eps", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "nn", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "Model", "ModelForCausalLM", "ModelLMHead", "ModelLMHeadModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "device", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "device", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormLMHead": ["LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "view"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "to", "token_type_ids", "torch", "tuple", "view"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "view"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "roberta_prelayernorm/modeling_roberta_prelayernorm.py:RobertaPreLayerNormForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "univnet/modeling_univnet.py:UnivNetModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "r", "torch", "waveform_lengths", "waveforms"], "univnet/modeling_univnet.py:UnivNetKernelPredictorResidualBlock": ["Conv1d", "Dropout", "FloatTensor", "ModelConfig", "ModelKernelPredictorResidualBlock", "Module", "True", "__init__", "apply_weight_norm", "bias", "channels", "class", "config", "conv1", "conv2", "def", "dropout", "dropout_prob", "forward", "functional", "hasattr", "hidden_states", "if", "kernel_predictor_conv_size", "kernel_predictor_dropout", "kernel_size", "leaky_relu", "leaky_relu_slope", "model_in_channels", "nn", "padding", "parametrizations", "remove_weight_norm", "residual", "return", "self", "super", "torch", "utils", "weight_norm"], "univnet/modeling_univnet.py:UnivNetKernelPredictor": ["Conv1d", "FloatTensor", "ModelConfig", "ModelKernelPredictor", "ModelKernelPredictorResidualBlock", "Module", "ModuleList", "True", "_", "__init__", "apply_weight_norm", "batch_size", "bias", "bias_channels", "bias_conv", "bias_hidden_states", "biases", "class", "config", "contiguous", "conv_in_channels", "conv_kernel_size", "conv_layers", "conv_out_channels", "def", "for", "forward", "functional", "hasattr", "hidden_states", "if", "in", "input_conv", "int", "kernel_channels", "kernel_conv", "kernel_hidden_states", "kernel_predictor_conv_size", "kernel_predictor_hidden_channels", "kernel_predictor_num_blocks", "kernels", "layer", "leaky_relu", "leaky_relu_slope", "model_hidden_channels", "nn", "num_blocks", "num_mel_bins", "padding", "parametrizations", "range", "remove_weight_norm", "resblock", "resblocks", "resnet_hidden_channels", "resnet_in_channels", "resnet_kernel_size", "return", "self", "seq_length", "shape", "spectrogram", "super", "torch", "utils", "view", "weight_norm"], "univnet/modeling_univnet.py:UnivNetLvcResidualBlock": ["Conv1d", "Dim", "FloatTensor", "ModelConfig", "ModelLvcResidualBlock", "Module", "Please", "ValueError", "_", "__init__", "and", "apply_weight_norm", "are", "batch", "be", "bias", "bildsk", "biokl", "bolsd", "but", "channels_last_3d", "check", "class", "config", "constant", "contiguous", "conv", "correct", "def", "dilation", "einsum", "f", "forward", "functional", "got", "hasattr", "hidden_channels", "hidden_states", "hop_size", "if", "in_length", "int", "kernel", "kernel_length", "kernel_size", "leaky_relu", "leaky_relu_slope", "location_variable_convolution", "make", "memory_format", "model_hidden_channels", "nn", "of", "or", "out_channels", "output_hidden_states", "pad", "padding", "parametrizations", "raise", "remove_weight_norm", "residual", "return", "self", "shape", "should", "sigmoid", "super", "sure", "tanh", "they", "to", "torch", "transpose", "unfold", "unsqueeze", "utils", "view", "weight_norm"], "univnet/modeling_univnet.py:UnivNetLvcBlock": ["ConvTranspose1d", "FloatTensor", "ModelConfig", "ModelKernelPredictor", "ModelLvcBlock", "ModelLvcResidualBlock", "Module", "ModuleList", "__init__", "apply_weight_norm", "bias", "biases", "class", "cond_hop_length", "config", "convt_pre", "def", "dilations", "enumerate", "for", "forward", "functional", "hasattr", "hidden_channels", "hidden_states", "hop_size", "i", "if", "in", "int", "kernel", "kernel_predictor", "kernel_size", "kernels", "layer", "layer_id", "leaky_relu", "leaky_relu_slope", "len", "lvc_hop_size", "model_hidden_channels", "nn", "num_blocks", "output_padding", "padding", "parametrizations", "range", "remove_weight_norm", "resblock", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblock_stride_sizes", "resblocks", "return", "self", "spectrogram", "stride", "super", "torch", "utils", "weight_norm"], "univnet/modeling_univnet.py:UnivNetModel": ["Conv1d", "ConvTranspose1d", "FloatTensor", "Generator", "Linear", "ModelConfig", "ModelLvcBlock", "ModelModel", "ModelModelOutput", "ModuleList", "None", "Optional", "PreTrainedModel", "The", "Union", "ValueError", "_", "__init__", "_init_weights", "and", "append", "apply_weight_norm", "are", "auto_docstring", "batch", "be", "bias", "bool", "but", "class", "config", "conv_post", "conv_pre", "data", "def", "device", "dim", "dtype", "elif", "else", "equal", "expected", "f", "for", "forward", "functional", "generator", "hasattr", "hidden_states", "hop_length", "hop_lengths", "i", "if", "in", "initializer_range", "input_features", "is", "isinstance", "kernel_size", "layer", "layer_id", "leaky_relu", "leaky_relu_slope", "len", "lvc_hop_size", "main_input_name", "mean", "model_hidden_channels", "model_in_channels", "module", "nn", "noise_sequence", "noise_sequence_batch_size", "noise_sequence_batched", "noise_sequence_shape", "normal_", "not", "num_kernels", "num_layers", "of", "outputs", "padding", "padding_mask", "padding_mask_batch_size", "padding_mode", "parametrizations", "post_init", "r", "raise", "randn", "range", "reflect", "remove_weight_norm", "repeat", "resblock", "resblock_kernel_sizes", "resblock_stride_sizes", "resblocks", "return", "return_dict", "self", "shape", "size", "spectrogram_batch_size", "spectrogram_batched", "spectrogram_length", "squeeze", "std", "stride", "sum", "super", "tanh", "the", "to", "torch", "transpose", "tuple", "two", "unsqueeze", "use_return_dict", "utils", "waveform", "waveform_lengths", "waveforms", "weight", "weight_norm", "zero_"], "fnet/modeling_fnet.py:_two_dim_matmul": ["_two_dim_matmul", "bij", "bnk", "complex64", "def", "einsum", "jk", "matrix_dim_one", "matrix_dim_two", "ni", "return", "seq_length", "shape", "torch", "type", "x"], "fnet/modeling_fnet.py:two_dim_matmul": ["Model_dim_matmul", "_Model_dim_matmul", "def", "matrix_dim_Model", "matrix_dim_one", "return", "x"], "fnet/modeling_fnet.py:fftn": ["Model", "axis", "def", "fft", "for", "in", "ndim", "out", "range", "return", "reversed", "torch", "x"], "fnet/modeling_fnet.py:FNetEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Linear", "Model", "Module", "None", "__init__", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "projection", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "fnet/modeling_fnet.py:FNetBasicFourierTransform": ["DFT", "Model", "Module", "SciPy", "TPU", "Using", "__init__", "_init_fourier_transform", "and", "calculation", "class", "complex64", "config", "def", "dft", "dft_mat_hidden", "dft_mat_seq", "dim", "dtype", "elif", "else", "fast", "fft", "fftn", "for", "forward", "found", "fourier", "fourier_transform", "hidden_size", "hidden_states", "if", "instead", "is", "is_scipy_available", "linalg", "logging", "matrix", "matrix_dim_one", "matrix_dim_two", "max_position_embeddings", "needed", "nn", "not", "optimized", "outputs", "partial", "real", "register_buffer", "return", "self", "super", "tensor", "torch", "tpu_short_seq_length", "transform", "two_dim_matmul", "use_tpu_fourier_optimizations", "warning"], "fnet/modeling_fnet.py:FNetBasicOutput": ["LayerNorm", "Model", "Module", "__init__", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super"], "fnet/modeling_fnet.py:FNetFourierTransform": ["Model", "ModelBasicFourierTransform", "ModelBasicOutput", "Module", "__init__", "class", "config", "def", "forward", "fourier_output", "hidden_states", "nn", "output", "outputs", "return", "self", "self_outputs", "super"], "fnet/modeling_fnet.py:FNetIntermediate": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "fnet/modeling_fnet.py:FNetOutput": ["Dropout", "LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "fnet/modeling_fnet.py:FNetLayer": ["GradientCheckpointingLayer", "Model", "ModelFourierTransform", "ModelIntermediate", "ModelOutput", "__init__", "apply_chunking_to_forward", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "fourier", "fourier_output", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "outputs", "return", "self", "self_fourier_outputs", "seq_len_dim", "super"], "fnet/modeling_fnet.py:FNetEncoder": ["BaseModelOutput", "False", "Model", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "fnet/modeling_fnet.py:FNetPooler": ["Linear", "Model", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "fnet/modeling_fnet.py:FNetPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "fnet/modeling_fnet.py:FNetLMPredictionHead": ["Linear", "Model", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "meta", "nn", "return", "self", "super", "torch", "transform", "type", "vocab_size", "zeros"], "fnet/modeling_fnet.py:FNetOnlyMLMHead": ["Model", "ModelLMPredictionHead", "Module", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super"], "fnet/modeling_fnet.py:FNetOnlyNSPHead": ["Linear", "Model", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "fnet/modeling_fnet.py:FNetPreTrainingHeads": ["Linear", "Model", "ModelLMPredictionHead", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "fnet/modeling_fnet.py:FNetPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "fnet/modeling_fnet.py:FNetForPreTrainingOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "fnet/modeling_fnet.py:FNetModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "TPU", "The", "True", "Union", "ValueError", "You", "__init__", "add_pooling_layer", "and", "at", "auto_docstring", "batch_size", "be", "being", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "equal", "expand", "forward", "get_input_embeddings", "hasattr", "have", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "length", "long", "model", "not", "optimizations", "or", "output_hidden_states", "passed", "pooler", "pooler_output", "position_ids", "post_init", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence", "sequence_output", "set", "set_input_embeddings", "should", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tpu_short_seq_length", "tuple", "use_return_dict", "use_tpu_fourier_optimizations", "using", "value", "when", "word_embeddings", "zeros"], "fnet/modeling_fnet.py:FNetForPreTraining": ["CrossEntropyLoss", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelPreTrainingHeads", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "next_sentence_label", "next_sentence_loss", "not", "output", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "fnet/modeling_fnet.py:FNetForMaskedLM": ["CrossEntropyLoss", "MaskedLMOutput", "Model", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "fnet/modeling_fnet.py:FNetForNextSentencePrediction": ["CrossEntropyLoss", "FutureWarning", "Model", "ModelModel", "ModelOnlyNSPHead", "ModelPreTrainedModel", "NextSentencePredictorOutput", "None", "Optional", "Tensor", "The", "Union", "__init__", "a", "and", "argument", "auto_docstring", "be", "bool", "class", "cls", "config", "def", "deprecated", "else", "forward", "future", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "output", "output_hidden_states", "outputs", "pooled_output", "pop", "position_ids", "post_init", "r", "removed", "return", "return_dict", "self", "seq_relationship_scores", "super", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "version", "view", "warn", "warnings", "will"], "fnet/modeling_fnet.py:FNetForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "dtype", "elif", "else", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "fnet/modeling_fnet.py:FNetForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "fnet/modeling_fnet.py:FNetForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "fnet/modeling_fnet.py:FNetForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "mobilenet_v1/modeling_mobilenet_v1.py:apply_tf_padding": ["Conv2d", "Model_tf_padding", "Tensor", "constant", "conv_layer", "def", "else", "features", "functional", "if", "in_height", "in_width", "kernel_height", "kernel_size", "kernel_width", "max", "nn", "pad", "pad_along_height", "pad_along_width", "pad_bottom", "pad_left", "pad_right", "pad_top", "padding", "return", "shape", "stride", "stride_height", "stride_width", "torch"], "mobilenet_v1/modeling_mobilenet_v1.py:MobileNetV1ConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Input", "ModelConfig", "ModelConvLayer", "Module", "None", "Optional", "Output", "Tensor", "True", "Union", "ValueError", "__init__", "activation", "affine", "apply_tf_padding", "are", "bias", "bool", "by", "channels", "class", "config", "convolution", "def", "divisible", "elif", "else", "eps", "f", "features", "forward", "groups", "hidden_act", "if", "in_channels", "int", "is", "isinstance", "kernel_size", "layer_norm_eps", "momentum", "nn", "normalization", "not", "num_features", "out_channels", "padding", "padding_mode", "raise", "return", "self", "str", "stride", "super", "tf_padding", "torch", "track_running_stats", "use_activation", "use_normalization", "zeros"], "mobilenet_v1/modeling_mobilenet_v1.py:MobileNetV1PreTrainedModel": ["BatchNorm2d", "Conv2d", "False", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "Union", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "mobilenet_v1/modeling_mobilenet_v1.py:MobileNetV1Model": ["AdaptiveAvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "ModelConfig", "ModelConvLayer", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "NotImplementedError", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "all_hidden_states", "append", "auto_docstring", "bool", "class", "config", "conv_stem", "def", "depth", "depth_multiplier", "else", "enumerate", "flatten", "for", "forward", "groups", "have", "heads_to_prune", "hidden_states", "i", "if", "in", "in_channels", "int", "is", "kernel_size", "last_hidden_state", "layer", "layer_module", "max", "min_depth", "nn", "not", "num_channels", "or", "out_channels", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "r", "raise", "range", "return", "return_dict", "self", "specify", "start_dim", "stride", "strides", "super", "to", "torch", "tuple", "use_return_dict", "v"], "mobilenet_v1/modeling_mobilenet_v1.py:MobileNetV1ForImageClassification": ["Dropout", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "classifier_dropout_prob", "config", "convolution", "def", "dropout", "else", "forward", "hidden_states", "if", "inplace", "is", "labels", "last_hidden_size", "layer", "logits", "loss", "loss_function", "nn", "not", "num_labels", "out_channels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "jetmoe/modeling_jetmoe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "jetmoe/modeling_jetmoe.py:JetMoeParallelExperts": ["F", "ModelParallelExperts", "Module", "None", "Parameter", "__init__", "append", "cat", "class", "def", "dim", "empty", "expert_size", "for", "forward", "i", "in", "input_list", "input_size", "inputs", "int", "linear", "nn", "num_experts", "output_list", "output_size", "range", "results", "return", "self", "split", "super", "torch", "weight"], "jetmoe/modeling_jetmoe.py:JetMoeTopKGating": ["False", "Linear", "ModelTopKGating", "Module", "_", "__init__", "batch_gates", "batch_index", "bias", "class", "def", "device", "dim", "div", "dtype", "expert_size", "flatten", "float", "forward", "gates", "hidden_states", "index_sorted_experts", "input_size", "int", "layer", "logits", "long", "nn", "num_experts", "return", "rounding_mode", "scatter", "self", "size", "softmax", "sort", "sum", "super", "tolist", "top_k", "top_k_experts", "top_k_gates", "top_k_indices", "top_k_logits", "topk", "torch", "trunc", "type_as", "zeros"], "jetmoe/modeling_jetmoe.py:JetMoeMoE": ["ACT2FN", "ModelConfig", "ModelMoE", "ModelParallelExperts", "ModelTopKGating", "Module", "None", "Parameter", "_", "__init__", "activation", "activation_function", "batch_gates", "batch_index", "bias", "bsz", "chunk", "chunked_hidden_states", "class", "config", "def", "device", "dim", "dtype", "emb_size", "empty", "expert_inputs", "expert_outputs", "expert_size", "forward", "hidden_size", "hidden_states", "index_add", "input_linear", "input_size", "intermediate_size", "layer_input", "layer_output", "length", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "output_linear", "reshape", "return", "router", "router_logits", "self", "size", "super", "top_k", "torch", "view", "zeros"], "jetmoe/modeling_jetmoe.py:JetMoeMoA": ["ModelConfig", "ModelMoA", "ModelParallelExperts", "ModelTopKGating", "Module", "None", "NotImplementedError", "Parameter", "This", "__init__", "and", "batch_gates", "batch_index", "bias", "bsz", "call", "class", "config", "def", "device", "doesn", "dtype", "emb_size", "empty", "expert_inputs", "expert_outputs", "expert_size", "forward", "hidden_size", "index_add", "index_sorted_experts", "input_linear", "input_size", "k", "kv_channels", "layer_input", "layer_output", "length", "map", "module", "nn", "num_experts", "num_experts_per_tok", "num_key_value_heads", "num_local_experts", "output_linear", "raise", "reduce", "reshape", "return", "router", "router_logits", "self", "size", "super", "support", "t", "top_k", "topo_info", "torch", "view", "zeros"], "jetmoe/modeling_jetmoe.py:JetMoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "jetmoe/modeling_jetmoe.py:JetMoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "jetmoe/modeling_jetmoe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "jetmoe/modeling_jetmoe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "jetmoe/modeling_jetmoe.py:JetMoeAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelMoA", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "chunk", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "dropout", "dtype", "during", "errors", "experts", "f", "float32", "forward", "functional", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key_states", "kv_channels", "kv_proj", "kv_projection_size", "layer_idx", "lead", "logger", "make", "map", "math", "matmul", "new_name", "nn", "not", "num_attention_heads", "num_experts_per_tok", "num_heads", "num_key_value_heads", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "query_states", "raise", "recommended", "reduce", "repeat", "reshape", "return", "rotary_emb", "router_logits", "self", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "top_k", "topo_info", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "value_states", "version", "view", "warning_once", "when", "will", "without"], "jetmoe/modeling_jetmoe.py:JetMoeSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "chunk", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "dim", "does", "dropout_p", "eager", "else", "experts", "forward", "from", "functional", "head_dim", "hidden_states", "if", "implementation", "is", "is_causal", "key_states", "kv_proj", "kv_projection_size", "layer_idx", "loading", "logger", "manual", "map", "model", "new_name", "nn", "not", "num_heads", "num_key_value_heads", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "query_states", "reduce", "removed", "repeat", "required", "reshape", "return", "rotary_emb", "router_logits", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "top_k", "topo_info", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "jetmoe/modeling_jetmoe.py:JetMoeFlashAttention2": ["Cache", "False", "FloatTensor", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "The", "Union", "We", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "apply_rotary_pos_emb", "args", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache_kwargs", "cache_position", "cast", "casted", "chunk", "class", "config", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dim", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "experts", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "key_states", "kv_proj", "kv_projection_size", "kwargs", "layer", "layer_idx", "layers", "logger", "map", "might", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "or", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "query_states", "reduce", "related", "repeat", "reshape", "return", "rotary_emb", "router_logits", "seems", "self", "silently", "sin", "size", "states", "super", "target_dtype", "the", "this", "to", "top_k", "topo_info", "torch", "training", "transpose", "tuple", "type", "upcasted", "update", "use_cache", "use_top_left_mask", "value_states", "version", "view", "warning_once", "weight", "will", "you"], "jetmoe/modeling_jetmoe.py:JetMoeBlock": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelBlock", "ModelConfig", "ModelMoE", "ModelRMSNorm", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "Union", "__init__", "_attn_implementation", "attention_mask", "attn_output", "attn_router_logits", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "layer_idx", "mlp", "mlp_router_logits", "new_name", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "return", "self", "self_attention", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version", "x_mlp"], "jetmoe/modeling_jetmoe.py:JetMoePreTrainedModel": ["Embedding", "False", "Linear", "ModelBlock", "ModelConfig", "ModelMoA", "ModelMoE", "ModelParallelExperts", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "jetmoe/modeling_jetmoe.py:JetMoeModel": ["AttentionMaskConverter", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelBlock", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "list", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_ids", "post_init", "raise", "range", "reshape", "return", "rms_norm_eps", "router_logits", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "type", "unsqueeze", "use_cache", "using_compilable_cache", "vocab_size", "warning_once", "with", "xpu"], "jetmoe/modeling_jetmoe.py:JetMoeForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "aux_loss", "aux_loss_coef", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_logits", "self", "slice", "slice_indices", "super", "tie_word_embeddings", "to", "torch", "use_cache", "vocab_size", "weight"], "jetmoe/modeling_jetmoe.py:JetMoeForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "dinov3_convnext/modeling_dinov3_convnext.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "dinov3_convnext/modeling_dinov3_convnext.py:DINOv3ConvNextDropPath": ["Model", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "dinov3_convnext/modeling_dinov3_convnext.py:DINOv3ConvNextLayerNorm": ["LayerNorm", "Model", "NotImplementedError", "Tensor", "Unsupported", "__init__", "args", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "dinov3_convnext/modeling_dinov3_convnext.py:DINOv3ConvNextLayer": ["ACT2FN", "Conv2d", "Identity", "Linear", "Model", "ModelConfig", "ModelDropPath", "ModelNorm", "Module", "Parameter", "Tensor", "True", "__init__", "activation_fn", "channels", "class", "config", "def", "depthwise_conv", "drop_path", "else", "eps", "features", "float", "forward", "full", "gamma", "groups", "hidden_act", "if", "int", "kernel_size", "layer_norm", "layer_norm_eps", "layer_scale_init_value", "nn", "padding", "permute", "pointwise_conv1", "pointwise_conv2", "requires_grad", "residual", "return", "self", "super", "torch"], "dinov3_convnext/modeling_dinov3_convnext.py:DINOv3ConvNextStage": ["Conv2d", "Model", "ModelConfig", "ModelLayer", "ModelLayerNorm", "Module", "ModuleList", "Tensor", "__init__", "channels", "channels_first", "class", "config", "data_format", "def", "depths", "downsample_layers", "drop_path", "drop_path_rate", "drop_path_rates", "else", "eps", "features", "for", "forward", "hidden_sizes", "i", "if", "in", "in_channels", "int", "kernel_size", "layer", "layer_norm_eps", "layers", "linspace", "nn", "np", "num_channels", "num_previous_layers", "num_stage_layers", "num_total_layers", "out_channels", "range", "return", "self", "stage_idx", "stride", "sum", "super", "tolist", "torch"], "dinov3_convnext/modeling_dinov3_convnext.py:DINOv3ConvNextPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLayer", "ModelLayerNorm", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "gamma", "if", "initializer_range", "is", "isinstance", "layer_scale_init_value", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "dinov3_convnext/modeling_dinov3_convnext.py:DINOv3ConvNextModel": ["AdaptiveAvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "LayerNorm", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelStage", "ModuleList", "None", "Optional", "__init__", "all_hidden_states", "append", "auto_docstring", "bool", "can_return_tuple", "cat", "class", "config", "def", "dim", "else", "eps", "flatten", "for", "forward", "hidden_sizes", "hidden_states", "if", "in", "last_hidden_state", "layer_norm", "layer_norm_eps", "nn", "num_stages", "or", "output_hidden_states", "pixel_values", "pool", "pooled_output", "pooler_output", "post_init", "range", "return", "self", "stage", "stage_idx", "stages", "super", "torch", "transpose", "tuple"], "splinter/modeling_splinter.py:SplinterEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "__init__", "absolute", "arange", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "tuple", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "splinter/modeling_splinter.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "splinter/modeling_splinter.py:SplinterSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "bool", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_states", "kwargs", "multiple", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "query", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_states", "view"], "splinter/modeling_splinter.py:SplinterSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "splinter/modeling_splinter.py:SplinterAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "splinter/modeling_splinter.py:SplinterIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "splinter/modeling_splinter.py:SplinterOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "splinter/modeling_splinter.py:SplinterLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "splinter/modeling_splinter.py:SplinterEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "splinter/modeling_splinter.py:SplinterPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "splinter/modeling_splinter.py:SplinterModel": ["BaseModelOutput", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "can_return_tuple", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "splinter/modeling_splinter.py:SplinterFullyConnectedLayer": ["ACT2FN", "LayerNorm", "Linear", "ModelFullyConnectedLayer", "Module", "Tensor", "__init__", "act_fn", "class", "def", "dense", "forward", "gelu", "hidden_act", "hidden_states", "input_dim", "inputs", "nn", "output_dim", "return", "self", "super", "torch"], "splinter/modeling_splinter.py:QuestionAwareSpanSelectionHead": ["False", "Linear", "ModelAwareSpanSelectionHead", "ModelFullyConnectedLayer", "Module", "_", "__init__", "bias", "class", "config", "def", "dim", "end_classifier", "end_logits", "end_reps", "end_transform", "forward", "gather", "gathered_reps", "hidden_size", "hidden_states", "index", "inputs", "matmul", "nn", "permute", "positions", "query_end_reps", "query_end_transform", "query_start_reps", "query_start_transform", "repeat", "return", "self", "size", "start_classifier", "start_logits", "start_reps", "start_transform", "super", "torch", "unsqueeze"], "splinter/modeling_splinter.py:SplinterForQuestionAnswering": ["CrossEntropyLoss", "False", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "Model_qass", "None", "Optional", "QuestionAnsweringModelOutput", "QuestionAwareSpanSelectionHead", "Tensor", "True", "Union", "__init__", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "clamp_", "class", "config", "def", "device", "dim", "dtype", "else", "end_logits", "end_loss", "end_positions", "eq", "finfo", "forward", "head_mask", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "int", "is", "layout", "len", "long", "loss", "loss_fct", "min", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "question_position_for_each_example", "question_positions", "question_positions_were_none", "question_token_id", "r", "return", "return_dict", "self", "sequence_output", "size", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "unsqueeze", "use_return_dict", "zeros"], "splinter/modeling_splinter.py:SplinterForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "end_logits", "hidden_states", "loss", "r", "start_logits", "torch", "tuple"], "splinter/modeling_splinter.py:SplinterForPreTraining": ["CrossEntropyLoss", "LongTensor", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "Model_qass", "None", "Optional", "QuestionAwareSpanSelectionHead", "Tensor", "TypeError", "Union", "__init__", "_prepare_question_positions", "and", "arange", "attention_mask", "attention_mask_for_each_question", "attentions", "auto_docstring", "batch_size", "be", "bincount", "bool", "calculate", "cat", "clamp_", "class", "cols", "config", "def", "device", "dim", "dtype", "elif", "else", "end_logits", "end_loss", "end_positions", "expand", "finfo", "flat_positions", "for", "forward", "full", "head_mask", "hidden_states", "if", "ignore_index", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "long", "loss", "loss_fct", "max", "min", "must", "n", "not", "num_questions", "order", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "position_ids", "positions", "post_init", "question_positions", "question_token_id", "r", "raise", "return", "return_dict", "rows", "self", "sequence_length", "sequence_output", "size", "specified", "start_logits", "start_loss", "start_positions", "super", "the", "to", "token_type_ids", "torch", "total_loss", "tuple", "unsqueeze", "use_return_dict", "used", "view", "when", "where"], "vitpose/modeling_vitpose.py:VitPoseEstimatorOutput": ["FloatTensor", "ModelEstimatorOutput", "ModelOutput", "None", "Optional", "attentions", "class", "heatmaps", "hidden_states", "loss", "r", "torch", "tuple"], "vitpose/modeling_vitpose.py:VitPosePreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "Union", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "dtype", "elif", "fill_", "float32", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "vitpose/modeling_vitpose.py:flip_back": ["Model", "Model_back", "Model_pairs", "ValueError", "batch_size", "be", "channels", "clone", "combined", "def", "for", "gaussian", "heatmap", "height", "if", "in", "left", "ndim", "not", "num_keypoints", "or", "output_Modelped", "output_Modelped_back", "raise", "reshape", "return", "right", "shape", "should", "target", "target_type", "tolist", "width"], "vitpose/modeling_vitpose.py:VitPoseSimpleDecoder": ["Conv2d", "False", "ModelConfig", "ModelSimpleDecoder", "Module", "None", "Optional", "ReLU", "Tensor", "Upsample", "__init__", "activation", "align_corners", "backbone_config", "bilinear", "class", "config", "conv", "def", "flip_back", "flip_pairs", "forward", "heatmaps", "hidden_size", "hidden_state", "if", "is", "kernel_size", "mode", "nn", "not", "num_labels", "padding", "return", "scale_factor", "self", "stride", "super", "torch", "upsampling"], "vitpose/modeling_vitpose.py:VitPoseClassicDecoder": ["BatchNorm2d", "Conv2d", "ConvTranspose2d", "False", "ModelClassicDecoder", "ModelConfig", "Module", "None", "Optional", "ReLU", "Tensor", "__init__", "backbone_config", "batchnorm1", "batchnorm2", "bias", "class", "config", "conv", "deconv1", "deconv2", "def", "flip_back", "flip_pairs", "forward", "heatmaps", "hidden_size", "hidden_state", "if", "is", "kernel_size", "nn", "not", "num_labels", "padding", "relu1", "relu2", "return", "self", "stride", "super", "torch"], "vitpose/modeling_vitpose.py:VitPoseForPoseEstimation": ["BackboneOutput", "ModelClassicDecoder", "ModelConfig", "ModelEstimatorOutput", "ModelForPoseEstimation", "ModelPreTrainedModel", "ModelSimpleDecoder", "None", "NotImplementedError", "Optional", "Tensor", "The", "Training", "TransformersKwargs", "Unpack", "ValueError", "__init__", "a", "an", "attentions", "attribute", "auto_docstring", "backbone", "backbone_config", "batch_size", "can_return_tuple", "class", "config", "contiguous", "dataset_index", "def", "else", "feature_maps", "flip_pairs", "forward", "forward_with_filtered_kwargs", "hasattr", "have", "head", "heatmaps", "hidden_size", "hidden_states", "if", "image_size", "is", "kwargs", "labels", "load_backbone", "loss", "not", "outputs", "patch_height", "patch_size", "patch_width", "permute", "pixel_values", "post_init", "r", "raise", "reshape", "return", "self", "sequence_output", "shape", "should", "super", "supported", "torch", "use_simple_decoder", "yet"], "gpt2/modeling_gpt2.py:eager_attention_forward": ["Model_attention_forward", "None", "attention_mask", "attn_dropout", "attn_output", "attn_weights", "bias", "causal_mask", "def", "device", "dim", "dtype", "finfo", "float", "full", "functional", "head_mask", "if", "is", "is_cross_attention", "key", "key_length", "kwargs", "layer_idx", "mask_value", "matmul", "min", "module", "nn", "not", "query", "query_length", "return", "scale_attn_by_inverse_layer_idx", "scale_attn_weights", "shape", "size", "softmax", "to", "torch", "transpose", "type", "value", "where"], "gpt2/modeling_gpt2.py:GPT2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Conv1D", "Dropout", "EncoderDecoderCache", "Error", "False", "FloatTensor", "If", "LongTensor", "Model", "Module", "None", "Optional", "Please", "RuntimeError", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_attn_implementation", "_upcast_and_reordered_attn", "alpha", "and", "as", "attention", "attention_interface", "attention_mask", "attn_dropout", "attn_output", "attn_pdrop", "attn_weights", "autocast", "baddbmm", "be", "beta", "bias", "bool", "bsz", "by", "c_attn", "c_proj", "cache_position", "cat", "causal_mask", "class", "config", "contiguous", "cross", "cross_attention_cache", "curr_past_key_value", "def", "defined", "deprecate_kwarg", "device", "dim", "divisible", "dk", "does", "dropout", "dtype", "eager", "eager_attention_forward", "else", "embed_dim", "empty", "enabled", "encoder_attention_mask", "encoder_hidden_states", "f", "find_pruneable_heads_and_indices", "finfo", "float", "float32", "forward", "functional", "get", "got", "hasattr", "have", "head_dim", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "index_attn", "instantiate", "is", "is_causal", "is_cross_attention", "is_updated", "isinstance", "k", "k_seq_len", "key", "key_length", "key_states", "keys", "kwargs", "layer_idx", "layers", "len", "make", "mask_value", "masked_bias", "matmul", "max_position_embeddings", "max_positions", "min", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "ones", "or", "output_attentions", "p", "past_key_value", "past_key_values", "persistent", "prune_conv1d_layer", "prune_heads", "pruned_heads", "q", "q_attn", "q_seq_len", "query", "query_length", "query_states", "raise", "register_buffer", "reorder_and_upcast_attn", "reshape", "resid_dropout", "resid_pdrop", "return", "scale_attn_by_inverse_layer_idx", "scale_attn_weights", "scale_factor", "self", "self_attention_cache", "set", "shape", "shape_kv", "shape_q", "size", "softmax", "split", "split_size", "super", "sure", "tensor", "the", "to", "torch", "training", "transpose", "tril", "tuple", "type", "union", "upcasting", "update", "used", "using_eager", "value", "value_states", "values", "version", "view", "weights", "where", "with"], "gpt2/modeling_gpt2.py:GPT2MLP": ["ACT2FN", "Conv1D", "Dropout", "FloatTensor", "Model", "Module", "Optional", "__init__", "act", "activation_function", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "resid_pdrop", "return", "self", "super", "torch", "tuple"], "gpt2/modeling_gpt2.py:GPT2Block": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "LayerNorm", "LongTensor", "Model", "ModelAttention", "ModelMLP", "None", "Optional", "Tensor", "True", "Union", "ValueError", "__init__", "add_cross_attention", "are", "attention", "attention_mask", "attn", "attn_output", "be", "bool", "by", "cache_position", "class", "config", "cross", "cross_attn_output", "cross_attn_weights", "crossattention", "def", "deprecate_kwarg", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_hidden_states", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "inner_dim", "instantiated", "is", "is_cross_attention", "kwargs", "layer_idx", "layer_norm_epsilon", "layers", "ln_1", "ln_2", "ln_cross_attn", "mlp", "n_inner", "new_name", "nn", "not", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "raise", "residual", "return", "self", "self_attn_weights", "setting", "super", "to", "torch", "tuple", "use_cache", "version", "with"], "gpt2/modeling_gpt2.py:GPT2SequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelConfig", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "gpt2/modeling_gpt2.py:GPT2PreTrainedModel": ["Conv1D", "Embedding", "LayerNorm", "Linear", "Model", "ModelBlock", "ModelConfig", "None", "PreTrainedModel", "True", "__init__", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "c_proj", "class", "config", "data", "def", "elif", "fill_", "for", "if", "in", "initializer_range", "inputs", "is", "is_parallelizable", "isinstance", "kwargs", "math", "mean", "module", "n_layer", "name", "named_parameters", "nn", "normal_", "not", "p", "padding_idx", "past_key_values", "self", "sqrt", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "gpt2/modeling_gpt2.py:GPT2DoubleHeadsModelOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "mc_logits", "mc_loss", "past_key_values", "r", "torch", "tuple"], "gpt2/modeling_gpt2.py:GPT2Model": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DEPARALLELIZE_DOCSTRING", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "FutureWarning", "LayerNorm", "Like", "LongTensor", "Model", "ModelBlock", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "PARALLELIZE_DOCSTRING", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_supports_param_buffer_assignment", "_use_sdpa", "a", "add_cross_attention", "add_start_docstrings", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "also", "an", "and", "arange", "assert_device_map", "at", "attention_mask", "attentions", "attn", "auto_docstring", "balanced", "batch_size", "be", "block", "bool", "both", "but", "cache_position", "call", "can", "cannot", "causal_mask", "checkpointing", "class", "config", "cpu", "create_causal_mask", "cross_attentions", "cuda", "cuda_device", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "drop", "dtype", "e", "either", "elif", "else", "embd_pdrop", "embed_dim", "empty_cache", "encoder_attention_mask", "encoder_batch_size", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "first_device", "flash_attention_2", "for", "forward", "from_legacy_cache", "from_pretrained", "g", "get_device_map", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "h", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "index", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "isinstance", "it", "items", "k", "keys", "kwargs", "last_device", "last_hidden_state", "layer", "layer_idx", "layer_norm_epsilon", "len", "ln_f", "load", "logger", "mask", "max", "max_position_embeddings", "min", "model", "model_parallel", "module_name", "n_layer", "ndim", "needs", "new_embeddings", "nn", "not", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "own", "parallelize", "pass", "past_key_values", "past_seen_tokens", "position_embeds", "position_ids", "post_init", "provide", "prune_heads", "r", "raise", "range", "removed", "return", "return_dict", "same", "sdpa", "self", "set_device", "set_input_embeddings", "shape", "should", "size", "so", "specify", "str", "super", "tgt_len", "the", "time", "to", "token_type_embeds", "token_type_ids", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "v4", "v5", "view", "vocab_size", "warn", "warn_if_padding_and_no_attention_mask", "warning_once", "warnings", "will", "with", "wpe", "wte", "you", "your"], "gpt2/modeling_gpt2.py:GPT2LMHeadModel": ["Cache", "CausalLMOutputWithCrossAttentions", "DEPARALLELIZE_DOCSTRING", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Like", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PARALLELIZE_DOCSTRING", "Tensor", "Transformers", "True", "Union", "You", "__init__", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "bias", "bool", "but", "cache_position", "call", "can", "class", "config", "cpu", "cross_attentions", "cuda", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "else", "empty_cache", "encoder_attention_mask", "encoder_hidden_states", "first_device", "for", "forward", "from_pretrained", "get_device_map", "h", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instance", "int", "is", "isinstance", "it", "kwargs", "labels", "len", "lm_head", "load", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_parallel", "module_name", "n_embd", "needs", "nn", "not", "of", "output", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "position_ids", "post_init", "provide", "r", "range", "removed", "return", "return_dict", "self", "set_device", "should", "slice", "slice_indices", "so", "super", "the", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "v5", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "gpt2/modeling_gpt2.py:GPT2DoubleHeadsModel": ["Cache", "CrossEntropyLoss", "DEPARALLELIZE_DOCSTRING", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Like", "Linear", "LongTensor", "Model", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelSequenceSummary", "None", "Optional", "PARALLELIZE_DOCSTRING", "Transformers", "True", "Union", "You", "__init__", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "bias", "bool", "but", "cache_position", "call", "can", "class", "config", "contiguous", "cpu", "cuda", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "else", "empty_cache", "first_device", "for", "forward", "from_pretrained", "get_device_map", "h", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instance", "is", "it", "kwargs", "labels", "len", "lm_head", "lm_logits", "lm_loss", "load", "logits", "loss", "loss_fct", "mc_labels", "mc_logits", "mc_loss", "mc_token_ids", "model", "model_parallel", "module_name", "multiple_choice_head", "n_embd", "needs", "nn", "not", "num_labels", "of", "output", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "position_ids", "post_init", "provide", "r", "range", "removed", "return", "return_dict", "self", "set_device", "shift_labels", "shift_logits", "should", "size", "so", "squeeze", "super", "the", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "v5", "view", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "gpt2/modeling_gpt2.py:GPT2ForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "device_map", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "model_parallel", "multi_label_classification", "n_embd", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "sequence_length", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "gpt2/modeling_gpt2.py:GPT2ForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "device_map", "dropout", "elif", "else", "forward", "hasattr", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "model_parallel", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view"], "gpt2/modeling_gpt2.py:GPT2ForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "device", "device_map", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "model_parallel", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "to", "token_type_ids", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "ibert/modeling_ibert.py:IBertEmbeddings": ["Dropout", "False", "IntLayerNorm", "LayerNorm", "Model", "Module", "None", "QuantAct", "QuantEmbedding", "__init__", "absolute", "act_bit", "arange", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "def", "device", "dropout", "dtype", "else", "embedding_act_bit", "embedding_bit", "embeddings", "embeddings_act1", "embeddings_act2", "embeddings_scaling_factor", "eps", "expand", "force_dequant", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "identity", "identity_scaling_factor", "if", "input_ids", "input_shape", "inputs_embeds", "inputs_embeds_scaling_factor", "is", "layer_norm_eps", "ln_input_bit", "ln_output_bit", "long", "max_position_embeddings", "nn", "not", "output_activation", "output_bit", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_embeddings_scaling_factor", "position_ids", "quant_mode", "register_buffer", "return", "self", "sequence_length", "size", "super", "to", "token_type_embeddings", "token_type_embeddings_scaling_factor", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "weight_bit", "word_embeddings", "zeros"], "ibert/modeling_ibert.py:IBertSelfAttention": ["BERT", "Dropout", "False", "I", "IntSoftmax", "Model", "Module", "None", "QuantAct", "QuantLinear", "The", "True", "ValueError", "_", "__init__", "a", "absolute", "act_bit", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_probs_scaling_factor", "attention_scores", "attention_scores_scaling_factor", "batch_size", "bias", "bias_bit", "class", "config", "context_layer", "context_layer_scaling_factor", "contiguous", "def", "dropout", "else", "embedding_size", "f", "for", "force_dequant", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "hidden_states_scaling_factor", "if", "int", "is", "key", "key_activation", "key_layer", "key_layer_scaling_factor", "math", "matmul", "mixed_key_layer", "mixed_key_layer_scaling_factor", "mixed_query_layer", "mixed_query_layer_scaling_factor", "mixed_value_layer", "mixed_value_layer_scaling_factor", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "only", "output_activation", "output_attentions", "output_scaling_factor", "outputs", "per_channel", "permute", "position_embedding_type", "quant_mode", "query", "query_activation", "query_layer", "query_layer_scaling_factor", "raise", "return", "scale", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "supports", "the", "torch", "transpose", "value", "value_activation", "value_layer", "value_layer_scaling_factor", "view", "weight_bit"], "ibert/modeling_ibert.py:IBertSelfOutput": ["Dropout", "IntLayerNorm", "LayerNorm", "Model", "Module", "QuantAct", "QuantLinear", "True", "__init__", "act_bit", "bias", "bias_bit", "class", "config", "def", "dense", "dropout", "eps", "force_dequant", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "hidden_states_scaling_factor", "identity", "identity_scaling_factor", "input_tensor", "input_tensor_scaling_factor", "layer_norm_eps", "ln_input_act", "ln_input_bit", "ln_output_bit", "nn", "output_activation", "output_bit", "per_channel", "quant_mode", "return", "self", "super", "weight_bit"], "ibert/modeling_ibert.py:IBertAttention": ["False", "Model", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "attention_output_scaling_factor", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "hidden_states_scaling_factor", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "outputs_scaling_factor", "prune_heads", "prune_linear_layer", "pruned_heads", "quant_mode", "query", "return", "self", "self_outputs", "self_outputs_scaling_factor", "set", "super", "union", "value"], "ibert/modeling_ibert.py:IBertIntermediate": ["BERT", "I", "IntGELU", "Model", "Module", "QuantAct", "QuantLinear", "True", "ValueError", "__init__", "act_bit", "bias", "bias_bit", "class", "config", "def", "dense", "for", "force_dequant", "forward", "gelu", "hidden_act", "hidden_size", "hidden_states", "hidden_states_scaling_factor", "if", "intermediate_act_fn", "intermediate_size", "nn", "only", "output_activation", "per_channel", "quant_mode", "raise", "return", "self", "super", "supports", "weight_bit"], "ibert/modeling_ibert.py:IBertOutput": ["Dropout", "IntLayerNorm", "LayerNorm", "Model", "Module", "QuantAct", "QuantLinear", "True", "__init__", "act_bit", "bias", "bias_bit", "class", "config", "def", "dense", "dropout", "eps", "force_dequant", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "hidden_states_scaling_factor", "identity", "identity_scaling_factor", "input_tensor", "input_tensor_scaling_factor", "intermediate_size", "layer_norm_eps", "ln_input_act", "ln_input_bit", "ln_output_bit", "nn", "output_activation", "output_bit", "per_channel", "quant_mode", "return", "self", "super", "weight_bit"], "ibert/modeling_ibert.py:IBertLayer": ["False", "Model", "ModelAttention", "ModelIntermediate", "ModelOutput", "Module", "None", "QuantAct", "__init__", "act_bit", "attention", "attention_mask", "attention_output", "attention_output_scaling_factor", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "hidden_states_scaling_factor", "intermediate", "intermediate_output", "intermediate_output_scaling_factor", "layer_output", "layer_output_scaling_factor", "nn", "output", "output_attentions", "outputs", "pre_intermediate_act", "pre_output_act", "quant_mode", "return", "self", "self_attention_outputs", "self_attention_outputs_scaling_factor", "seq_len_dim", "super"], "ibert/modeling_ibert.py:IBertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "False", "Model", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "class", "config", "cross_attentions", "def", "else", "enumerate", "for", "forward", "head_mask", "hidden_states", "hidden_states_scaling_factor", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "quant_mode", "range", "return", "return_dict", "self", "super", "tuple", "v"], "ibert/modeling_ibert.py:IBertPooler": ["Linear", "Model", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "quant_mode", "return", "self", "super"], "ibert/modeling_ibert.py:IBertPreTrainedModel": ["BERT", "Embedding", "I", "IntLayerNorm", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMHead", "None", "NotImplementedError", "PreTrainedModel", "QuantEmbedding", "QuantLinear", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "for", "if", "initializer_range", "is", "isinstance", "mean", "module", "new_num_tokens", "nn", "normal_", "not", "padding_idx", "raise", "resize_token_embeddings", "self", "std", "supported", "weight", "zero_"], "ibert/modeling_ibert.py:IBertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "FloatTensor", "LongTensor", "Model", "ModelEmbeddings", "ModelEncoder", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embedding_output_scaling_factor", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "quant_mode", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "ibert/modeling_ibert.py:IBertForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_head", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "ibert/modeling_ibert.py:IBertLMHead": ["LayerNorm", "Linear", "Model", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "ibert/modeling_ibert.py:IBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "ibert/modeling_ibert.py:IBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "ibert/modeling_ibert.py:IBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "ibert/modeling_ibert.py:IBertClassificationHead": ["Dropout", "Linear", "Model", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch"], "ibert/modeling_ibert.py:IBertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "ibert/modeling_ibert.py:create_position_ids_from_input_ids": ["Model_position_ids_from_input_ids", "cumsum", "def", "dim", "incremental_indices", "input_ids", "int", "long", "mask", "ne", "padding_idx", "past_key_values_length", "return", "torch", "type_as"], "depth_pro/modeling_depth_pro.py:DepthProOutput": ["FloatTensor", "ModelOutput", "None", "Optional", "Union", "attentions", "class", "features", "hidden_states", "last_hidden_state", "list", "r", "torch", "tuple"], "depth_pro/modeling_depth_pro.py:DepthProDepthEstimatorOutput": ["FloatTensor", "ModelModelEstimatorOutput", "ModelOutput", "None", "Optional", "attentions", "class", "field_of_view", "hidden_states", "loss", "predicted_Model", "r", "torch", "tuple"], "depth_pro/modeling_depth_pro.py:split_to_patches": ["F", "Model_to_patches", "Tensor", "batch_size", "def", "float", "height", "if", "int", "kernel_size", "num_channels", "overlap_ratio", "patch_size", "patches", "permute", "pixel_values", "reshape", "return", "shape", "stride", "torch", "torch_int", "unfold", "width"], "depth_pro/modeling_depth_pro.py:reshape_features": ["Model", "Model_features", "Tensor", "def", "hidden_size", "hidden_states", "n_samples", "permute", "return", "seq_len", "shape", "size", "torch", "torch_int"], "depth_pro/modeling_depth_pro.py:merge_patches": ["Model_patches", "Modeld", "Tensor", "_", "append", "batch_size", "box", "box_h", "box_w", "boxes", "boxes_in_row", "cat", "def", "dim", "else", "for", "h", "hidden_size", "i", "if", "in", "int", "min", "n_patches", "n_patches_per_batch", "new_out_size", "out_size", "pad_bottom", "pad_left", "pad_right", "pad_top", "padding", "paddings", "patches", "permute", "range", "reshape", "return", "shape", "sqrt_n_patches_per_batch", "torch", "torch_int", "w"], "depth_pro/modeling_depth_pro.py:reconstruct_feature_maps": ["F", "False", "Model_feature_maps", "Tensor", "align_corners", "batch_size", "bilinear", "def", "features", "float", "hidden_state", "int", "interpolate", "merge_patches", "mode", "output_size", "padding", "reshape_features", "return", "size", "torch", "tuple"], "depth_pro/modeling_depth_pro.py:DepthProPatchEncoder": ["AutoModel", "F", "False", "Image", "ModelConfig", "ModelPatchEncoder", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "align_corners", "append", "base_height", "base_width", "batch_size", "be", "bilinear", "cat", "class", "config", "def", "dim", "encodings", "exponent_value", "f", "features", "for", "forward", "from_config", "head_mask", "height", "hidden_state", "i", "if", "image_model_config", "image_size", "in", "intermediate_feature_dims", "intermediate_features", "intermediate_hook_ids", "interpolate", "is", "len", "list", "log2", "math", "merge_padding_value", "min", "mode", "model", "n_intermediate_hooks", "n_patches_per_scaled_image", "n_scaled_images", "nn", "num_channels", "out_size", "output_height", "output_hidden_states", "output_size", "output_width", "overlap_ratio", "padding", "patch_model_config", "patch_size", "patches", "pixel_values", "raise", "range", "ratio", "reconstruct_feature_maps", "return", "scale_factor", "scaled", "scaled_images", "scaled_images_feature_dims", "scaled_images_features", "scaled_images_last_hidden_state", "scaled_images_overlap_ratios", "scaled_images_ratios", "self", "shape", "size", "small", "split_to_patches", "split_with_sizes", "super", "to", "too", "torch", "torch_int", "when", "width", "with", "x"], "depth_pro/modeling_depth_pro.py:DepthProImageEncoder": ["AutoModel", "F", "False", "ModelConfig", "ModelImageEncoder", "ModelOutput", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "align_corners", "attentions", "base_height", "base_width", "batch_size", "bilinear", "bool", "class", "config", "def", "encodings", "exponent_value", "features", "forward", "from_config", "head_mask", "height", "hidden_states", "if", "image_model_config", "image_size", "interpolate", "last_hidden_state", "log2", "math", "mode", "model", "nn", "not", "num_channels", "out_size", "output_attentions", "output_hidden_states", "output_size", "padding", "patch_size", "pixel_values", "reconstruct_feature_maps", "return", "return_dict", "self", "shape", "size", "super", "torch", "torch_int", "tuple", "width"], "depth_pro/modeling_depth_pro.py:DepthProEncoder": ["False", "ModelConfig", "ModelEncoder", "ModelImageEncoder", "ModelOutput", "ModelPatchEncoder", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "attentions", "batch_size", "bool", "class", "config", "def", "features", "forward", "head_mask", "height", "hidden_states", "if", "image_encoder", "image_encodings", "image_features", "intermediate_feature_dims", "intermediate_hook_ids", "last_hidden_state", "len", "merge_padding_value", "n_intermediate_hooks", "n_scaled_images", "nn", "not", "num_channels", "output_attentions", "output_hidden_states", "patch_encoder", "patch_features", "pixel_values", "return", "return_dict", "scaled_images_feature_dims", "scaled_images_overlap_ratios", "scaled_images_ratios", "self", "shape", "super", "torch", "tuple", "width"], "depth_pro/modeling_depth_pro.py:DepthProFeatureUpsampleBlock": ["Conv2d", "ConvTranspose2d", "False", "ModelConfig", "ModelFeatureUpsampleBlock", "Module", "ModuleList", "Tensor", "True", "__init__", "append", "bias", "bool", "class", "config", "def", "else", "features", "for", "forward", "i", "if", "in", "in_channels", "input_dims", "int", "intermediate_dims", "kernel_size", "layer", "layers", "n_upsample_layers", "nn", "out_channels", "output_dims", "padding", "proj", "range", "return", "self", "stride", "super", "torch", "use_proj"], "depth_pro/modeling_depth_pro.py:DepthProFeatureUpsample": ["False", "ModelConfig", "ModelFeatureUpsample", "ModelFeatureUpsampleBlock", "Module", "ModuleList", "Tensor", "True", "__init__", "append", "bias", "block", "class", "config", "def", "else", "enumerate", "feature_dims", "features", "for", "forward", "fusion_hidden_size", "hidden_size", "i", "if", "image_block", "image_model_config", "in", "input_dims", "intermediate", "intermediate_dims", "intermediate_feature_dims", "intermediate_hook_ids", "len", "list", "n_intermediate_hooks", "n_scaled_images", "n_upsample_layers", "nn", "output_dims", "patch_model_config", "range", "return", "scaled_images", "scaled_images_feature_dims", "scaled_images_ratios", "self", "super", "torch", "use_proj"], "depth_pro/modeling_depth_pro.py:DepthProFeatureProjection": ["Conv2d", "False", "Identity", "ModelConfig", "ModelFeatureProjection", "Module", "ModuleList", "Tensor", "__init__", "and", "append", "bias", "class", "combined_feature_dims", "config", "def", "else", "enumerate", "features", "for", "forward", "fusion_hidden_size", "i", "if", "in", "in_channels", "intermediate_feature_dims", "kernel_size", "len", "list", "nn", "out_channels", "padding", "projected_features", "projection", "projections", "return", "scaled_images_feature_dims", "self", "stride", "super", "torch", "upsampled_feature"], "depth_pro/modeling_depth_pro.py:DepthProNeck": ["Conv2d", "ModelConfig", "ModelFeatureProjection", "ModelFeatureUpsample", "ModelNeck", "Module", "Tensor", "True", "__init__", "bias", "cat", "class", "config", "def", "dim", "feature_projection", "feature_upsample", "features", "forward", "fuse_image_with_low_res", "global_features", "in_channels", "kernel_size", "list", "nn", "out_channels", "padding", "return", "scaled_images_feature_dims", "self", "stride", "super", "torch"], "depth_pro/modeling_depth_pro.py:DepthProPreTrainedModel": ["Conv2d", "ConvTranspose2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreActResidualLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fan_out", "fill_", "fov_model", "if", "init", "initializer_range", "is", "isinstance", "kaiming_normal_", "main_input_name", "mean", "mode", "module", "nn", "nonlinearity", "normal_", "not", "pixel_values", "relu", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "depth_pro/modeling_depth_pro.py:DepthProModel": ["FloatTensor", "ModelEncoder", "ModelModel", "ModelNeck", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "encodings", "features", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "image_encoder", "is", "last_hidden_state", "model", "neck", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "depth_pro/modeling_depth_pro.py:DepthProPreActResidualLayer": ["BatchNorm2d", "Conv2d", "ModelConfig", "ModelPreActResidualLayer", "Module", "None", "ReLU", "Tensor", "__init__", "activation1", "activation2", "batch_norm1", "batch_norm2", "bias", "class", "config", "convolution1", "convolution2", "def", "else", "forward", "fusion_hidden_size", "hidden_state", "if", "is", "kernel_size", "nn", "not", "padding", "residual", "return", "self", "stride", "super", "torch", "use_batch_norm", "use_batch_norm_in_fusion_residual", "use_bias_in_fusion_residual"], "depth_pro/modeling_depth_pro.py:DepthProFeatureFusionLayer": ["Conv2d", "ConvTranspose2d", "False", "ModelConfig", "ModelFeatureFusionLayer", "ModelPreActResidualLayer", "Module", "None", "Optional", "Tensor", "True", "__init__", "bias", "bool", "class", "config", "deconv", "def", "forward", "fusion_hidden_size", "hidden_state", "if", "in_channels", "is", "kernel_size", "nn", "not", "out_channels", "padding", "projection", "residual", "residual_layer1", "residual_layer2", "return", "self", "stride", "super", "torch", "use_deconv"], "depth_pro/modeling_depth_pro.py:DepthProFeatureFusionStage": ["False", "ModelFeatureFusionLayer", "ModelFeatureFusionStage", "Module", "ModuleList", "None", "Tensor", "ValueError", "_", "__init__", "append", "class", "config", "def", "does", "else", "f", "final", "for", "forward", "fused_hidden_state", "fused_hidden_states", "hidden_state", "hidden_states", "if", "in", "intermediate", "intermediate_hook_ids", "is", "layer", "len", "list", "match", "nn", "not", "num_layers", "raise", "range", "return", "scaled_images_ratios", "self", "super", "torch", "use_deconv", "zip"], "depth_pro/modeling_depth_pro.py:DepthProFovEncoder": ["AutoModel", "F", "False", "Linear", "ModelConfig", "ModelFovEncoder", "Module", "None", "Optional", "Tensor", "__init__", "align_corners", "base_height", "base_width", "batch_size", "bilinear", "class", "config", "def", "encodings", "exponent_value", "features", "forward", "fov_model_config", "from_config", "fusion_hidden_size", "head_mask", "height", "hidden_size", "hidden_state", "image_model_config", "image_size", "interpolate", "log2", "math", "mode", "model", "neck", "nn", "num_channels", "out_size", "output_size", "padding", "patch_size", "pixel_values", "reconstruct_feature_maps", "return", "self", "shape", "size", "super", "torch", "torch_int", "width"], "depth_pro/modeling_depth_pro.py:DepthProFovHead": ["Conv2d", "F", "False", "ModelConfig", "ModelFovHead", "Module", "ModuleList", "ReLU", "Tensor", "True", "__init__", "align_corners", "append", "bilinear", "ceil", "class", "config", "def", "features", "final_in_channels", "final_kernel_size", "for", "forward", "fusion_hidden_size", "i", "image_model_config", "image_size", "in", "in_channels", "interpolate", "kernel_size", "layer", "layers", "math", "mode", "nn", "num_fov_head_layers", "out_channels", "out_size", "padding", "patch_size", "range", "return", "self", "size", "stride", "super", "torch", "torch_int"], "depth_pro/modeling_depth_pro.py:DepthProFovModel": ["Conv2d", "ModelConfig", "ModelFovEncoder", "ModelFovHead", "ModelFovModel", "Module", "None", "Optional", "ReLU", "Tensor", "True", "__init__", "activation", "class", "config", "conv", "def", "flatten", "forward", "fov_encoder", "fov_features", "fov_output", "fusion_hidden_size", "global_features", "head", "head_mask", "inplace", "kernel_size", "nn", "padding", "pixel_values", "return", "self", "stride", "super", "torch"], "depth_pro/modeling_depth_pro.py:DepthProDepthEstimationHead": ["Conv2d", "ConvTranspose2d", "ModelModelEstimationHead", "Module", "ModuleList", "ReLU", "Tensor", "True", "__init__", "bias", "class", "config", "def", "dim", "features", "for", "forward", "fusion_hidden_size", "hidden_states", "in", "in_channels", "kernel_size", "layer", "layers", "nn", "out_channels", "padding", "predicted_Model", "return", "self", "squeeze", "stride", "super", "torch"], "depth_pro/modeling_depth_pro.py:DepthProForDepthEstimation": ["FloatTensor", "LongTensor", "Model", "ModelFeatureFusionStage", "ModelForModelEstimation", "ModelFovModel", "ModelModel", "ModelModelEstimationHead", "ModelModelEstimatorOutput", "ModelPreTrainedModel", "Model_outputs", "None", "NotImplementedError", "Optional", "Tensor", "Training", "True", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "detach", "else", "features", "features_for_fov", "field_of_view", "for", "forward", "fov", "fov_model", "fused_hidden_states", "fusion_stage", "global_features", "head", "head_mask", "hidden_states", "if", "implemented", "in", "is", "labels", "loss", "not", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "predicted_Model", "r", "raise", "return", "return_dict", "self", "super", "torch", "tuple", "use_fov_model", "use_return_dict", "v", "yet"], "vitdet/modeling_vitdet.py:VitDetEmbeddings": ["Absolute", "Conv2d", "Expected", "False", "Iterable", "Make", "ModelEmbeddings", "Module", "None", "Parameter", "Tensor", "True", "ValueError", "__init__", "a", "abc", "abs_pos_embeddings", "align_corners", "be", "bicubic", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "f", "forward", "functional", "get_absolute_positions", "got", "has_cls_token", "height", "hidden_size", "if", "image_size", "in", "int", "interpolate", "is", "is_tracing", "isinstance", "jit", "kernel_size", "match", "math", "mode", "must", "new_abs_pos_embeddings", "nn", "not", "num_channels", "num_patches", "num_position", "num_positions", "number", "of", "one", "or", "patch_size", "permute", "pixel", "pixel_values", "position", "position_embeddings", "pretrain_image_size", "projection", "raise", "reshape", "return", "self", "set", "shape", "size", "sqrt", "square", "stride", "super", "sure", "that", "the", "torch", "use_absolute_position_embeddings", "values", "width", "with", "zeros"], "vitdet/modeling_vitdet.py:get_rel_pos": ["Model_rel_pos", "None", "arange", "def", "else", "functional", "if", "int", "interpolate", "k_coords", "k_size", "linear", "long", "max", "max_rel_dist", "mode", "nn", "permute", "q_coords", "q_size", "rel_pos", "rel_pos_resized", "relative_coords", "reshape", "return", "shape", "size", "torch"], "vitdet/modeling_vitdet.py:add_decomposed_relative_positions": ["Model_decomposed_relative_positions", "None", "_", "attn", "batch_size", "bhwc", "bhwk", "def", "dim", "einsum", "get_rel_pos", "hkc", "k_size", "keys_height", "keys_width", "q_size", "queries", "queries_height", "queries_width", "r_q", "rel_pos_h", "rel_pos_w", "relative_height", "relative_weight", "relative_width", "reshape", "return", "shape", "torch", "view", "wkc"], "vitdet/modeling_vitdet.py:VitDetAttention": ["False", "Linear", "ModelAttention", "Module", "None", "Parameter", "_", "__init__", "add_decomposed_relative_positions", "attention_probs", "attention_scores", "batch_size", "bias", "class", "config", "def", "dim", "else", "forward", "head_dim", "height", "hidden_size", "hidden_state", "if", "input_size", "keys", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "permute", "proj", "qkv", "qkv_bias", "queries", "rel_pos_h", "rel_pos_w", "reshape", "return", "scale", "self", "shape", "softmax", "super", "torch", "transpose", "unbind", "use_relative_position_embeddings", "values", "view", "width", "zeros"], "vitdet/modeling_vitdet.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "vitdet/modeling_vitdet.py:VitDetDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "vitdet/modeling_vitdet.py:VitDetLayerNorm": ["ModelLayerNorm", "Module", "None", "Parameter", "True", "__init__", "bias", "class", "def", "eps", "forward", "keepdim", "mean", "nn", "normalized_shape", "ones", "pow", "return", "s", "self", "sqrt", "super", "torch", "u", "weight", "x", "zeros"], "vitdet/modeling_vitdet.py:VitDetResBottleneckBlock": ["ACT2FN", "Conv2d", "False", "ModelLayerNorm", "ModelResBottleneckBlock", "Module", "__init__", "act1", "act2", "bias", "bottleneck_channels", "children", "class", "config", "conv1", "conv2", "conv3", "def", "for", "forward", "hidden_act", "in", "in_channels", "layer", "nn", "norm1", "norm2", "norm3", "out", "out_channels", "padding", "return", "self", "super", "x"], "vitdet/modeling_vitdet.py:VitDetMlp": ["ACT2FN", "Dropout", "Linear", "ModelMlp", "Module", "None", "Tensor", "__init__", "act", "class", "config", "def", "drop", "dropout_prob", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "in_features", "int", "nn", "return", "self", "super", "torch", "x"], "vitdet/modeling_vitdet.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "functional", "height", "hidden_state", "nn", "num_channels", "pad", "pad_height", "pad_width", "padded_height", "padded_width", "permute", "return", "shape", "view", "width"], "vitdet/modeling_vitdet.py:window_unpartition": ["Model_size", "Model_unpartition", "Models", "batch_size", "contiguous", "def", "height", "height_width", "hidden_state", "pad_height_width", "padded_height", "padded_width", "permute", "return", "shape", "view", "width"], "vitdet/modeling_vitdet.py:VitDetLayer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelMlp", "ModelResBottleneckBlock", "None", "Optional", "Tensor", "Union", "__init__", "attention", "bool", "bottleneck_channels", "class", "config", "def", "dim", "drop_path", "drop_path_rate", "else", "eps", "float", "forward", "head_mask", "height", "hidden_features", "hidden_size", "hidden_states", "if", "image_size", "in_channels", "in_features", "input_size", "int", "isinstance", "layer_norm_eps", "list", "mlp", "mlp_ratio", "nn", "norm1", "norm2", "out_channels", "output_attentions", "outputs", "pad_height_width", "patch_size", "permute", "residual", "return", "self", "self_attention_outputs", "shape", "shortcut", "super", "torch", "tuple", "use_residual_block", "width", "window_partition", "window_size", "window_unpartition"], "vitdet/modeling_vitdet.py:VitDetEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "append", "attentions", "bool", "class", "config", "cpu", "def", "depth", "device", "drop_path_rate", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "item", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "layers", "linspace", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "residual_block_indices", "return", "return_dict", "self", "super", "torch", "tuple", "use_residual_block", "v", "window_block_indices", "window_size", "x"], "vitdet/modeling_vitdet.py:caffe2_msra_fill": ["Model_msra_fill", "Module", "None", "bias", "constant_", "def", "fan_out", "if", "init", "is", "kaiming_normal_", "mode", "module", "nn", "nonlinearity", "not", "relu", "weight"], "vitdet/modeling_vitdet.py:VitDetPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "ModelResBottleneckBlock", "None", "PreTrainedModel", "True", "Union", "_init_weights", "_no_split_modules", "and", "base_model_prefix", "bias", "caffe2_msra_fill", "class", "config", "conv1", "conv2", "conv3", "data", "def", "dtype", "elif", "fill_", "float32", "for", "if", "in", "init", "initializer_range", "is", "isinstance", "layer", "main_input_name", "mean", "module", "nn", "norm1", "norm2", "norm3", "not", "pixel_values", "position_embeddings", "rel_pos_h", "rel_pos_w", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "use_relative_position_embeddings", "weight", "zero_"], "vitdet/modeling_vitdet.py:VitDetModel": ["BaseModelOutput", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_prune_heads", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "dict", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "int", "is", "items", "last_hidden_state", "layer", "list", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "projection", "prune_heads", "r", "raise", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "vitdet/modeling_vitdet.py:VitDetBackbone": ["BackboneMixin", "BackboneOutput", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "_", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "get_input_embeddings", "hidden_size", "hidden_state", "hidden_states", "if", "in", "is", "not", "num_features", "num_hidden_layers", "out_features", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "projection", "r", "range", "return", "return_dict", "self", "stage", "stage_names", "super", "torch", "use_return_dict", "zip"], "textnet/modeling_textnet.py:TextNetConvLayer": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "Identity", "ModelConfig", "ModelConvLayer", "Module", "None", "Tensor", "__init__", "activation", "activation_function", "batch_norm", "batch_norm_eps", "bias", "class", "config", "conv", "def", "else", "forward", "hidden_states", "if", "is", "isinstance", "kernel_size", "nn", "not", "padding", "return", "self", "stem_act_func", "stem_kernel_size", "stem_num_channels", "stem_out_channels", "stem_stride", "stride", "super", "torch", "tuple"], "textnet/modeling_textnet.py:TextNetRepConvLayer": ["BatchNorm2d", "Conv2d", "False", "ModelConfig", "ModelRepConvLayer", "Module", "None", "ReLU", "Tensor", "__init__", "activation_function", "and", "batch_norm_eps", "bias", "class", "config", "def", "else", "eps", "forward", "hidden_states", "horizontal_batch_norm", "horizontal_conv", "horizontal_outputs", "horizontal_padding", "id_out", "if", "in_channels", "int", "is", "kernel_size", "main_batch_norm", "main_conv", "main_outputs", "nn", "not", "num_channels", "num_features", "out_channels", "padding", "r", "rbr_identity", "return", "self", "stride", "super", "torch", "vertical_batch_norm", "vertical_conv", "vertical_outputs", "vertical_padding"], "textnet/modeling_textnet.py:TextNetStage": ["ModelConfig", "ModelRepConvLayer", "ModelStage", "Module", "ModuleList", "__init__", "append", "block", "class", "config", "conv_layer_kernel_sizes", "conv_layer_strides", "def", "depth", "for", "forward", "hidden_sizes", "hidden_state", "in", "in_channels", "int", "kernel_size", "len", "nn", "num_layers", "out_channels", "return", "self", "stage", "stage_config", "stage_in_channel_size", "stage_out_channel_size", "stride", "super", "zip"], "textnet/modeling_textnet.py:TextNetEncoder": ["BaseModelOutputWithNoAttention", "ModelConfig", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "append", "bool", "class", "config", "conv_layer_kernel_sizes", "def", "else", "for", "forward", "hidden_state", "hidden_states", "if", "in", "last_hidden_state", "len", "nn", "not", "num_stages", "output", "output_hidden_states", "range", "return", "return_dict", "self", "stage", "stage_ix", "stages", "super", "torch"], "textnet/modeling_textnet.py:TextNetPreTrainedModel": ["BatchNorm2d", "Conv2d", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "textnet/modeling_textnet.py:TextNetModel": ["AdaptiveAvgPool2d", "Any", "BaseModelOutputWithPoolingAndNoAttention", "ModelConvLayer", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "encoder_outputs", "forward", "hidden_state", "hidden_states", "if", "is", "last_hidden_state", "list", "nn", "not", "output", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "return", "return_dict", "self", "stem", "super", "tuple", "use_return_dict"], "textnet/modeling_textnet.py:TextNetForImageClassification": ["AdaptiveAvgPool2d", "Flatten", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "__init__", "auto_docstring", "avg_pool", "bool", "class", "classifier", "config", "def", "else", "fc", "flatten", "for", "forward", "hidden_sizes", "hidden_states", "if", "in", "is", "labels", "last_hidden_state", "layer", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "use_return_dict"], "textnet/modeling_textnet.py:TextNetBackbone": ["BackboneMixin", "BackboneOutput", "False", "Model", "ModelBackbone", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "enumerate", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_features", "out_features", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "stage", "stage_names", "super", "tuple", "use_return_dict"], "gptj/modeling_gptj.py:create_sinusoidal_positions": ["Model_sinusoidal_positions", "Tensor", "arange", "cat", "cos", "def", "dim", "dtype", "einsum", "float", "i", "int", "int64", "inv_freq", "j", "num_pos", "return", "sin", "sinusoid_inp", "torch"], "gptj/modeling_gptj.py:get_embed_positions": ["Model_embed_positions", "def", "device", "embed_positions", "position_ids", "repeat", "return", "shape", "to"], "gptj/modeling_gptj.py:rotate_every_two": ["Model_every_two", "Tensor", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "gptj/modeling_gptj.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "Tensor", "cos", "def", "repeat_interleave", "return", "rotate_every_two", "sin", "tensor", "torch"], "gptj/modeling_gptj.py:GPTJAttention": ["Cache", "Dropout", "False", "FloatTensor", "Input", "Instantiating", "Linear", "LongTensor", "Model", "Module", "None", "Optional", "Please", "Tensor", "True", "Union", "ValueError", "__class__", "__init__", "__name__", "_attn", "_get_embed_positions", "_merge_heads", "_split_heads", "a", "and", "apply_rotary_pos_emb", "attention_mask", "attn_dropout", "attn_head_size", "attn_output", "attn_pdrop", "attn_weights", "be", "bias", "bool", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "cat", "causal_mask", "class", "config", "contiguous", "cos", "create_sinusoidal_positions", "creating", "def", "device", "dim", "divisible", "dtype", "during", "elif", "else", "embed_dim", "embed_positions", "errors", "f", "float32", "forward", "functional", "gather", "get_default_dtype", "get_embed_positions", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "is", "is_causal", "is_torch_fx_proxy", "is_tracing", "jit", "k_pass", "k_proj", "k_rot", "key", "layer_idx", "layer_past", "lead", "len", "logger", "make", "matmul", "max_position_embeddings", "max_positions", "must", "new_shape", "nn", "not", "num_attention_heads", "of", "one", "or", "out_proj", "output_attentions", "partial_rotation_size", "passing", "permute", "pos_embd_dim", "position_ids", "provide", "q_pass", "q_proj", "q_rot", "query", "raise", "rank", "recommended", "repeat", "repeated_position_ids", "resid_dropout", "resid_pdrop", "return", "rotary", "rotary_dim", "scale_attn", "self", "shape", "should", "sin", "sincos", "size", "softmax", "split", "sqrt", "super", "sure", "tensor", "the", "this", "to", "torch", "transpose", "tuple", "unsqueeze", "update", "use_cache", "used", "v_proj", "value", "view", "warning_once", "when", "will", "without"], "gptj/modeling_gptj.py:GPTJFlashAttention2": ["Cache", "False", "FloatTensor", "LongTensor", "Model", "ModelAttention", "None", "Optional", "Tensor", "The", "True", "Union", "We", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_get_embed_positions", "_pre_quantization_dtype", "_split_heads", "apply_rotary_pos_emb", "args", "attention_dropout", "attention_mask", "attn_output", "attn_pdrop", "attn_weights", "back", "be", "bool", "cache_kwargs", "cache_position", "cast", "casted", "cat", "class", "config", "contiguous", "cos", "cpu", "def", "device", "device_type", "dim", "dropout", "dtype", "elif", "else", "embed_positions", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "gather", "get_autocast_dtype", "get_autocast_gpu_dtype", "get_embed_positions", "hasattr", "have", "head_dim", "head_mask", "hidden", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "is_torch_fx_proxy", "is_tracing", "jit", "k_pass", "k_proj", "k_rot", "key", "kwargs", "layer", "layer_idx", "layer_past", "layers", "logger", "might", "mps", "norm", "not", "num_attention_heads", "or", "out_proj", "output_attentions", "partial_rotation_size", "permute", "position_ids", "q_pass", "q_proj", "q_rot", "query", "query_length", "related", "repeat", "repeated_position_ids", "reshape", "resid_dropout", "return", "rotary_dim", "seems", "self", "shape", "silently", "sin", "sincos", "split", "states", "super", "target_dtype", "the", "this", "to", "torch", "training", "tuple", "type", "unsqueeze", "upcasted", "update", "use_cache", "use_top_left_mask", "v_proj", "value", "warning_once", "weight", "will", "you"], "gptj/modeling_gptj.py:GPTJMLP": ["ACT2FN", "Dropout", "FloatTensor", "Linear", "Model", "Module", "Optional", "__init__", "act", "activation_function", "class", "config", "def", "dropout", "embed_dim", "fc_in", "fc_out", "forward", "hidden_states", "intermediate_size", "n_embd", "nn", "resid_pdrop", "return", "self", "super", "torch"], "gptj/modeling_gptj.py:GPTJBlock": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "Model", "ModelMLP", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "Union", "__init__", "_attn_implementation", "attention_mask", "attn", "attn_outputs", "attn_weights", "bool", "cache_position", "class", "config", "def", "else", "eps", "feed_forward_hidden_states", "forward", "head_mask", "hidden_states", "if", "inner_dim", "is", "layer_idx", "layer_norm_epsilon", "layer_past", "ln_1", "mlp", "n_embd", "n_inner", "nn", "not", "output_attentions", "position_ids", "residual", "return", "self", "super", "torch", "tuple", "use_cache"], "gptj/modeling_gptj.py:GPTJPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "Model", "ModelBlock", "ModelConfig", "None", "PreTrainedModel", "True", "__init__", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_param_buffer_assignment", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "inputs", "is", "is_parallelizable", "isinstance", "kwargs", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "gptj/modeling_gptj.py:GPTJModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DEPARALLELIZE_DOCSTRING", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "FutureWarning", "LayerNorm", "Like", "LongTensor", "Model", "ModelBlock", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "PARALLELIZE_DOCSTRING", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "a", "add_start_docstrings", "all_hidden_states", "all_self_attentions", "also", "and", "any", "arange", "assert_device_map", "attention_mask", "attentions", "auto_docstring", "balanced", "batch_size", "be", "block", "bool", "but", "cache_position", "call", "can", "causal_mask", "checkpointing", "class", "clone", "config", "cpu", "cuda", "cuda_device", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "diagonal", "dictionary", "dim", "drop", "dtype", "else", "embd_pdrop", "embed_dim", "empty_cache", "enumerate", "eps", "exactly", "expand", "fill_value", "finfo", "first_device", "flash_attention_2", "flex_attention", "for", "forward", "from_pretrained", "full", "get_device_map", "get_head_mask", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "h", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "index", "input_ids", "input_tensor", "inputs_embeds", "instance", "int", "is", "is_compileable", "is_training", "isinstance", "it", "items", "k", "keys", "kwargs", "last_device", "last_hidden_state", "layer", "layer_idx", "layer_norm_epsilon", "layer_past", "layers", "len", "ln_f", "load", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "max", "min", "min_dtype", "model", "model_parallel", "module_name", "must", "n_embd", "n_layer", "needs", "new_embeddings", "nn", "not", "npu", "of", "one", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "own", "padding_mask", "parallelize", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_ids", "post_init", "provide", "r", "raise", "range", "removed", "reshape", "return", "return_dict", "sdpa", "self", "seq_length", "sequence_length", "set_device", "set_input_embeddings", "shape", "should", "size", "so", "specify", "staticmethod", "str", "super", "target_length", "the", "to", "token_type_embeds", "token_type_ids", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "v5", "values", "view", "vocab_size", "warn", "warning_once", "warnings", "will", "with", "wte", "xpu", "you", "your"], "gptj/modeling_gptj.py:GPTJForCausalLM": ["Cache", "CausalLMOutputWithPast", "DEPARALLELIZE_DOCSTRING", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Like", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PARALLELIZE_DOCSTRING", "Tensor", "Transformers", "True", "Union", "You", "__init__", "_tied_weights_keys", "a", "add_start_docstrings", "also", "and", "assert_device_map", "attention_mask", "attentions", "auto_docstring", "balanced", "be", "bool", "but", "cache_position", "call", "can", "class", "config", "cpu", "cuda", "def", "deparallelize", "deprecated", "device", "device_count", "device_map", "dictionary", "dtype", "else", "empty_cache", "first_device", "float32", "for", "forward", "from_pretrained", "get_device_map", "h", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instance", "is", "it", "kwargs", "labels", "len", "lm_head", "lm_logits", "load", "logits", "loss", "loss_function", "model", "model_parallel", "module_name", "n_embd", "needs", "nn", "not", "of", "output", "output_attentions", "output_hidden_states", "own", "parallelize", "past_key_values", "position_ids", "post_init", "provide", "r", "range", "removed", "return", "return_dict", "self", "set_device", "should", "so", "super", "the", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "v5", "vocab_size", "warn", "warnings", "weight", "will", "with", "you", "your"], "gptj/modeling_gptj.py:GPTJForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "device_map", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "model_parallel", "multi_label_classification", "n_embd", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "gptj/modeling_gptj.py:GPTJForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "device", "device_map", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "model_parallel", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "to", "token_type_ids", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "xcodec/modeling_xcodec.py:XcodecOutput": ["FloatTensor", "LongTensor", "ModelOutput", "None", "Optional", "audio_codes", "audio_values", "class", "torch"], "xcodec/modeling_xcodec.py:XcodecEncoderOutput": ["LongTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "audio_codes", "class", "torch"], "xcodec/modeling_xcodec.py:XcodecDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "audio_values", "class", "torch"], "xcodec/modeling_xcodec.py:ResidualUnit": ["Conv1d", "ELU", "False", "ModelConfig", "ModelUnit", "Module", "Tensor", "__init__", "activation", "bias", "class", "config", "conv1", "conv2", "def", "dilation", "forward", "groups", "hidden_state", "in_channels", "int", "kernel_size", "nn", "out_channels", "output_tensor", "padding", "return", "self", "stride", "super", "torch", "unit_kernel_size"], "xcodec/modeling_xcodec.py:SemanticEncoderBlock": ["Conv1d", "ModelConfig", "ModelEncoderBlock", "Module", "ModuleList", "ResidualUnit", "Tensor", "True", "__init__", "bias", "block_dilations", "class", "config", "conv", "def", "dilation", "else", "for", "forward", "hidden_state", "if", "in", "in_channels", "int", "kernel", "kernel_size", "nn", "out_channels", "padding", "res_units", "return", "self", "stride", "super", "torch", "unit"], "xcodec/modeling_xcodec.py:SemanticEncoder": ["Conv1d", "False", "ModelEncoder", "ModelEncoderBlock", "Model_hidden_size", "Module", "ModuleList", "Number", "Tensor", "ValueError", "__init__", "bias", "block", "channel_ratios", "class", "config", "conv", "conv_blocks", "def", "enumerate", "for", "forward", "hidden_state", "i", "if", "in", "in_channels", "int", "kernel_size", "len", "match", "must", "nn", "number", "of", "out_channels", "raise", "return", "self", "stride", "strides", "super", "the", "torch"], "xcodec/modeling_xcodec.py:SemanticDecoderBlock": ["Conv1d", "ConvTranspose1d", "False", "ModelConfig", "ModelDecoderBlock", "Module", "ModuleList", "ResidualUnit", "Tensor", "True", "__init__", "bias", "block_dilations", "class", "config", "conv", "def", "dilation", "else", "for", "forward", "hidden_state", "if", "in", "in_channels", "int", "kernel_size", "nn", "out_channels", "output_padding", "padding", "res_units", "return", "self", "stride", "super", "torch", "unit"], "xcodec/modeling_xcodec.py:SemanticDecoder": ["Conv1d", "False", "ModelDecoder", "ModelDecoderBlock", "Model_hidden_size", "Module", "ModuleList", "Tensor", "__init__", "bias", "block", "channel_ratios", "class", "config", "conv1", "conv2", "conv_blocks", "def", "else", "enumerate", "for", "forward", "hidden_state", "i", "if", "in", "in_channels", "int", "kernel_size", "len", "nn", "out_channels", "padding", "return", "self", "stride", "strides", "super", "torch"], "xcodec/modeling_xcodec.py:XcodecEuclideanCodebook": ["F", "ModelEuclideanCodebook", "Module", "Tensor", "True", "__init__", "class", "clone", "cluster_size", "codebook_dim", "codebook_size", "config", "decode", "def", "dim", "dist", "embed", "embed_avg", "embed_ind", "embedding", "encode", "hidden_states", "indices", "inited", "keepdim", "max", "nn", "pow", "quantize", "quantized", "register_buffer", "reshape", "return", "scaled_states", "self", "shape", "sum", "super", "t", "torch", "view", "zeros"], "xcodec/modeling_xcodec.py:XcodecVectorQuantization": ["ModelConfig", "ModelEuclideanCodebook", "ModelVectorQuantization", "Module", "__init__", "class", "codebook", "config", "decode", "def", "embed_in", "embed_ind", "encode", "hidden_states", "nn", "permute", "quantize", "return", "self", "super"], "xcodec/modeling_xcodec.py:XcodecResidualVectorQuantization": ["ModelConfig", "ModelResidualVectorQuantization", "ModelVectorQuantization", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "all_indices", "and", "append", "bandwidth", "bw_per_q", "class", "codebook_size", "codes", "config", "decode", "def", "device", "embeddings", "encode", "enumerate", "floor", "for", "frame_rate", "get_bandwidth_per_quantizer", "get_num_quantizers_for_bandwidth", "i", "if", "in", "indices", "int", "is", "log2", "math", "max", "nn", "not", "num_quantizers", "out_indices", "quantized", "quantized_out", "quantizer", "quantizers", "range", "residual", "return", "self", "stack", "super", "tensor", "torch"], "xcodec/modeling_xcodec.py:XcodecPreTrainedModel": ["AttributeError", "Conv1d", "ConvTranspose1d", "Embedding", "GroupNorm", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedAudioTokenizerBase", "Snake1d", "True", "ValueError", "__class__", "__name__", "_init_weights", "a", "acoustic_decoder", "acoustic_encoder", "alpha", "and", "apply_weight_norm", "b", "base_model_prefix", "bias", "block", "class", "config", "config_class", "constant_", "conv1", "conv2", "conv_t1", "data", "def", "elif", "except", "fill_", "for", "groups", "hasattr", "if", "in", "in_channels", "init", "initializer_range", "input_values", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "leave_parametrized", "m", "main_input_name", "math", "mean", "module", "modules", "name", "nn", "normal_", "not", "parametrizations", "parametrize", "pass", "remove_parametrizations", "remove_weight_norm", "res_unit", "res_unit1", "res_unit2", "res_unit3", "reset_parameters", "self", "sqrt", "std", "submodule", "torch", "trunc_normal_", "try", "uniform_", "utils", "weight", "weight_norm", "zero_"], "xcodec/modeling_xcodec.py:XcodecModel": ["Audio", "AutoModel", "ConvTranspose1d", "F", "False", "FloatTensor", "Identity", "Linear", "ModelDecoderOutput", "ModelEncoderOutput", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelResidualVectorQuantization", "Module", "None", "Optional", "Select", "SemanticDecoder", "SemanticEncoder", "Tanh", "Tensor", "This", "True", "Union", "ValueError", "__init__", "_adjust_dac_decoder", "_extract_semantic_features", "acoustic_decoder", "acoustic_encoder", "acoustic_model", "acoustic_model_config", "and", "audio_codes", "audio_values", "auto_docstring", "bandwidth", "be", "bool", "but", "cat", "channels", "class", "config", "decode", "decoder", "decoder_semantic", "def", "detach", "dim", "doesn", "e_acoustic", "e_semantic", "e_semantic_input", "elif", "else", "embeddings", "encode", "encoder", "encoder_semantic", "eval", "f", "fc", "fc1", "fc2", "float", "for", "forward", "from_config", "got", "hasattr", "hidden_size", "hidden_states", "hop_length", "if", "in", "input_values", "is", "isinstance", "length", "mean", "model", "module", "modules", "mono", "must", "nn", "no_grad", "not", "of", "one", "output_hidden_states", "output_padding", "outputs", "pad", "post_init", "quantized", "quantized_acoustic", "quantizer", "r", "raise", "return", "return_dict", "self", "semantic_model", "semantic_model_config", "shape", "stack", "stacked", "staticmethod", "stride", "super", "support", "t", "tanh", "target_bandwidths", "the", "torch", "transpose", "tuple", "unsqueeze", "with"], "udop/modeling_udop.py:BaseModelOutputWithAttentionMask": ["Cache", "FloatTensor", "ModelModelOutputWithAttentionMask", "ModelOutput", "None", "Optional", "attention_mask", "attentions", "class", "cross_attentions", "hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "udop/modeling_udop.py:get_visual_bbox": ["Model_visual_bbox", "arange", "def", "dim", "image_feature_pool_shape", "image_size", "patch_size", "repeat", "return", "stack", "torch", "transpose", "view", "visual_bbox_input", "visual_bbox_x", "visual_bbox_y"], "udop/modeling_udop.py:pad_sequence": ["Model_sequence", "Model_value", "Tensor", "cat", "def", "dim", "else", "if", "isinstance", "len", "m", "n", "ret", "return", "seq", "shape", "stack", "target_len", "tensor", "to", "torch"], "udop/modeling_udop.py:combine_image_text_embeddings": ["False", "Model_image_text_embeddings", "None", "True", "arange", "attention_mask", "bbox", "bool", "cat", "clip", "cols", "def", "device", "dim", "else", "flatten", "float64", "floor", "for", "full_like", "gather", "get_visual_bbox", "i", "if", "image_embeddings", "image_size", "in", "ind", "input_vision_patches", "inputs_embeds", "inputs_vision_patches", "is", "item", "len", "long", "max_len", "mean", "not", "num_patches", "ocr_points", "ocr_points_x", "ocr_points_y", "pad_sequence", "patch_inds", "patch_size", "range", "repeat", "repeated_vision_embeds", "return", "rows", "sequence_length", "size", "stack", "target_seg", "tensor", "to", "torch", "unsqueeze", "visual_attention_mask", "visual_bbox", "zeros_like", "zip"], "udop/modeling_udop.py:UdopPatchEmbeddings": ["Conv2d", "Input", "Iterable", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "class", "collections", "config", "def", "doesn", "else", "embeddings", "f", "flatten", "forward", "height", "hidden_size", "if", "image", "image_size", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "or", "patch_size", "pixel_values", "proj", "raise", "return", "self", "shape", "size", "stride", "super", "t", "transpose", "width"], "udop/modeling_udop.py:UdopPreTrainedModel": ["Conv2d", "Embedding", "False", "In", "Model", "ModelAttention", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelForConditionalGeneration", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "RelativePositionBiasBase", "See", "True", "Verify", "_can_compile_fullgraph", "_init_weights", "_keep_in_fp32_modules", "_shift_right", "all", "and", "assert", "base_model_prefix", "be", "bias", "class", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_start_token_id", "def", "defined", "docs", "dtype", "elif", "factor", "fill_", "float32", "for", "has", "has_relative_attention_bias", "hasattr", "if", "information", "init", "initializer_factor", "input_ids", "is", "isinstance", "it", "item", "k", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "nn", "normal_", "not", "num_heads", "o", "only", "pad_token_id", "padding_idx", "positive", "q", "relative_attention_bias", "return", "self", "set", "shape", "shared", "shifted_input_ids", "std", "supports_gradient_checkpointing", "that", "the", "tie_word_embeddings", "to", "torch", "transformer", "trunc_normal_", "usually", "v", "values", "weight", "wi", "wi_0", "wi_1", "wo", "zero_"], "udop/modeling_udop.py:UdopLayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "udop/modeling_udop.py:UdopDenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "udop/modeling_udop.py:UdopDenseGatedActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseGatedActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "udop/modeling_udop.py:UdopLayerFF": ["DenseReluDense", "Dropout", "ModelConfig", "ModelDenseActDense", "ModelDenseGatedActDense", "ModelLayerFF", "ModelLayerNorm", "Module", "__init__", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is_gated_act", "layer_norm", "layer_norm_epsilon", "nn", "return", "self", "super"], "udop/modeling_udop.py:UdopAttention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "if", "index", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "len", "list", "log", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "union", "unsqueeze", "update", "use_cache", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "udop/modeling_udop.py:UdopLayerSelfAttention": ["Dropout", "False", "ModelAttention", "ModelLayerNorm", "ModelLayerSelfAttention", "Module", "None", "Optional", "SelfAttention", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "udop/modeling_udop.py:UdopLayerCrossAttention": ["Dropout", "EncDecAttention", "False", "ModelAttention", "ModelLayerCrossAttention", "ModelLayerNorm", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "udop/modeling_udop.py:UdopBlock": ["False", "GradientCheckpointingLayer", "ModelBlock", "ModelLayerCrossAttention", "ModelLayerFF", "ModelLayerSelfAttention", "ModuleList", "None", "Optional", "True", "__init__", "and", "any", "append", "attention_mask", "attention_outputs", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "deprecate_kwarg", "do_cross_attention", "dtype", "encoder_attention_mask", "encoder_decoder_position_bias", "encoder_hidden_states", "finfo", "float16", "forward", "has_relative_attention_bias", "hidden_states", "if", "int", "is", "is_decoder", "isinf", "key_value_states", "layer", "layer_head_mask", "layer_idx", "max", "min", "new_name", "nn", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "return_dict", "self", "self_attention_outputs", "super", "torch", "use_cache", "version", "where"], "udop/modeling_udop.py:UdopCellEmbeddings": ["Embedding", "ModelCellEmbeddings", "Module", "__init__", "bbox", "class", "clip", "def", "embeddings", "forward", "hidden_size", "left_position_embeddings", "long", "lower_position_embeddings", "max_2d_position_embeddings", "nn", "return", "right_position_embeddings", "self", "super", "torch", "upper_position_embeddings", "x_position_embeddings", "y_position_embeddings"], "udop/modeling_udop.py:RelativePositionBiasBase": ["ABC", "AUGMENTATION_RANGE", "Any", "Embedding", "False", "ModelPositionBiasBase", "Model_attention_bias", "Model_attention_num_buckets", "Model_position", "Module", "None", "Optional", "Tensor", "True", "ValueError", "Wrong", "__init__", "abstractmethod", "and", "attention_mask", "augmentation", "bbox", "bidirectional", "class", "context_position", "cpu", "data", "def", "dict", "dim", "dimension", "else", "enumerate", "expand", "extra_head", "for", "forward", "get_Model_position", "get_Model_position_bucket", "get_bucket", "idx", "if", "in", "is_prefix", "level", "long", "max_distance", "memory_position", "new_bias", "nn", "not", "num_buckets", "num_heads", "num_prefix", "num_prefix_row", "numpy", "of", "pass", "permute", "positions", "prefix_bucket", "prepare_input", "raise", "random", "repeat", "return", "rp_bucket", "scaling_factor", "self", "size", "str", "sum", "super", "tensor", "to", "tokens", "torch", "training", "uniform", "values", "weight"], "udop/modeling_udop.py:RelativePositionBias1D": ["Any", "ModelPositionBias1D", "ModelPositionBiasBase", "Model_position", "No", "None", "Optional", "Tensor", "ValueError", "__init__", "arange", "attention_mask", "bbox", "class", "def", "device", "dict", "dtype", "features", "get_Model_position", "if", "kwargs", "long", "max_distance", "need", "prepare_input", "raise", "return", "scale", "scaling_factor", "self", "size", "str", "super", "to", "torch"], "udop/modeling_udop.py:RelativePositionBiasHorizontal": ["Any", "Bbox", "Model", "ModelPositionBiasBase", "ModelPositionBiasHorizontal", "Need", "None", "Optional", "Tensor", "ValueError", "__init__", "are", "as", "attention_mask", "bbox", "bboxes", "bias", "class", "def", "dict", "dim", "for", "get_Model_position", "horizontal", "horizontal_position", "if", "in", "is", "kwargs", "max_distance", "mean", "not", "of", "position", "prepare_input", "raise", "range", "required", "return", "scale", "scaling_factor", "self", "small", "str", "super", "the", "there", "to", "values"], "udop/modeling_udop.py:RelativePositionBiasVertical": ["Any", "Bbox", "Model", "ModelPositionBiasBase", "ModelPositionBiasVertical", "Need", "None", "Optional", "Tensor", "ValueError", "__init__", "are", "as", "attention_mask", "bbox", "bboxes", "bias", "class", "def", "dict", "dim", "for", "get_Model_position", "if", "in", "is", "kwargs", "max_distance", "mean", "not", "of", "position", "prepare_input", "raise", "range", "required", "return", "scale", "scaling_factor", "self", "small", "str", "super", "the", "there", "to", "values", "vertical", "vertical_position"], "udop/modeling_udop.py:RelativePositionBiasAggregated": ["Any", "ModelPositionBiasAggregated", "ModelPositionBiasBase", "Module", "ModuleList", "None", "Optional", "Sequence", "Tensor", "Union", "__init__", "attention_mask", "bbox", "bias", "biases", "class", "def", "dict", "float", "for", "forward", "in", "modules", "nn", "output", "return", "self", "str", "super"], "udop/modeling_udop.py:create_relative_bias": ["BIAS_CLASSES", "ModelConfig", "Model_relative_bias", "Number", "RelativePositionBiasBase", "Sequence", "ValueError", "append", "bias_kwargs", "bias_kwargs_org", "bias_list", "bias_type", "config", "deepcopy", "def", "else", "for", "hasattr", "heads", "if", "in", "match", "model", "model_num_heads", "must", "num", "num_attention_heads", "num_heads", "of", "pop", "raise", "relative_bias_args", "return", "the", "type"], "udop/modeling_udop.py:UdopStack": ["AttentionMaskConverter", "BaseModelOutputWithAttentionMask", "BlockMask", "Cache", "Dropout", "DynamicCache", "Empty", "EncoderDecoderCache", "False", "ModelBlock", "ModelCellEmbeddings", "ModelConfig", "ModelLayerNorm", "ModelPreTrainedModel", "ModelStack", "ModuleList", "None", "RelativePositionBias1D", "RelativePositionBiasAggregated", "SelfAttention", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_get_relative_bias", "_ignore_causal_mask_sdpa", "_max_length", "_prepare_4d_causal_attention_mask_with_cache_position", "_tie_or_clone_weights", "_tie_weights", "_unmask_unattended", "_update_causal_mask", "a", "all_attentions", "all_cross_attentions", "all_hidden_states", "and", "any", "arange", "as", "assert", "at", "attention_mask", "attentions", "batch", "batch_size", "bbox", "be", "bias", "biases", "block", "bool", "both", "cache_position", "can", "cannot", "causal_mask", "cell_2d_embedding", "class", "clone", "combine_image_text_embeddings", "config", "create_relative_bias", "cross_attentions", "cross_attn_head_mask", "cuda", "d_model", "decoder", "decoder_", "def", "device", "diagonal", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_patches", "embed_tokens", "embeddings", "encoder_attention_mask", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_states", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_extended_attention_mask", "get_head_mask", "get_max_cache_shape", "get_output_embeddings", "get_seq_length", "has_relative_attention_bias", "have", "head_mask", "hidden_size", "hidden_states", "i", "if", "image_embeddings", "image_size", "in", "initialize", "input_ids", "input_shape", "input_tensor", "inputs", "inputs_embeds", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_torchdynamo_compiling", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "logger", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "max_2d_position_embeddings", "max_length", "min", "min_dtype", "model", "new_embeddings", "nn", "not", "npu", "num_layers", "num_patches", "numel", "ones", "only", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "patch_size", "pixel_values", "position_bias", "raise", "range", "relative_attention_bias", "relative_bias", "relative_bias_list", "reshape", "return", "return_dict", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "used", "using_compilable_cache", "v", "valid", "view", "visual_bbox", "warning", "with", "xpu", "zeros", "zeros_like"], "udop/modeling_udop.py:UdopModel": ["Any", "Cache", "Embedding", "False", "LongTensor", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "True", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bbox", "bias", "biases", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "dict", "else", "embed_patches", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "enumerate", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "idx", "if", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "last_hidden_state", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output_attentions", "output_hidden_states", "past_key_values", "patch_embed", "pixel_values", "post_init", "proj", "r", "relative_attention_bias", "relative_bias", "return", "return_dict", "self", "set_input_embeddings", "shared", "str", "super", "tie_encoder_decoder", "torch", "tuple", "use_cache", "use_return_dict", "value", "visual_bbox", "vocab_size", "weight"], "udop/modeling_udop.py:UdopForConditionalGeneration": ["Any", "Cache", "CrossEntropyLoss", "Embedding", "False", "GenerationMixin", "Linear", "LongTensor", "ModelForConditionalGeneration", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "True", "__init__", "_shift_right", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bias", "biases", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "deepcopy", "def", "dict", "else", "embed_patches", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "input_ids", "inputs_embeds", "is", "is_decoder", "labels", "last_hidden_state", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output", "output_attentions", "output_hidden_states", "past_key_values", "patch_embed", "pixel_values", "post_init", "proj", "r", "relative_attention_bias", "relative_bias", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "shared", "size", "str", "super", "tie_encoder_decoder", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "view", "visual_bbox", "vocab_size", "weight"], "udop/modeling_udop.py:UdopEncoderModel": ["Any", "BaseModelOutputWithAttentionMask", "Embedding", "False", "FloatTensor", "ModelConfig", "ModelEncoderModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "SelfAttention", "Tensor", "Union", "__init__", "_prune_heads", "_tied_weights_keys", "attention_mask", "auto_docstring", "bbox", "bias", "biases", "block", "bool", "class", "config", "d_model", "deepcopy", "def", "dict", "else", "embed_patches", "embed_tokens", "encoder", "encoder_config", "encoder_outputs", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "items", "layer", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "patch_embed", "pixel_values", "post_init", "proj", "prune_heads", "r", "relative_attention_bias", "relative_bias", "return", "return_dict", "self", "set_input_embeddings", "shared", "str", "super", "torch", "tuple", "use_cache", "use_return_dict", "visual_bbox", "vocab_size", "weight"], "glm/modeling_glm.py:GlmMLP": ["ACT2FN", "False", "FloatTensor", "Linear", "ModelMLP", "Module", "__init__", "activation_fn", "bias", "chunk", "class", "config", "def", "dim", "down_proj", "forward", "gate", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch", "up_states"], "glm/modeling_glm.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "glm/modeling_glm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "glm/modeling_glm.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "glm/modeling_glm.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "repeat_interleave", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "glm/modeling_glm.py:GlmAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "glm/modeling_glm.py:GlmRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "glm/modeling_glm.py:GlmRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "glm/modeling_glm.py:GlmDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "glm/modeling_glm.py:GlmPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "glm/modeling_glm.py:GlmModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "glm/modeling_glm.py:GlmForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "glm/modeling_glm.py:GlmForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "glm/modeling_glm.py:GlmForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "ctrl/modeling_ctrl.py:angle_defn": ["Model_defn", "Model_rates", "d_model_size", "def", "i", "pos", "pow", "return", "torch"], "ctrl/modeling_ctrl.py:positional_encoding": ["Model_encoding", "angle_defn", "angle_rads", "arange", "cat", "cos", "cosines", "d_model_size", "def", "dim", "dtype", "int64", "pos_encoding", "position", "return", "sin", "sines", "to", "torch", "unsqueeze"], "ctrl/modeling_ctrl.py:scaled_dot_product_attention": ["Model_attention_logits", "Model_dot_product_attention", "None", "attention_mask", "attention_weights", "def", "dim", "dk", "head_mask", "if", "is", "k", "mask", "matmul", "matmul_qk", "nd", "not", "np", "ns", "output", "permute", "q", "return", "shape", "size", "softmax", "sqrt", "torch", "v"], "ctrl/modeling_ctrl.py:MultiHeadAttention": ["False", "Linear", "ModelHeadAttention", "Module", "None", "Wk", "Wq", "Wv", "__init__", "attention_head_size", "attention_mask", "attn", "batch_size", "cache_position", "class", "d_model_size", "def", "dense", "depth", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "if", "index", "int", "is", "k", "layer_idx", "layer_past", "len", "mask", "nn", "not", "num_heads", "original_size_attention", "output", "output_attentions", "permute", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "reshape", "return", "scaled_attention", "scaled_dot_product_attention", "self", "set", "shape", "split_into_heads", "super", "union", "update", "use_cache", "v", "x"], "ctrl/modeling_ctrl.py:point_wise_feed_forward_network": ["Linear", "Model_wise_feed_forward_network", "ReLU", "Sequential", "d_model_size", "def", "dff", "nn", "return"], "ctrl/modeling_ctrl.py:EncoderLayer": ["Dropout", "False", "LayerNorm", "ModelLayer", "Module", "MultiHeadAttention", "None", "__init__", "attention_mask", "attn_output", "attn_outputs", "cache_position", "class", "d_model_size", "def", "dff", "dropout1", "dropout2", "eps", "ffn", "ffn_output", "forward", "head_mask", "layer_idx", "layer_past", "layernorm1", "layernorm2", "mask", "multi_head_attention", "nn", "normed", "num_heads", "out1", "out2", "output_attentions", "outputs", "point_wise_feed_forward_network", "rate", "return", "self", "super", "use_cache", "x"], "ctrl/modeling_ctrl.py:CTRLPreTrainedModel": ["Conv1D", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "transformer", "weight", "zero_"], "ctrl/modeling_ctrl.py:CTRLModel": ["BaseModelOutputWithPast", "Cache", "Dropout", "DynamicCache", "Embedding", "EncoderLayer", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Tensor", "Transformers", "Union", "ValueError", "You", "__init__", "_prune_heads", "a", "all_attentions", "all_hidden_states", "an", "and", "arange", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "both", "cache_position", "cannot", "class", "config", "d_model_size", "def", "defined", "deprecated", "device", "dff", "dropout", "dtype", "e", "either", "elif", "else", "embd_pdrop", "enumerate", "eps", "finfo", "float", "for", "forward", "from_legacy_cache", "g", "get_head_mask", "get_input_embeddings", "get_seq_length", "h", "has", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "i", "if", "in", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "layer_idx", "layer_norm_epsilon", "layer_past", "layernorm", "logger", "long", "mask", "min", "multi_head_attention", "n_embd", "n_head", "n_layer", "n_positions", "new_embeddings", "nn", "not", "np", "num_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "pass", "past_key_values", "past_length", "pos_embeds", "pos_encoding", "position_ids", "positional_encoding", "post_init", "prune_heads", "r", "raise", "range", "removed", "resid_pdrop", "return", "return_dict", "same", "self", "seq_len", "set_input_embeddings", "shape", "should", "size", "specify", "sqrt", "super", "the", "time", "to", "token_type_embeds", "token_type_ids", "torch", "triu", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "w", "warn_if_padding_and_no_attention_mask", "warning_once", "will"], "ctrl/modeling_ctrl.py:CTRLLMHeadModel": ["Cache", "CausalLMOutputWithPast", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "Warning", "__init__", "_tied_weights_keys", "a", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "def", "else", "f", "for", "forward", "get_seq_length", "head_mask", "hidden_states", "if", "in", "input", "input_ids", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_function", "model_inputs", "n_embd", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "past_length", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "print", "r", "recognized", "remove_prefix_length", "return", "return_dict", "self", "shape", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "ctrl/modeling_ctrl.py:CTRLForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "classifier", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "n_embd", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "self", "sequence_length", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "token_type_ids", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "llama/modeling_llama.py:LlamaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "llama/modeling_llama.py:LlamaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "llama/modeling_llama.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "llama/modeling_llama.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "llama/modeling_llama.py:LlamaMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "llama/modeling_llama.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "llama/modeling_llama.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "llama/modeling_llama.py:LlamaAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "llama/modeling_llama.py:LlamaDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "llama/modeling_llama.py:LlamaPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "llama/modeling_llama.py:LlamaModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "llama/modeling_llama.py:LlamaForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "llama/modeling_llama.py:LlamaForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "llama/modeling_llama.py:LlamaForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "llama/modeling_llama.py:LlamaForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class"], "perceiver/modeling_perceiver.py:PerceiverModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "last_hidden_state", "logits", "r", "torch", "tuple"], "perceiver/modeling_perceiver.py:PerceiverDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "class", "cross_attentions", "logits", "r", "torch", "tuple"], "perceiver/modeling_perceiver.py:PerceiverMaskedLMOutput": ["FloatTensor", "ModelMaskedLMOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "perceiver/modeling_perceiver.py:PerceiverClassifierOutput": ["FloatTensor", "ModelClassifierOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "perceiver/modeling_perceiver.py:PerceiverEmbeddings": ["ModelEmbeddings", "Module", "Parameter", "__init__", "batch_size", "class", "config", "d_latents", "def", "expand", "forward", "int", "latents", "nn", "num_latents", "randn", "return", "self", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverSelfAttention": ["Dropout", "False", "FloatTensor", "Identity", "LayerNorm", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Softmax", "Tensor", "ValueError", "_", "__init__", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "be", "bool", "by", "channels_per_head", "class", "config", "context_layer", "contiguous", "def", "dim", "divisible", "dropout", "else", "f", "forward", "head_mask", "hidden_states", "hiddens", "if", "inputs", "inputs_mask", "is", "is_cross_attention", "key", "keys", "kv_dim", "layernorm1", "layernorm2", "math", "matmul", "must", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_heads", "output_attentions", "outputs", "permute", "q_dim", "q_head_dim", "qk_channels", "qk_channels_per_head", "queries", "query", "raise", "return", "self", "seq_len", "shape", "size", "sqrt", "super", "torch", "transpose", "transpose_for_scores", "tuple", "v_channels", "v_channels_per_head", "v_head_dim", "value", "values", "view", "x"], "perceiver/modeling_perceiver.py:PerceiverSelfOutput": ["Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "forward", "hidden_states", "input_channels", "nn", "output_channels", "return", "self", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "True", "Unknown", "ValueError", "__init__", "all_head_size", "and", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "cross_attention_shape_for_attention", "def", "dense", "dim", "elif", "else", "f", "find_pruneable_heads_and_indices", "for", "forward", "head_mask", "heads", "hidden_states", "if", "index", "input_channels", "inputs", "inputs_mask", "is", "is_cross_attention", "key", "kv", "kv_dim", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "output_channels", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "q_dim", "qk_channels", "query", "raise", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "use_query_residual", "v_channels", "value"], "perceiver/modeling_perceiver.py:PerceiverMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "class", "config", "def", "dense1", "dense2", "else", "forward", "hidden_act", "hidden_states", "if", "input_size", "intermediate_act_fn", "isinstance", "nn", "return", "self", "str", "super", "torch", "widening_factor"], "perceiver/modeling_perceiver.py:PerceiverLayer": ["False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelLayer", "ModelMLP", "Module", "None", "Optional", "Tensor", "True", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "attention_outputs", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "input_size", "inputs", "inputs_mask", "is_cross_attention", "kv_dim", "layer_output", "layernorm", "mlp", "nn", "num_heads", "output_attentions", "outputs", "q_dim", "qk_channels", "return", "self", "seq_len_dim", "super", "torch", "tuple", "use_query_residual", "v_channels", "widening_factor"], "perceiver/modeling_perceiver.py:PerceiverEncoder": ["BaseModelOutputWithCrossAttentions", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "ValueError", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "append", "attention_mask", "attentions", "be", "bool", "by", "class", "config", "cross_attention", "cross_attention_widening_factor", "cross_attentions", "d_latents", "def", "divisible", "else", "enumerate", "f", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "inputs", "inputs_mask", "is", "is_cross_attention", "kv_dim", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "must", "nn", "not", "num_blocks", "num_cross_attend_heads", "num_cross_attention_heads", "num_heads", "num_self_attend_heads", "num_self_attends_per_block", "num_self_attention_heads", "num_z_channels", "output_attentions", "output_hidden_states", "q_dim", "qk_channels", "raise", "range", "return", "return_dict", "self", "self_attends", "self_attention_layers", "self_attention_widening_factor", "super", "torch", "tuple", "use_query_residual", "v", "v_channels", "widening_factor"], "perceiver/modeling_perceiver.py:PerceiverPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelTrainablePositionEncoding", "None", "ParameterDict", "PreTrainedModel", "_init_weights", "and", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "for", "hasattr", "if", "in", "initializer_range", "inputs", "is", "isinstance", "latents", "main_input_name", "mean", "modality", "module", "nn", "normal_", "not", "padding_idx", "position_embeddings", "self", "std", "weight", "zero_"], "perceiver/modeling_perceiver.py:PerceiverModel": ["False", "FloatTensor", "Last", "Make", "ModelAbstractDecoder", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "PostprocessorType", "PreprocessorType", "Tensor", "Union", "ValueError", "_", "__init__", "_prune_heads", "and", "appropriately", "attention", "attention_mask", "attentions", "audio", "auto_docstring", "batch_size", "bool", "class", "config", "correspond", "cross_attentions", "d_model", "decoder", "decoder_outputs", "decoder_query", "def", "device", "dict", "dimension", "doesn", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "f", "for", "forward", "get_head_mask", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "image", "in", "input_preprocessor", "inputs", "inputs_mask", "inputs_without_pos", "interpolate_pos_encoding", "invert_attention_mask", "is", "items", "kv_dim", "label", "last_hidden_state", "latents", "layer", "logits", "modality_sizes", "not", "num_blocks", "num_channels", "num_self_attends_per_block", "of", "ones", "output_attentions", "output_hidden_states", "output_modality_sizes", "output_postprocessor", "post_init", "prune_heads", "query_mask", "r", "raise", "return", "return_dict", "self", "seq_length", "sequence_output", "set", "set_input_embeddings", "shape", "size", "str", "subsampled_output_points", "subsampled_points", "super", "sure", "t", "the", "to", "torch", "tuple", "use_return_dict", "value", "z"], "perceiver/modeling_perceiver.py:PerceiverForMaskedLM": ["CrossEntropyLoss", "False", "Model", "ModelBasicDecoder", "ModelConfig", "ModelEmbeddingDecoder", "ModelForMaskedLM", "ModelMaskedLMOutput", "ModelModel", "ModelPreTrainedModel", "ModelTextPreprocessor", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "cross_attentions", "d_latents", "decoder", "def", "elif", "else", "embedding_decoder", "embedding_layer", "embeddings", "final_project", "forward", "head_mask", "hidden_states", "if", "index_dims", "input_ids", "input_preprocessor", "inputs", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "max_position_embeddings", "not", "num_channels", "num_heads", "output", "output_attentions", "output_hidden_states", "output_index_dims", "output_num_channels", "outputs", "post_init", "qk_channels", "r", "raise", "return", "return_dict", "self", "super", "text_preprocessor", "torch", "trainable_position_encoding_kwargs", "trainable_position_encoding_kwargs_decoder", "tuple", "use", "use_query_residual", "use_return_dict", "v_channels", "view", "vocab_size"], "perceiver/modeling_perceiver.py:PerceiverForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelClassificationDecoder", "ModelClassifierOutput", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "ModelTextPreprocessor", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "cross_attentions", "d_latents", "decoder", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "index_dims", "input_ids", "input_preprocessor", "inputs", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_channels", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "torch", "trainable_position_encoding_kwargs", "trainable_position_encoding_kwargs_decoder", "tuple", "use", "use_query_residual", "use_return_dict", "view"], "perceiver/modeling_perceiver.py:PerceiverForImageClassificationLearned": ["False", "Model", "ModelClassificationDecoder", "ModelClassifierOutput", "ModelForImageClassificationLearned", "ModelImagePreprocessor", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "concat", "concat_or_add_pos", "config", "conv1x1", "cross_attentions", "d_latents", "decoder", "def", "elif", "else", "forward", "head_mask", "hidden_states", "if", "image_size", "index_dims", "input_preprocessor", "inputs", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "not", "num_channels", "num_labels", "out_channels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "position_encoding_type", "post_init", "prep_type", "project_pos_dim", "r", "raise", "return", "return_dict", "self", "spatial_downsample", "super", "torch", "trainable", "trainable_position_encoding_kwargs", "trainable_position_encoding_kwargs_decoder", "trainable_position_encoding_kwargs_preprocessor", "tuple", "use", "use_query_residual", "use_return_dict"], "perceiver/modeling_perceiver.py:PerceiverForImageClassificationFourier": ["False", "Model", "ModelClassificationDecoder", "ModelClassifierOutput", "ModelForImageClassificationFourier", "ModelImagePreprocessor", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "concat_pos", "config", "cross_attentions", "d_latents", "decoder", "def", "elif", "else", "forward", "fourier_position_encoding_kwargs", "fourier_position_encoding_kwargs_preprocessor", "head_mask", "hidden_states", "if", "index_dims", "input_preprocessor", "inputs", "is", "labels", "logits", "loss", "loss_function", "max_resolution", "not", "num_bands", "num_channels", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pixels", "post_init", "prep_type", "r", "raise", "return", "return_dict", "self", "sine_only", "spatial_downsample", "super", "torch", "trainable_position_encoding_kwargs", "trainable_position_encoding_kwargs_decoder", "tuple", "use", "use_query_residual", "use_return_dict"], "perceiver/modeling_perceiver.py:PerceiverForImageClassificationConvProcessing": ["False", "Model", "ModelClassificationDecoder", "ModelClassifierOutput", "ModelForImageClassificationConvProcessing", "ModelImagePreprocessor", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "concat_pos", "config", "conv", "cross_attentions", "d_latents", "decoder", "def", "elif", "else", "forward", "fourier", "fourier_position_encoding_kwargs", "fourier_position_encoding_kwargs_preprocessor", "head_mask", "hidden_states", "if", "index_dims", "input_preprocessor", "inputs", "is", "labels", "logits", "loss", "loss_function", "max_resolution", "not", "num_bands", "num_channels", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "position_encoding_type", "post_init", "prep_type", "r", "raise", "return", "return_dict", "self", "sine_only", "spatial_downsample", "super", "torch", "trainable_position_encoding_kwargs", "trainable_position_encoding_kwargs_decoder", "tuple", "use", "use_query_residual", "use_return_dict"], "perceiver/modeling_perceiver.py:PerceiverForOpticalFlow": ["False", "Model", "ModelClassifierOutput", "ModelForOpticalFlow", "ModelImagePreprocessor", "ModelModel", "ModelOpticalFlowDecoder", "ModelPreTrainedModel", "None", "NotImplementedError", "Optical", "Optional", "Tensor", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "concat_pos", "config", "conv_after_patching", "conv_after_patching_in_channels", "cross_attentions", "decoder", "def", "else", "flow", "forward", "fourier", "fourier_position_encoding_kwargs", "fourier_position_encoding_kwargs_decoder", "fourier_position_encoding_kwargs_preprocessor", "head_mask", "hidden_states", "if", "image_preprocessor", "input_preprocessor", "inputs", "is", "labels", "logits", "loss", "max_resolution", "not", "num_bands", "num_channels", "output", "output_attentions", "output_hidden_states", "output_image_shape", "output_num_channels", "outputs", "patches", "position_encoding_type", "post_init", "prep_type", "r", "raise", "rescale_factor", "return", "return_dict", "self", "sine_only", "spatial_downsample", "super", "supported", "temporal_downsample", "torch", "train_size", "training", "tuple", "use_query_residual", "use_return_dict", "yet"], "perceiver/modeling_perceiver.py:PerceiverForMultimodalAutoencoding": ["False", "Model", "ModelAudioPostprocessor", "ModelAudioPreprocessor", "ModelBasicDecoder", "ModelBasicVideoAutoencodingDecoder", "ModelClassificationDecoder", "ModelClassificationPostprocessor", "ModelClassifierOutput", "ModelConfig", "ModelForMultimodalAutoencoding", "ModelImagePreprocessor", "ModelModel", "ModelMultimodalDecoder", "ModelMultimodalPostprocessor", "ModelMultimodalPreprocessor", "ModelOneHotPreprocessor", "ModelPreTrainedModel", "ModelProjectionPostprocessor", "Multimodal", "None", "NotImplementedError", "Optional", "Tensor", "True", "Union", "__init__", "_label_trainable_num_channels", "attention_mask", "attentions", "audio", "audio_samples_per_frame", "auto_docstring", "autoencoding", "bool", "class", "concat_pos", "concat_preprocessed_input", "config", "cross_attentions", "decoder", "def", "dict", "else", "forward", "fourier", "fourier_position_encoding_kwargs", "head_mask", "hidden_states", "if", "image", "image_decoder", "image_size", "in_channels", "index_dims", "input_preprocessor", "inputs", "is", "label", "labels", "logits", "loss", "mask_probs", "max_resolution", "min_padding_size", "modalities", "n_audio_samples", "not", "num_bands", "num_channels", "num_frames", "num_outputs", "out_channels", "output", "output_attentions", "output_hidden_states", "output_index_dims", "output_num_channels", "output_postprocessor", "output_shape", "outputs", "patches", "position_encoding_only", "position_encoding_type", "post_init", "prep_type", "r", "raise", "return", "return_dict", "samples_per_patch", "self", "sine_only", "spatial_downsample", "str", "subsampled_output_points", "super", "supported", "temporal_downsample", "torch", "trainable", "trainable_position_encoding_kwargs", "training", "tuple", "use_query_residual", "use_return_dict", "yet"], "perceiver/modeling_perceiver.py:build_position_encoding": ["Identity", "Linear", "Make", "ModelFourierPositionEncoding", "ModelTrainablePositionEncoding", "Model_position_encoding", "None", "Unknown", "ValueError", "def", "elif", "else", "encoding", "f", "fourier", "fourier_position_encoding_kwargs", "if", "nn", "not", "out_channels", "output_pos_enc", "pass", "position", "position_encoding_type", "positions_projection", "project_pos_dim", "raise", "return", "sure", "to", "trainable", "trainable_position_encoding_kwargs", "type"], "perceiver/modeling_perceiver.py:PerceiverAbstractDecoder": ["ABCMeta", "ModelAbstractDecoder", "Module", "None", "NotImplementedError", "abc", "abstractmethod", "class", "decoder_query", "def", "forward", "inputs", "inputs_without_pos", "metaclass", "modality_sizes", "nn", "num_query_channels", "property", "query", "query_mask", "raise", "self", "subsampled_points", "z"], "perceiver/modeling_perceiver.py:PerceiverProjectionDecoder": ["FloatTensor", "Linear", "ModelAbstractDecoder", "ModelProjectionDecoder", "None", "Optional", "Tensor", "__init__", "class", "classifier", "config", "d_latents", "decoder_query", "def", "dim", "forward", "inputs", "inputs_without_pos", "logits", "mean", "modality_sizes", "nn", "num_labels", "query", "query_mask", "return", "self", "subsampled_points", "super", "torch", "z"], "perceiver/modeling_perceiver.py:PerceiverBasicDecoder": ["False", "FloatTensor", "Identity", "Linear", "ModelAbstractDecoder", "ModelBasicDecoder", "ModelConfig", "ModelDecoderOutput", "ModelLayer", "None", "Optional", "Tensor", "True", "Value", "ValueError", "You", "__init__", "attention_mask", "batch_size", "bool", "broadcast_to", "build_position_encoding", "calculate", "cannot", "cat", "channels", "class", "concat_preprocessed_input", "config", "construct", "cpu", "cross_attentions", "d_latents", "decoder", "decoder_query", "decoding_cross_attention", "def", "device", "dim", "dtype", "elif", "else", "final_layer", "final_project", "for", "forward", "fourier", "from_numpy", "head_mask", "if", "in", "index_dims", "indices", "inputs", "inputs_mask", "inputs_without_pos", "int", "is", "is_cross_attention", "kv_dim", "layer_outputs", "logits", "modality_sizes", "nn", "none", "not", "np", "num_channels", "num_heads", "num_query_channels", "number", "of", "output", "output_attentions", "output_index_dims", "output_num_channels", "output_position_encodings", "output_size", "pos", "pos_emb", "position_encoding_kwargs", "position_encoding_only", "position_encoding_type", "positions_projection", "project_pos_dim", "property", "q_dim", "qk_channels", "queries", "query", "query_mask", "raise", "required", "reshape", "return", "self", "set", "shape", "stack", "str", "subsampled_index_dims", "subsampled_points", "super", "tensor", "to", "torch", "trainable", "unravel_index", "use_query_residual", "v_channels", "when", "widening_factor", "x", "z"], "perceiver/modeling_perceiver.py:PerceiverClassificationDecoder": ["False", "FloatTensor", "ModelAbstractDecoder", "ModelBasicDecoder", "ModelClassificationDecoder", "ModelDecoderOutput", "None", "Optional", "Tensor", "__init__", "bool", "class", "config", "cross_attentions", "decoder", "decoder_kwargs", "decoder_outputs", "decoder_query", "def", "forward", "inputs", "inputs_without_pos", "int", "logits", "modality_sizes", "num_labels", "num_query_channels", "output_attentions", "output_index_dims", "output_num_channels", "property", "query", "query_mask", "return", "self", "subsampled_points", "super", "torch", "z"], "perceiver/modeling_perceiver.py:PerceiverOpticalFlowDecoder": ["False", "FloatTensor", "FlowDecoder", "ModelAbstractDecoder", "ModelBasicDecoder", "ModelDecoderOutput", "ModelOpticalFlowDecoder", "None", "Optional", "Tensor", "ValueError", "__init__", "bool", "class", "config", "cross_attentions", "decoder", "decoder_kwargs", "decoder_outputs", "decoder_query", "def", "doesn", "forward", "if", "inputs", "inputs_without_pos", "int", "is", "list", "logits", "modality_sizes", "not", "num_query_channels", "output_attentions", "output_image_shape", "output_num_channels", "preds", "property", "query", "query_mask", "raise", "rescale_factor", "reshape", "return", "self", "shape", "subsampled_points", "subsampling", "super", "support", "t", "torch", "yet", "z"], "perceiver/modeling_perceiver.py:PerceiverBasicVideoAutoencodingDecoder": ["Expected", "FloatTensor", "ModelAbstractDecoder", "ModelBasicDecoder", "ModelBasicVideoAutoencodingDecoder", "ModelConfig", "ModelDecoderOutput", "None", "Optional", "Tensor", "ValueError", "__init__", "class", "config", "cross_attentions", "decoder", "decoder_kwargs", "decoder_outputs", "decoder_query", "def", "f", "forward", "got", "if", "inputs", "inputs_without_pos", "int", "len", "list", "logits", "modality_sizes", "num_query_channels", "output_index_dims", "output_num_channels", "output_shape", "position_encoding_type", "property", "query", "query_mask", "raise", "rank", "reshape", "return", "self", "shape", "str", "subsampled_points", "super", "torch", "z"], "perceiver/modeling_perceiver.py:restructure": ["Mapping", "ModalitySizeType", "Model", "Tensor", "def", "for", "in", "index", "inp", "inputs", "keys", "modality", "modality_sizes", "outputs", "return", "size", "sorted", "str", "torch"], "perceiver/modeling_perceiver.py:PerceiverMultimodalDecoder": ["False", "FloatTensor", "ModelAbstractDecoder", "ModelBasicDecoder", "ModelConfig", "ModelMultimodalDecoder", "ModuleDict", "None", "Optional", "Parameter", "ParameterDict", "Tensor", "_", "__init__", "bool", "broadcast_to", "cat", "class", "common_channel_size", "config", "decoder", "decoder_kwargs", "decoder_outputs", "decoder_queries", "decoder_query", "def", "dict", "dim", "embed", "for", "forward", "get", "if", "in", "input_without_pos", "inputs", "inputs_without_pos", "int", "is", "items", "keys", "max", "max_channel_size", "min_padding_size", "modalities", "modality", "modality_sizes", "nn", "none", "not", "np", "num_channels", "num_outputs", "num_query_channels", "or", "output_attentions", "output_index_dims", "output_num_channels", "padding", "pos", "position_encoding_type", "prod", "property", "query", "query_mask", "randn", "reshape", "restructure", "return", "self", "shape", "sorted", "str", "subsampled_index_dims", "subsampled_points", "super", "torch", "x", "z"], "perceiver/modeling_perceiver.py:space_to_depth": ["Frames", "Model_to_depth", "Tensor", "ValueError", "batch", "batch_size", "be", "channels", "contiguous", "def", "elif", "else", "frames", "height", "if", "int", "len", "num_channels", "of", "or", "permute", "raise", "rank", "return", "shape", "should", "spatial_block_size", "temporal_block_size", "time", "torch", "view", "width"], "perceiver/modeling_perceiver.py:Conv2dSamePadding": ["Model", "ModelSamePadding", "ZeroPad2d", "__add__", "__init__", "_conv_forward", "args", "bias", "class", "def", "for", "forward", "in", "input", "k", "kernel_size", "kwargs", "nn", "reduce", "return", "self", "super", "weight", "zero_pad_2d"], "perceiver/modeling_perceiver.py:Conv2DDownsample": ["BatchNorm2d", "False", "Identity", "MaxPool2d", "Model", "ModelDDownsample", "ModeldSamePadding", "Module", "ReLU", "Tensor", "True", "__init__", "batchnorm", "bias", "bool", "class", "def", "else", "forward", "if", "in_channels", "inputs", "int", "kernel_size", "max_pool", "nn", "num_features", "num_layers", "out", "out_channels", "relu", "return", "self", "stride", "super", "torch", "use_batchnorm"], "perceiver/modeling_perceiver.py:generate_fourier_features": ["False", "Model_fourier_features", "None", "True", "batch_size", "cat", "concat_pos", "cos", "def", "dim", "else", "end", "expand", "for", "freq_bands", "if", "in", "linspace", "max_resolution", "min_freq", "np", "num_bands", "per_pos_features", "pi", "pos", "prod", "res", "reshape", "return", "shape", "sin", "sine_only", "stack", "start", "steps", "torch"], "perceiver/modeling_perceiver.py:build_linear_positions": ["Model_linear_positions", "_linspace", "array_index_grid", "def", "dim", "dim_ranges", "dtype", "end", "float32", "for", "ij", "in", "index_dims", "indexing", "linspace", "meshgrid", "n_xels_per_dim", "output_range", "return", "stack", "start", "steps", "torch"], "perceiver/modeling_perceiver.py:PerceiverAbstractPositionEncoding": ["ABCMeta", "ModelAbstractPositionEncoding", "Module", "NotImplementedError", "abc", "abstractmethod", "args", "batch_size", "class", "def", "forward", "int", "kwargs", "metaclass", "nn", "num_dimensions", "output_size", "pos", "property", "raise", "self"], "perceiver/modeling_perceiver.py:PerceiverTrainablePositionEncoding": ["False", "ModelAbstractPositionEncoding", "ModelTrainablePositionEncoding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_index_dims", "_num_channels", "align_corners", "and", "args", "batch_size", "bicubic", "bool", "class", "def", "expand", "forward", "functional", "height", "if", "index_dim", "index_dims", "input_size", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "isinstance", "jit", "kwargs", "len", "mode", "new_height", "new_width", "nn", "not", "np", "num_channels", "num_dimensions", "num_positions", "output_size", "permute", "position_embeddings", "prod", "property", "randn", "reshape", "return", "self", "shape", "size", "squeeze", "super", "torch", "torch_int", "width"], "perceiver/modeling_perceiver.py:_check_or_build_spatial_positions": ["None", "Spatial", "ValueError", "_check_or_build_spatial_positions", "batch_size", "build_linear_positions", "def", "dimensions", "else", "expand", "features", "have", "if", "index_dims", "is", "len", "np", "number", "of", "pos", "prod", "raise", "reshape", "return", "shape", "the", "torch", "wrong"], "perceiver/modeling_perceiver.py:PerceiverFourierPositionEncoding": ["False", "FloatTensor", "ModelAbstractPositionEncoding", "ModelFourierPositionEncoding", "None", "Optional", "True", "__init__", "_check_or_build_spatial_positions", "batch_size", "class", "concat_pos", "def", "device", "dtype", "encoding_size", "forward", "fourier_pos_enc", "generate_fourier_features", "if", "index_dims", "int", "len", "list", "max_resolution", "not", "num_bands", "num_dimensions", "num_dims", "output_size", "pos", "property", "return", "self", "sine_only", "super", "to", "torch"], "perceiver/modeling_perceiver.py:AbstractPreprocessor": ["ModelPreprocessor", "Module", "NotImplementedError", "class", "def", "int", "nn", "num_channels", "property", "raise", "self"], "perceiver/modeling_perceiver.py:PerceiverTextPreprocessor": ["AbstractPreprocessor", "Embedding", "False", "LongTensor", "ModelConfig", "ModelTextPreprocessor", "None", "Optional", "Tensor", "True", "__init__", "arange", "bool", "class", "config", "d_model", "def", "device", "embedding_dim", "embeddings", "embeddings_without_pos", "forward", "inputs", "int", "interpolate_pos_encoding", "max_position_embeddings", "network_input_is_1d", "nn", "num_channels", "num_embeddings", "pos", "position_embeddings", "position_ids", "property", "return", "self", "seq_length", "shape", "super", "torch", "vocab_size"], "perceiver/modeling_perceiver.py:PerceiverEmbeddingDecoder": ["ModelConfig", "ModelEmbeddingDecoder", "Module", "None", "Parameter", "Tensor", "__init__", "batch_size", "bias", "class", "config", "d_model", "def", "embedding_layer", "forward", "hidden_states", "matmul", "nn", "output", "reshape", "return", "self", "seq_len", "shape", "super", "torch", "transpose", "vocab_size", "weight", "zeros"], "perceiver/modeling_perceiver.py:PerceiverMultimodalPostprocessor": ["False", "Mapping", "Modality", "ModelMultimodalPostprocessor", "Module", "ModuleDict", "None", "Optional", "PostprocessorType", "Tensor", "ValueError", "__init__", "a", "be", "bool", "class", "def", "dictionary", "for", "forward", "if", "in", "input", "input_is_dict", "inputs", "is", "items", "modalities", "modality", "modality_sizes", "nn", "not", "outputs", "pos", "postprocessor", "raise", "restructure", "return", "self", "should", "sizes", "specified", "str", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverClassificationPostprocessor": ["Linear", "ModelClassificationPostprocessor", "ModelConfig", "Module", "None", "Optional", "Tensor", "__init__", "class", "classifier", "config", "def", "forward", "in_channels", "inputs", "int", "logits", "modality_sizes", "nn", "num_labels", "pos", "return", "self", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverAudioPostprocessor": ["Invalid", "Linear", "ModelAudioPostprocessor", "ModelConfig", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "class", "classifier", "config", "def", "forward", "if", "in", "in_channels", "inputs", "int", "logits", "modality_sizes", "nn", "not", "patches", "pos", "postproc_type", "raise", "reshape", "return", "samples_per_patch", "self", "shape", "str", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverProjectionPostprocessor": ["Linear", "ModelProjectionPostprocessor", "Module", "None", "Optional", "Tensor", "__init__", "class", "classifier", "def", "forward", "in_channels", "inputs", "int", "logits", "modality_sizes", "nn", "out_channels", "pos", "return", "self", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverImagePreprocessor": ["AbstractPreprocessor", "Conv1x1", "Conv2DDownsample", "Conv2d", "False", "Identity", "Invalid", "Linear", "ModelImagePreprocessor", "None", "Only", "Optional", "Prep_type", "Tensor", "True", "Unsupported", "ValueError", "__init__", "_build_network_inputs", "add", "and", "batch_size", "bool", "build_position_encoding", "cat", "ceil", "class", "concat", "concat_or_add_pos", "config", "conv", "conv1x1", "conv2d_use_batchnorm", "conv_after_patches", "conv_after_patching", "conv_after_patching_in_channels", "convnet", "convnet_1x1", "convnet_num_layers", "convnet_num_layers_is_int", "data", "def", "device", "dim", "does", "downsample", "downsampling", "dtype", "elif", "else", "expected", "f", "for", "format", "forward", "fourier", "if", "in", "in_channels", "index_dims", "indices", "inp_dim", "input_size", "inputs", "inputs_with_pos", "inputs_without_pos", "int", "interpolate_pos_encoding", "invalid", "is", "is_temporal", "kernel_size", "len", "list", "log", "math", "modality_sizes", "ndim", "network_input_is_1d", "nn", "not", "np", "num_channels", "num_dimensions", "num_layers", "of", "or", "out_channels", "output_size", "patches", "permute", "pixels", "pos", "pos_dim", "pos_enc", "position_embeddings", "position_encoding_kwargs", "position_encoding_type", "positions_projection", "powers", "prep_type", "prod", "project_pos_dim", "property", "raise", "reshape", "return", "round", "self", "sh", "shape", "space_to_depth", "spatial", "spatial_block_size", "spatial_downsample", "squeeze", "str", "stride", "super", "temporal", "temporal_block_size", "temporal_downsample", "time", "torch", "trainable", "use_batchnorm", "value", "with"], "perceiver/modeling_perceiver.py:PerceiverOneHotPreprocessor": ["AbstractPreprocessor", "ModelConfig", "ModelOneHotPreprocessor", "None", "Optional", "Tensor", "True", "__init__", "bool", "class", "config", "def", "forward", "inputs", "int", "network_input_is_1d", "num_channels", "num_labels", "pos", "property", "return", "self", "super", "torch"], "perceiver/modeling_perceiver.py:PerceiverAudioPreprocessor": ["AbstractPreprocessor", "Concat_or_pos", "False", "ModelAudioPreprocessor", "None", "Optional", "Prep_type", "Tensor", "True", "ValueError", "__init__", "_build_network_inputs", "add", "batch_size", "be", "bool", "build_position_encoding", "can", "cat", "class", "concat", "concat_or_add_pos", "config", "def", "device", "dim", "dtype", "elif", "else", "f", "forward", "fourier", "if", "in", "index_dims", "inputs", "inputs_with_pos", "inputs_without_pos", "int", "interpolate_pos_encoding", "invalid", "is", "modality_sizes", "network_input_is_1d", "not", "num_channels", "only", "or", "out_channels", "output_size", "patches", "pos", "pos_dim", "pos_enc", "position_embeddings", "position_encoding_kwargs", "position_encoding_type", "positions_projection", "prep_type", "project_pos_dim", "property", "raise", "reshape", "return", "samples_per_patch", "self", "shape", "str", "super", "torch", "trainable"], "perceiver/modeling_perceiver.py:PerceiverMultimodalPreprocessor": ["AbstractPreprocessor", "False", "Mapping", "ModelMultimodalPreprocessor", "ModuleDict", "None", "Optional", "Parameter", "ParameterDict", "PreprocessorOutputType", "PreprocessorType", "Tensor", "True", "_", "__init__", "batch_size", "bernoulli", "bool", "broadcast_to", "cat", "class", "common_channel_size", "def", "device", "dim", "else", "expand", "final_inputs", "float", "for", "forward", "full", "if", "in", "inputs", "inputs_without_pos", "int", "interpolate_pos_encoding", "is", "items", "k", "keys", "mask", "mask_prob", "mask_probs", "mask_token", "max", "max_channel_size", "min_padding_size", "modalities", "modality", "modality_sizes", "network_input_is_1d", "nn", "not", "num_channels", "num_samples", "output", "output_padded", "padded", "padded_ls", "padding", "pos", "pos_enc", "preprocessor", "processor", "property", "randn", "return", "self", "shape", "sorted", "str", "super", "to", "torch", "unsqueeze"], "dab_detr/modeling_dab_detr.py:DabDetrDecoderOutput": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "ModelDecoderOutput", "None", "Optional", "class", "intermediate_hidden_states", "r", "reference_points", "torch", "tuple"], "dab_detr/modeling_dab_detr.py:DabDetrModelOutput": ["FloatTensor", "ModelModelOutput", "None", "Optional", "Seq2SeqModelOutput", "class", "intermediate_hidden_states", "r", "reference_points", "torch", "tuple"], "dab_detr/modeling_dab_detr.py:DabDetrObjectDetectionOutput": ["FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "dab_detr/modeling_dab_detr.py:DabDetrFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "dab_detr/modeling_dab_detr.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "dab_detr/modeling_dab_detr.py:DabDetrConvEncoder": ["ModelConfig", "ModelConvEncoder", "Module", "None", "Tensor", "__init__", "append", "backbone", "bool", "channels", "class", "config", "def", "feature_map", "feature_maps", "features", "float", "for", "forward", "functional", "in", "intermediate_channel_sizes", "interpolate", "load_backbone", "mask", "model", "nn", "no_grad", "out", "pixel_mask", "pixel_values", "replace_batch_norm", "return", "self", "shape", "size", "super", "to", "torch", "with"], "dab_detr/modeling_dab_detr.py:DabDetrConvModel": ["ModelConvModel", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "dab_detr/modeling_dab_detr.py:DabDetrSinePositionEmbedding": ["ModelConfig", "ModelSinePositionEmbedding", "Module", "No", "None", "ValueError", "__init__", "arange", "cat", "class", "config", "copy_", "cos", "cumsum", "def", "device", "dim", "dim_tx", "dim_ty", "dtype", "embedding_dim", "flatten", "float32", "forward", "hidden_size", "if", "is", "mask", "math", "mul_", "nn", "permute", "pi", "pixel", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "provided", "raise", "return", "scale", "self", "sin", "sine_position_embedding_scale", "stack", "super", "temperature_height", "temperature_width", "torch", "x_embed", "y_embed"], "dab_detr/modeling_dab_detr.py:gen_sine_position_embeddings": ["Model_sine_position_embeddings", "None", "Unknown", "ValueError", "arange", "cat", "cos", "def", "device", "dim", "dim_t", "div", "dtype", "else", "f", "flatten", "float32", "floor", "h_embed", "hidden_size", "if", "math", "pi", "pos", "pos_h", "pos_tensor", "pos_w", "pos_x", "pos_y", "raise", "return", "rounding_mode", "scale", "shape", "sin", "size", "stack", "to", "torch", "w_embed", "x_embed", "y_embed"], "dab_detr/modeling_dab_detr.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "dab_detr/modeling_dab_detr.py:DetrAttention": ["False", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "and", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "bool", "but", "by", "class", "config", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "hidden_states_original", "if", "is", "k_proj", "key_states", "key_value_states", "matmul", "must", "nn", "not", "num_heads", "object_queries", "of", "out_proj", "output_attentions", "p", "q_len", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "should", "size", "softmax", "super", "to", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "dab_detr/modeling_dab_detr.py:DabDetrAttention": ["False", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "and", "attention_dropout", "attention_head_dim", "attention_heads", "attention_mask", "attn_output", "attn_probs", "attn_weights", "batch_size", "be", "bias", "bool", "but", "by", "class", "config", "contiguous", "decoder_attention_heads", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float32", "forward", "functional", "got", "hidden_size", "hidden_states", "if", "is", "is_cross", "key_states", "matmul", "must", "nn", "not", "num_heads", "of", "output_attentions", "output_dim", "output_proj", "p", "q_len", "query_states", "raise", "reshape", "return", "scaling", "self", "should", "size", "softmax", "super", "to", "torch", "training", "transpose", "tuple", "value_states", "values_head_dim", "view"], "dab_detr/modeling_dab_detr.py:DabDetrDecoderLayerSelfAttention": ["LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayerSelfAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "def", "dropout", "forward", "functional", "hidden_size", "hidden_states", "key", "key_content", "key_pos", "key_states", "nn", "output_attentions", "p", "query", "query_content", "query_pos", "query_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_key_content_proj", "self_attn_key_pos_proj", "self_attn_layer_norm", "self_attn_query_content_proj", "self_attn_query_pos_proj", "self_attn_value_proj", "super", "torch", "training", "value", "value_states"], "dab_detr/modeling_dab_detr.py:DabDetrDecoderLayerCrossAttention": ["False", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayerCrossAttention", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "and", "attention_mask", "batch_size", "bool", "cat", "class", "config", "cross_attn", "cross_attn_key_content_proj", "cross_attn_key_pos_proj", "cross_attn_layer_norm", "cross_attn_query_content_proj", "cross_attn_query_pos_proj", "cross_attn_query_pos_sine_proj", "cross_attn_value_proj", "cross_attn_weights", "decoder_attention_heads", "def", "dim", "dropout", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "functional", "height_width", "hidden_size", "hidden_states", "if", "is", "is_cross", "is_first", "keep_query_pos", "key", "key_content", "key_pos", "key_states", "n_model", "nn", "not", "num_queries", "object_queries", "or", "output_attentions", "p", "query", "query_content", "query_pos", "query_position_embeddings", "query_sine_embed", "residual", "return", "self", "shape", "super", "torch", "training", "value", "value_states", "view"], "dab_detr/modeling_dab_detr.py:DabDetrDecoderLayerFFN": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayerFFN", "Module", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "class", "config", "decoder_ffn_dim", "def", "dropout", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_size", "hidden_states", "keep_query_pos", "nn", "p", "residual", "return", "self", "super", "torch", "training"], "dab_detr/modeling_dab_detr.py:DabDetrEncoderLayer": ["ACT2FN", "DetrAttention", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderLayer", "None", "Optional", "Tensor", "__init__", "activation_fn", "activation_function", "attention_mask", "attn_weights", "bool", "class", "config", "def", "dropout", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_size", "hidden_states", "if", "nn", "object_queries", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "dab_detr/modeling_dab_detr.py:DabDetrDecoderLayer": ["False", "GradientCheckpointingLayer", "ModelConfig", "ModelDecoderLayer", "ModelDecoderLayerCrossAttention", "ModelDecoderLayerFFN", "ModelDecoderLayerSelfAttention", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "class", "config", "cross_attn", "cross_attn_weights", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "hidden_states", "if", "is_first", "mlp", "object_queries", "output_attentions", "outputs", "query_position_embeddings", "query_sine_embed", "return", "self", "self_attn", "self_attn_weights", "super", "torch"], "dab_detr/modeling_dab_detr.py:DabDetrMLP": ["Linear", "ModelMLP", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "input_tensor", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "zip"], "dab_detr/modeling_dab_detr.py:DabDetrPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelConvEncoder", "ModelDecoderLayer", "ModelEncoderLayer", "ModelForObjectDetection", "ModelMHAttentionMap", "ModelPreTrainedModel", "None", "PReLU", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bbox_predictor", "bias", "bias_value", "class", "class_embed", "config", "constant_", "data", "def", "elif", "fill_", "gain", "if", "init", "init_std", "init_xavier_std", "initializer_bias_prior_prob", "is", "isinstance", "k_linear", "layers", "log", "main_input_name", "math", "mean", "model", "module", "nn", "normal_", "not", "num_labels", "or", "padding_idx", "pixel_values", "prior_prob", "q_linear", "r", "reset_parameters", "self", "std", "weight", "xavier_std", "xavier_uniform_", "zero_", "zeros_"], "dab_detr/modeling_dab_detr.py:DabDetrEncoder": ["BaseModelOutput", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelMLP", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "dtype", "else", "encoder_layer", "encoder_layers", "encoder_states", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "norm", "normalize_before", "not", "object_queries", "output_attentions", "output_hidden_states", "pos_scales", "post_init", "query_scale", "r", "range", "return", "return_dict", "scaled_object_queries", "self", "super", "tuple", "use_return_dict", "v"], "dab_detr/modeling_dab_detr.py:DabDetrDecoder": ["False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelMLP", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "__init__", "_prepare_4d_attention_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "append", "attentions", "bbox_embed", "bool", "class", "config", "cross_attentions", "decoder_layer", "decoder_layers", "def", "detach", "dropout", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gen_sine_position_embeddings", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_shape", "inputs_embeds", "intermediate", "intermediate_hidden_states", "inverse_sigmoid", "is", "is_first", "last_hidden_state", "layer_id", "layer_outputs", "layernorm", "layers", "memory_key_padding_mask", "new_reference_points", "nn", "not", "num_layers", "obj_center", "object_queries", "output_attentions", "output_hidden_states", "output_intermediate_hidden_states", "output_reference_points", "pos_transformation", "post_init", "query_dim", "query_pos", "query_position_embeddings", "query_scale", "query_sine_embed", "r", "range", "ref_anchor_head", "ref_point_head", "ref_points", "reference_anchor_size", "reference_points", "return", "return_dict", "self", "sigmoid", "size", "stack", "super", "tgt_len", "torch", "tuple", "unsqueeze", "use_return_dict", "v"], "dab_detr/modeling_dab_detr.py:DabDetrModel": ["Backbone", "BaseModelOutput", "Conv2d", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModelSinePositionEmbedding", "None", "Optional", "True", "Union", "ValueError", "_", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "aux_loss", "auxiliary_loss", "backbone", "batch_size", "be", "bool", "but", "class", "config", "conv_encoder", "cross_attentions", "data", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "device", "does", "downsampled", "elif", "else", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "feature_map", "features", "flatten", "flattened_features", "flattened_mask", "for", "forward", "freeze_backbone", "get_encoder", "height", "hidden_size", "hidden_states", "if", "in", "input_projection", "inputs_embeds", "int", "intermediate_channel_sizes", "intermediate_hidden_states", "inverse_sigmoid", "is", "isinstance", "kernel_size", "last_hidden_state", "len", "logger", "mask", "memory_key_padding_mask", "model", "name", "named_parameters", "nn", "not", "num_patterns", "num_queries", "object_queries", "object_queries_list", "ones", "output", "output_attentions", "output_hidden_states", "param", "patterns", "permute", "pixel", "pixel_mask", "pixel_values", "post_init", "projected_feature_map", "queries", "query_dim", "query_position_embeddings", "query_refpoint_embeddings", "r", "raise", "random_refpoints_xy", "reference_points", "reference_position_embeddings", "repeat", "requires_grad", "requires_grad_", "return", "return_dict", "self", "shape", "should", "super", "torch", "tuple", "type", "unfreeze_backbone", "uniform_", "unsqueeze", "use_return_dict", "warning", "weight", "width", "zeros"], "dab_detr/modeling_dab_detr.py:DabDetrMHAttentionMap": ["Dropout", "Linear", "ModelMHAttentionMap", "Module", "None", "Optional", "Tensor", "True", "__init__", "bias", "bnchw", "bqnc", "bqnhw", "class", "conv2d", "def", "dim", "dropout", "dtype", "einsum", "finfo", "flatten", "float", "forward", "functional", "hidden_dim", "if", "is", "k", "k_linear", "keys_per_head", "mask", "masked_fill", "min", "nn", "normalize_fact", "not", "num_heads", "q", "q_linear", "queries_per_head", "query_dim", "return", "self", "shape", "size", "softmax", "std", "super", "torch", "unsqueeze", "view", "weight", "weights"], "dab_detr/modeling_dab_detr.py:DabDetrForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForObjectDetection", "ModelMLP", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_bbox_embed", "_set_aux_loss", "_tied_weights_keys", "a", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "b", "bbox_embed", "bbox_predictor", "bbox_with_refinement", "bias", "bool", "class", "class_embed", "config", "cross_attentions", "d", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "def", "device", "dict", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "hidden_size", "if", "in", "inputs_embeds", "intermediate_hidden_states", "inverse_sigmoid", "is", "jit", "labels", "last_hidden_state", "layers", "list", "logits", "loss", "loss_dict", "loss_function", "model", "model_outputs", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs_class", "outputs_coord", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "query_dim", "r", "reference_before_sigmoid", "reference_points", "return", "return_dict", "self", "sigmoid", "super", "torch", "tuple", "unused", "use_return_dict", "weight", "zip"], "reformer/modeling_reformer.py:ReformerDynamicCache": ["Any", "Cache", "FloatTensor", "Iterable", "KeyError", "ModelDynamicCache", "None", "Optional", "Tensor", "__getitem__", "__init__", "__iter__", "__len__", "_distributed_cache_data", "_seen_tokens", "access", "append", "attempted", "buckets", "buckets_cache", "cache", "cache_kwargs", "cat", "class", "classmethod", "cls", "def", "device", "dict", "dim", "else", "f", "for", "from_legacy_cache", "get_seq_length", "has", "if", "in", "index", "int", "is", "layer", "layer_idx", "layers", "legacy_cache", "len", "list", "not", "numel", "only", "past_buckets_states", "raise", "range", "return", "self", "shape", "states", "states_cache", "str", "tensor", "to", "to_legacy_cache", "torch", "tuple", "update", "with", "yield"], "reformer/modeling_reformer.py:_stable_argsort": ["_stable_argsort", "arange", "argsort", "def", "device", "dim", "expand", "return", "scale_offset", "scaled_vector", "shape", "torch", "vector", "view"], "reformer/modeling_reformer.py:_get_least_common_mult_chunk_len": ["NotImplementedError", "Only", "Select", "_get_least_common_mult_chunk_len", "and", "attn", "attn_layers", "attn_types", "attn_types_set", "but", "config", "def", "elif", "else", "exist", "f", "from", "if", "layer", "lcm", "len", "local", "local_attn_chunk_length", "lsh", "lsh_attn_chunk_length", "np", "only", "raise", "return", "set", "types"], "reformer/modeling_reformer.py:_get_min_chunk_len": ["NotImplementedError", "Only", "Select", "_get_min_chunk_len", "and", "attn", "attn_layers", "attn_types", "attn_types_set", "but", "config", "def", "elif", "else", "exist", "f", "from", "if", "layer", "len", "local", "local_attn_chunk_length", "lsh", "lsh_attn_chunk_length", "min", "only", "raise", "return", "set", "types"], "reformer/modeling_reformer.py:AxialPositionEmbeddings": ["Got", "If", "Make", "ModelPositionEmbeddings", "Model_pos_embd_dim", "Model_pos_embds", "Model_pos_embds_dim", "Model_pos_shape", "Module", "Parameter", "ParameterList", "True", "ValueError", "You", "__init__", "_get_least_common_mult_chunk_len", "append", "at", "ax_shape", "axis", "batch_size", "broadcasted_weights", "cat", "changing", "class", "config", "consider", "def", "dim", "dropout", "dropout2d", "dropped_transposed_weights", "dropped_weights", "dtype", "else", "enumerate", "expand", "f", "factors", "float32", "for", "forward", "functional", "hidden_dropout_prob", "hidden_size", "i", "if", "in", "index_select", "is", "item", "least", "least_common_mult_chunk_length", "len", "length", "make", "max", "max_position_id", "might", "mul", "multiply", "nn", "ones", "or", "p", "padding", "position_encodings", "position_ids", "prod", "raise", "range", "reduce", "required_pos_encodings_columns", "reshape", "return", "self", "sequence", "sequence_length", "shape", "sum", "super", "sure", "that", "to", "torch", "training", "transpose", "transposed_weights", "tuple", "unsqueeze", "want", "weight", "weights", "your"], "reformer/modeling_reformer.py:PositionEmbeddings": ["Embedding", "ModelEmbeddings", "Model_embeddings", "Model_ids", "Module", "__init__", "class", "config", "def", "dropout", "embedding", "forward", "functional", "hidden_dropout_prob", "hidden_size", "max_Model_embeddings", "nn", "p", "return", "self", "super", "training"], "reformer/modeling_reformer.py:ReformerEmbeddings": ["AxialPositionEmbeddings", "Embedding", "Length", "ModelEmbeddings", "Module", "None", "PositionEmbeddings", "Sequence", "ValueError", "__init__", "arange", "axial_pos_embds", "be", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "equal", "expand", "f", "forward", "functional", "has", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "less", "long", "max_position_embeddings", "nn", "not", "or", "p", "position_embeddings", "position_ids", "raise", "return", "self", "seq_length", "shape", "size", "start_idx_pos_encodings", "super", "than", "to", "torch", "training", "unsqueeze", "vocab_size", "word_embeddings"], "reformer/modeling_reformer.py:EfficientAttentionMixin": ["Input", "ModelAttentionMixin", "None", "ValueError", "_look_adjacent", "_merge_hidden_size_dims", "_split_hidden_size_dim", "_split_seq_length_dim_to", "and", "append", "attn_head_size", "batch_size", "be", "but", "cat", "class", "def", "dim", "dim_factor_1", "dim_factor_2", "elif", "else", "f", "for", "i", "if", "in", "is", "len", "new_x_shape", "num_attn_heads", "num_chunks_after", "num_chunks_before", "of", "one", "permute", "raise", "range", "rank", "reshape", "return", "self", "shape", "should", "size", "slices", "split_dim_shape", "torch", "transpose", "vector", "vectors", "view", "x"], "reformer/modeling_reformer.py:LSHSelfAttention": ["At", "EfficientAttentionMixin", "False", "If", "Linear", "Make", "Model", "ModelOutput", "Module", "None", "ReverseSort", "Setting", "The", "There", "True", "__init__", "_attend", "_compute_attn_mask", "_expand_to_indices_in_relevant_chunk", "_gather_by_expansion", "_get_relevant_hid_states_and_buckets", "_get_sorted_bucket_idx_and_undo_sorted_bucket_idx", "_hash_vectors", "_len_and_dim_norm", "_len_norm", "_look_adjacent", "_merge_hidden_size_dims", "_query_per_attn_head", "_set_num_buckets", "_split_hidden_size_dim", "_split_seq_length_dim_to", "_stable_argsort", "_value_per_attn_head", "a", "ahr", "all_head_size", "an", "and", "apply", "arange", "are", "argmax", "assert", "at", "attention_head_size", "attention_mask", "attention_probs", "auto", "balh", "balr", "batch_size", "be", "bias", "bit_length", "bmhtr", "bmtd", "bool", "bucket_factor", "bucket_idx", "bucket_idx_batch_offset", "buckets", "buckets_cache", "buckets_mask", "but", "cache_position", "cat", "causal_mask", "chunk_length", "chunk_sequence_indices", "class", "clone", "concat_buckets", "config", "cur_product", "cur_sum", "def", "del", "detach", "device", "dim", "div", "do_standard_self_attention", "dropout", "dtype", "einsum", "elif", "else", "end_dim", "epsilon", "equals", "even", "exists_cache", "exp", "expand", "expanded_idxs", "expanded_start_indices", "f", "flatten", "float16", "floor", "for", "forward", "functional", "gather", "ge", "generation", "half", "has", "hash_seed", "have", "head_mask", "hidden_size", "hidden_states", "idxs", "if", "in", "increase_num_buckets", "index_select", "indices", "input", "int", "is", "is_decoder", "isinstance", "item", "keepdim", "key_indices", "key_value_bucket_idx", "key_value_hidden_states", "key_vectors", "kwargs", "language", "last", "layer_idx", "len", "length", "logger", "logits", "logsumexp", "long", "lsh_attention_probs_dropout_prob", "lsh_attn_chunk_length", "lsh_num_chunks_after", "lsh_num_chunks_before", "make", "manual_seed", "mask", "mask_value", "mask_value_float16", "mask_value_float32", "matmul", "max", "max_bucket", "max_position_embeddings", "mdhr", "mean", "moment", "mul", "ndim", "ne", "new", "new_ones", "nn", "no_grad", "nonzero", "norm_x", "not", "np", "num_attention_heads", "num_bucket", "num_buckets", "num_buckets_limit", "num_buckets_pow_2", "num_chunks_after", "num_chunks_before", "num_hashes", "number", "numel", "of", "offset", "offset_buckets", "offsets", "one", "ones_like", "only", "or", "out_vectors", "output_attentions", "p", "passed", "past_buckets", "past_buckets_states", "past_states", "per_head_query_key", "per_head_value", "persistent", "possible", "probs_vectors", "query_bucket_idx", "query_buckets", "query_indices", "query_key", "query_key_dot_shape", "query_key_dots", "query_key_vectors", "query_vectors", "randn", "random_rotations", "reduce", "register_buffer", "regressive", "relevant_bucket_idx", "relevant_bucket_idx_chunk", "relevant_bucket_idx_chunk_all_batch", "relevant_hidden_states", "repeat", "reshape", "return", "rotated_vectors", "rotated_vectors_factor", "rotation_size", "rotations_shape", "rounding_mode", "rsqrt", "scatter_", "self", "self_mask", "self_mask_value", "self_mask_value_float16", "self_mask_value_float32", "sequence", "sequence_length", "set", "shape", "should", "size", "sorted_bucket_idx", "sorted_bucket_idx_per_hash", "sqrt", "sqrt_num", "squeeze", "start_dim", "start_indices_chunk", "states_cache", "sum", "super", "sure", "tensor", "that", "the", "there", "time", "to", "torch", "total_chunk_size", "training", "transpose", "tuple", "undo_sorted_bucket_idx", "unsqueeze", "use_cache", "value", "value_vectors", "variance", "vectors", "view", "warning", "weight", "when", "where", "with", "word", "x"], "reformer/modeling_reformer.py:ReverseSort": ["Function", "ModelSort", "None", "backward", "class", "ctx", "def", "expand", "expanded_sort_indices", "expanded_undo_sort_indices", "forward", "gather", "grad_logits", "grad_out_vectors", "logits", "no_grad", "out_vectors", "return", "shape", "sorted_bucket_idx", "staticmethod", "torch", "undo_sorted_bucket_idx", "unsqueeze", "with"], "reformer/modeling_reformer.py:LocalSelfAttention": ["EfficientAttentionMixin", "False", "If", "Linear", "ModelSelfAttention", "ModelSelfAttentionOutput", "Model_attention_probs_dropout_prob", "Model_attn_chunk_length", "Model_num_chunks_after", "Model_num_chunks_before", "Module", "None", "There", "True", "__init__", "_compute_attn_mask", "_look_adjacent", "_merge_hidden_size_dims", "_retrieve_relevant_hidden_states", "_split_hidden_size_dim", "_split_seq_length_dim_to", "all_head_size", "an", "and", "arange", "are", "assert", "attention_head_size", "attention_mask", "attention_probs", "batch_size", "be", "bias", "bool", "buckets", "buckets_cache", "but", "caching", "cat", "causal_mask", "chunk_length", "class", "config", "def", "del", "device", "dim", "do_standard_self_attention", "dropout", "dtype", "else", "end_dim", "error", "exp", "expand", "f", "flatten", "float16", "forward", "functional", "ge", "half", "head_mask", "hidden_size", "hidden_states", "hidden_states_and_buckets", "if", "indices", "is", "is_decoder", "keepdim", "key", "key_indices", "key_value_hidden_states", "key_vectors", "kwargs", "last", "layer_idx", "len", "logits", "logsumexp", "make", "mask", "mask_value", "mask_value_float16", "mask_value_float32", "matmul", "nn", "not", "np", "num_attention_heads", "num_chunks_after", "num_chunks_before", "numel", "of", "out_vectors", "output_attentions", "p", "pad_token_id", "past_buckets", "past_buckets_states", "past_states", "persistent", "previous_hidden_states", "query", "query_indices", "query_key_dots", "query_key_dots_shape", "query_key_vectors", "query_vectors", "register_buffer", "repeat", "return", "seems", "self", "sequence_length", "set", "shape", "should", "sqrt", "start_dim", "start_position", "states_cache", "staticmethod", "super", "sure", "tensor", "to", "torch", "training", "transpose", "unsqueeze", "use", "use_cache", "value", "value_vectors", "when", "where"], "reformer/modeling_reformer.py:ReformerSelfOutput": ["False", "Linear", "ModelSelfOutput", "Module", "__init__", "all_head_size", "attention_head_size", "bias", "class", "config", "def", "dense", "dropout", "forward", "functional", "hidden_dropout_prob", "hidden_size", "hidden_states", "nn", "num_attention_heads", "p", "return", "self", "super", "training"], "reformer/modeling_reformer.py:ReformerAttention": ["AttentionOutput", "False", "LSHSelfAttention", "LayerNorm", "LocalSelfAttention", "ModelAttention", "ModelSelfOutput", "Module", "None", "NotImplementedError", "Only", "Select", "__init__", "and", "attention_mask", "attention_output", "attention_probs", "attn", "attn_layers", "buckets", "buckets_cache", "but", "cache_position", "class", "config", "def", "elif", "else", "eps", "exist", "f", "forward", "from", "got", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "is", "layer", "layer_id", "layer_idx", "layer_norm", "layer_norm_eps", "len", "local", "lsh", "nn", "not", "num_hashes", "only", "orig_sequence_length", "output", "output_attentions", "past_buckets_states", "raise", "return", "self", "self_attention", "self_attention_outputs", "set", "states", "states_cache", "super", "types", "update", "use_cache"], "reformer/modeling_reformer.py:ReformerFeedForwardDense": ["ACT2FN", "Linear", "ModelFeedForwardDense", "Module", "__init__", "act_fn", "class", "config", "def", "dense", "dropout", "else", "feed_forward_size", "forward", "functional", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "isinstance", "nn", "p", "return", "self", "str", "super", "training"], "reformer/modeling_reformer.py:ReformerFeedForwardOutput": ["Linear", "ModelFeedForwardOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "feed_forward_size", "forward", "functional", "hidden_dropout_prob", "hidden_size", "hidden_states", "nn", "p", "return", "self", "super", "training"], "reformer/modeling_reformer.py:ChunkReformerFeedForward": ["LayerNorm", "ModelFeedForwardDense", "ModelFeedForwardOutput", "ModelModelFeedForward", "Model_size_feed_forward", "Module", "__init__", "apply_Modeling_to_forward", "attention_output", "class", "config", "def", "dense", "eps", "forward", "forward_Model", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "output", "return", "self", "seq_len_dim", "super"], "reformer/modeling_reformer.py:ReformerLayer": ["ChunkModelFeedForward", "False", "If", "ModelAttention", "ModelBackwardOutput", "ModelLayer", "ModelModel", "ModelOutput", "Module", "None", "True", "__init__", "_init_attention_seed", "_init_feed_forward_seed", "and", "assert", "attention", "attention_mask", "attention_probs", "attention_seed", "attn_output", "attn_outputs", "backward", "backward_pass", "buckets", "class", "config", "cuda", "current_device", "def", "default_generators", "del", "detach", "device_idx", "else", "enable_grad", "feed_forward", "feed_forward_seed", "forward", "grad", "grad_attn_output", "grad_hidden_states", "hasattr", "head_mask", "hidden_states", "if", "int", "into", "its", "layer_id", "len", "make", "manual_seed", "maxsize", "mode", "model", "next_attn_output", "nn", "no_grad", "num_hashes", "orig_sequence_length", "output", "output_attentions", "past_buckets_states", "prev_attn_output", "put", "requires_grad", "res_hidden_states", "retain_graph", "return", "seed", "self", "super", "sure", "sys", "the", "to", "torch", "train", "training", "use", "use_cache", "variations", "want", "with", "you"], "reformer/modeling_reformer.py:_ReversibleFunction": ["Function", "ModelBackwardOutput", "None", "True", "_ReversibleFunction", "after", "all_attentions", "all_buckets", "all_hidden_states", "append", "assert", "attention_mask", "attention_probs", "attn_output", "backpropagation", "backward", "backward_pass", "be", "buckets", "cat", "chunk", "class", "ctx", "def", "del", "detach", "dim", "empty", "enumerate", "for", "forward", "grad_attn_output", "grad_hidden_states", "have", "head_mask", "hidden_states", "idx", "if", "in", "is", "layer", "layer_head_mask", "layer_id", "layer_outputs", "layers", "len", "next_attn_output", "num_hashes", "orig_sequence_length", "output", "output_attentions", "output_hidden_states", "past_buckets_states", "prev_attn_output", "return", "save_for_backward", "saved_tensors", "staticmethod", "to", "torch", "use_cache", "zip"], "reformer/modeling_reformer.py:ReformerEncoder": ["False", "LayerNorm", "ModelDynamicCache", "ModelEncoder", "ModelEncoderOutput", "ModelLayer", "Module", "ModuleList", "None", "Passing", "Transformers", "You", "_ReversibleFunction", "__init__", "a", "all_attentions", "all_hidden_states", "an", "and", "apply", "attention_mask", "be", "cat", "class", "config", "def", "deprecated", "dim", "dropout", "e", "elif", "else", "eps", "for", "forward", "from_legacy_cache", "functional", "g", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "i", "if", "in", "instance", "instead", "is", "isinstance", "layer_norm", "layer_norm_eps", "layers", "logger", "next_cache", "nn", "num_hashes", "num_hidden_layers", "of", "orig_sequence_length", "output_attentions", "output_hidden_states", "p", "pass", "past_buckets_states", "past_key_values", "range", "removed", "return", "self", "should", "super", "torch", "training", "tuple", "use_cache", "v4", "warning_once", "will"], "reformer/modeling_reformer.py:ReformerOnlyLMHead": ["False", "Linear", "ModelOnlyLMHead", "Module", "None", "Parameter", "__init__", "_tie_weights", "apply_chunking_to_forward", "bias", "chunk_size_lm_head", "class", "config", "decoder", "def", "device", "else", "forward", "forward_chunk", "hidden_size", "hidden_states", "if", "meta", "nn", "return", "self", "seq_len_dim", "super", "torch", "type", "vocab_size", "zeros"], "reformer/modeling_reformer.py:ReformerPreTrainedModel": ["AxialPositionEmbeddings", "DUMMY_INPUTS", "DUMMY_MASK", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "attention_mask", "axial_norm_std", "base_model_prefix", "bias", "class", "config", "data", "def", "dummy_inputs", "elif", "fill_", "for", "if", "in", "init", "initializer_range", "input_ids", "input_mask", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "property", "return", "self", "std", "tensor", "torch", "weight", "weights", "zero_"], "reformer/modeling_reformer.py:ReformerModelOutput": ["FloatTensor", "LongTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "list", "past_buckets_states", "r", "torch", "tuple"], "reformer/modeling_reformer.py:ReformerModelWithLMHeadOutput": ["FloatTensor", "LongTensor", "ModelModelWithLMHeadOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "past_buckets_states", "r", "torch", "tuple"], "reformer/modeling_reformer.py:ReformerModel": ["If", "Input", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "Please", "Select", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_get_least_common_mult_chunk_len", "_get_min_chunk_len", "_pad_to_mult_of_chunk_length", "_prune_heads", "a", "all_attentions", "all_hidden_states", "and", "arange", "are", "assert", "at", "attention", "attention_mask", "attentions", "attn", "attn_layers", "auto_docstring", "automatically", "batch_size", "be", "bool", "both", "but", "can", "cannot", "cat", "chunk_length", "class", "common", "config", "consider", "def", "device", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "empty", "encoder", "encoder_outputs", "expand", "f", "for", "form", "forward", "from", "full", "get_head_mask", "get_input_embeddings", "got", "has", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "ids", "if", "in", "inference", "input", "input_ids", "input_shape", "inputs_embeds", "int", "is", "is_attention_chunked", "items", "last_hidden_state", "layer", "least", "least_common_mult_chunk_length", "len", "length", "list", "local", "logger", "long", "lsh", "min_chunk_length", "multiple", "must_pad_to_match_chunk_length", "not", "num_hashes", "num_hidden_layers", "of", "one", "ones", "only", "or", "orig_sequence_length", "output_attentions", "output_hidden_states", "pad_attention_mask", "pad_token_id", "padded", "padded_input_ids", "padded_inputs_embeds", "padded_position_ids", "padded_seq_length", "padding", "padding_length", "past_buckets_states", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "sequence", "sequence_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "start_idx_pos_encodings", "super", "the", "time", "to", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "used", "v", "value", "warn_if_padding_and_no_attention_mask", "warning_once", "word_embeddings", "zeros"], "reformer/modeling_reformer.py:ReformerModelWithLMHead": ["CausalLMOutput", "GenerationMixin", "If", "Model", "ModelDynamicCache", "ModelModel", "ModelModelWithLMHead", "ModelModelWithLMHeadOutput", "ModelOnlyLMHead", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Tensor", "True", "Union", "Warning", "__init__", "_reorder_cache", "_tied_weights_keys", "a", "and", "append", "assert", "attention_mask", "attentions", "attn_layers", "auto_docstring", "beam_idx", "bias", "bool", "buckets", "causal", "class", "config", "decoder", "def", "device", "else", "enabled", "f", "for", "forward", "from_legacy_cache", "get_output_embeddings", "head_mask", "hidden_states", "if", "in", "index_select", "input", "input_ids", "inputs_embeds", "int", "is", "is_decoder", "isinstance", "items", "key", "kwargs", "labels", "list", "lm_head", "local", "local_num_chunks_after", "logits", "loss", "loss_function", "lsh", "lsh_num_chunks_after", "make", "mask", "model_inputs", "new_embeddings", "not", "num_hashes", "numel", "or", "output", "output_attentions", "output_hidden_states", "past_buckets_states", "past_key_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "print", "r", "recognized", "reord_buckets", "reord_hidden_states", "reord_past_buckets_states", "return", "return_dict", "self", "sequence_output", "set", "set_output_embeddings", "super", "sure", "that", "to", "torch", "tuple", "use", "use_cache", "use_return_dict", "value", "vocab_size", "want", "weight", "you"], "reformer/modeling_reformer.py:ReformerForMaskedLM": ["CrossEntropyLoss", "False", "If", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyLMHead", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "assert", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "bool", "class", "config", "decoder", "def", "directional", "else", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "is_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "num_hashes", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "torch", "tuple", "use", "use_cache", "use_return_dict", "view", "vocab_size", "want", "weight", "you"], "reformer/modeling_reformer.py:ReformerForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "True", "Union", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "causal", "class", "classification", "classifier", "config", "def", "disable", "dtype", "elif", "else", "for", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "masking", "might", "multi_label_classification", "not", "num_hashes", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence", "sequence_output", "single_label_classification", "squeeze", "super", "to", "torch", "tuple", "use_return_dict", "view", "want", "warning"], "reformer/modeling_reformer.py:ReformerClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch"], "reformer/modeling_reformer.py:ReformerForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "int", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_hashes", "num_labels", "output", "output_attentions", "output_hidden_states", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_cache", "use_return_dict"], "efficientloftr/modeling_efficientloftr.py:KeypointMatchingOutput": ["FloatTensor", "ModelMatchingOutput", "ModelOutput", "Models", "None", "Optional", "attentions", "class", "hidden_states", "matches", "matching_scores", "r", "torch", "tuple"], "efficientloftr/modeling_efficientloftr.py:compute_embeddings": ["Model_embeddings", "Tensor", "cumsum", "def", "device", "dtype", "emb", "embed_height", "embed_width", "hidden_size", "i_indices", "int", "inv_freq", "j_indices", "ones", "return", "torch", "unsqueeze", "zeros"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRRotaryEmbedding": ["False", "LongTensor", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "ROPE_INIT_FUNCTIONS", "Tensor", "_", "__init__", "and", "autocast", "class", "compute_embeddings", "config", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "emb", "embed_height", "embed_width", "enabled", "expand", "feats_height", "feats_width", "float", "forward", "hidden_size", "if", "inv_freq", "inv_freq_expanded", "isinstance", "mps", "nn", "no_grad", "persistent", "position_ids", "q_aggregation_kernel_size", "q_aggregation_stride", "register_buffer", "repeat_interleave", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "tuple", "type", "with", "x"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRConvNormLayer": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "Identity", "ModelConvNormLayer", "Module", "None", "__init__", "activation", "batch_norm_eps", "bias", "class", "config", "conv", "def", "else", "forward", "hidden_state", "if", "in_channels", "is", "kernel_size", "nn", "norm", "out_channels", "padding", "return", "self", "stride", "super"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRRepVGGBlock": ["ACT2FN", "BatchNorm2d", "GradientCheckpointingLayer", "Identity", "ModelConfig", "ModelConvNormLayer", "ModelRepVGGBlock", "None", "Tensor", "__init__", "activation", "activation_function", "and", "block_idx", "class", "config", "conv1", "conv2", "def", "else", "forward", "hidden_states", "identity", "identity_out", "if", "in_channels", "int", "is", "kernel_size", "nn", "not", "out_channels", "padding", "return", "self", "stage_block_in_channels", "stage_block_out_channels", "stage_block_stride", "stage_idx", "stride", "super", "torch"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRRepVGGStage": ["ModelConfig", "ModelRepVGGBlock", "ModelRepVGGStage", "Module", "ModuleList", "Tensor", "__init__", "append", "block", "block_idx", "blocks", "class", "config", "def", "for", "forward", "hidden_states", "in", "int", "nn", "range", "return", "self", "stage_idx", "stage_num_blocks", "super", "torch"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRepVGG": ["ModelConfig", "ModelRepVGGStage", "ModelepVGG", "Module", "ModuleList", "Tensor", "__init__", "append", "class", "config", "def", "for", "forward", "hidden_states", "in", "len", "list", "nn", "outputs", "range", "return", "self", "stage", "stage_idx", "stage_stride", "stages", "super", "torch"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRAggregationLayer": ["Conv2d", "False", "LayerNorm", "MaxPool2d", "ModelAggregationLayer", "ModelConfig", "Module", "None", "Optional", "Tensor", "__init__", "bias", "class", "config", "def", "else", "encoder_hidden_states", "forward", "groups", "hidden_size", "hidden_states", "if", "is", "is_cross_attention", "kernel_size", "kv_aggregation", "kv_aggregation_kernel_size", "kv_aggregation_stride", "kv_states", "nn", "norm", "not", "padding", "permute", "q_aggregation", "q_aggregation_kernel_size", "q_aggregation_stride", "query_states", "return", "self", "stride", "super", "torch", "tuple"], "efficientloftr/modeling_efficientloftr.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "rot_x", "stack", "torch", "x", "x1", "x2"], "efficientloftr/modeling_efficientloftr.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "float", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "efficientloftr/modeling_efficientloftr.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "efficientloftr/modeling_efficientloftr.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "class", "config", "contiguous", "cos", "current_states", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "encoder_hidden_states", "forward", "getattr", "head_dim", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_cross_attention", "k_proj", "key_states", "kwargs", "layer_idx", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "seq_len", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "unsqueeze_dim", "v_proj", "value_states", "view"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRMLP": ["ACT2FN", "False", "LayerNorm", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "__init__", "activation", "bias", "class", "config", "def", "fc1", "fc2", "forward", "hidden_size", "hidden_states", "intermediate_size", "layer_norm", "mlp_activation_function", "nn", "return", "self", "super", "torch"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRAggregatedAttention": ["False", "ModelAggregatedAttention", "ModelAggregationLayer", "ModelAttention", "ModelConfig", "ModelMLP", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "aggregated_encoder_hidden_states", "aggregated_h", "aggregated_hidden_states", "aggregated_w", "aggregation", "align_corners", "attention", "attn_output", "batch_size", "bilinear", "cat", "class", "config", "def", "dim", "embed_dim", "encoder_hidden_states", "forward", "functional", "hidden_states", "int", "intermediate_states", "interpolate", "kwargs", "layer_idx", "mlp", "mode", "nn", "output_states", "permute", "position_embeddings", "q_aggregation_kernel_size", "reshape", "return", "scale_factor", "self", "shape", "super", "torch", "tuple"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRLocalFeatureTransformerLayer": ["GradientCheckpointingLayer", "ModelAggregatedAttention", "ModelConfig", "ModelLocalFeatureTransformerLayer", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "batch_size", "class", "config", "cross_attention", "def", "dim", "embed_dim", "features_0", "features_1", "forward", "height", "hidden_states", "int", "kwargs", "layer_idx", "position_embeddings", "reshape", "return", "self", "self_attention", "shape", "stack", "super", "torch", "tuple", "width"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRLocalFeatureTransformer": ["ModelConfig", "ModelLocalFeatureTransformer", "ModelLocalFeatureTransformerLayer", "Module", "ModuleList", "Tensor", "TransformersKwargs", "Unpack", "__init__", "class", "config", "def", "for", "forward", "hidden_states", "i", "in", "kwargs", "layer", "layer_idx", "layers", "nn", "num_attention_layers", "position_embeddings", "range", "return", "self", "super", "torch", "tuple"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTROutConvBlock": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "ModelConfig", "ModelOutConvBlock", "Module", "Tensor", "__init__", "activation", "align_corners", "batch_norm", "bias", "bilinear", "class", "config", "def", "forward", "functional", "hidden_size", "hidden_states", "int", "intermediate_size", "interpolate", "kernel_size", "mlp_activation_function", "mode", "nn", "out_conv1", "out_conv2", "out_conv3", "padding", "residual_states", "return", "scale_factor", "self", "stride", "super", "torch"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRFineFusionLayer": ["Conv2d", "False", "ModelConfig", "ModelFineFusionLayer", "ModelOutConvBlock", "Module", "ModuleList", "Tensor", "_", "__init__", "align_corners", "append", "batch_size", "bias", "bilinear", "class", "coarse_features", "coarse_height", "coarse_width", "config", "def", "embed_dim", "enumerate", "fine_embed_dim", "fine_features", "fine_features_0", "fine_features_1", "fine_fusion_dims", "fine_height", "fine_kernel_size", "fine_width", "for", "forward", "forward_pyramid", "functional", "hidden_states", "i", "in", "int", "interpolate", "kernel_size", "layer", "len", "list", "mode", "nn", "out_conv", "out_conv_layers", "padding", "permute", "range", "reshape", "residual_features", "residual_states", "return", "reversed", "scale_factor", "self", "seq_len", "shape", "stride", "super", "torch", "tuple", "unfold"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRPreTrainedModel": ["BatchNorm2d", "Conv1d", "Conv2d", "FloatTensor", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelPreTrainedModel", "ModelRepVGGBlock", "Module", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "data", "def", "elif", "extract_one_channel_pixel_values", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "return", "self", "std", "supports_gradient_checkpointing", "torch", "weight", "zero_"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRModel": ["BackboneOutput", "FloatTensor", "Input", "LongTensor", "Model", "ModelConfig", "ModelLocalFeatureTransformer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModelepVGG", "None", "Optional", "TransformersKwargs", "Unpack", "ValueError", "_", "__init__", "a", "auto_docstring", "backbone", "batch_size", "be", "channels", "check_model_inputs", "class", "coarse_embed_dim", "coarse_features", "coarse_height", "coarse_width", "config", "cos", "def", "expand", "extract_one_channel_pixel_values", "feature_maps", "features", "forward", "height", "if", "is", "kwargs", "labels", "local_feature_transformer", "must", "ndim", "no", "not", "num_channels", "of", "or", "pixel_values", "position_embeddings", "post_init", "provided", "r", "raise", "reshape", "residual_features", "return", "rotary_emb", "self", "shape", "should", "sin", "size", "super", "tensor", "torch", "trainable", "tuple", "width"], "efficientloftr/modeling_efficientloftr.py:mask_border": ["Model_border", "Tensor", "Union", "bool", "border_margin", "def", "float", "if", "int", "return", "tensor", "torch", "value"], "efficientloftr/modeling_efficientloftr.py:create_meshgrid": ["False", "Model_meshgrid", "None", "Optional", "Tensor", "Union", "bool", "def", "device", "dim", "dtype", "grid", "height", "if", "ij", "indexing", "int", "linspace", "meshgrid", "normalized_coordinates", "permute", "return", "stack", "torch", "unsqueeze", "width", "xs", "ys"], "efficientloftr/modeling_efficientloftr.py:spatial_expectation2d": ["Model_expectation2d", "Tensor", "True", "batch_size", "bool", "cat", "create_meshgrid", "def", "device", "dtype", "embed_dim", "expected_x", "expected_y", "grid", "height", "input", "input_flat", "keepdim", "normalized_coordinates", "output", "pos_x", "pos_y", "r", "reshape", "return", "shape", "sum", "to", "torch", "view", "width"], "efficientloftr/modeling_efficientloftr.py:EfficientLoFTRForKeypointMatching": ["BackboneOutput", "False", "FloatTensor", "KeypointMatchingOutput", "LongTensor", "Model", "ModelConfig", "ModelFineFusionLayer", "ModelForKeypointMatching", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SuperGlue", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "_coarse_matching", "_fine_matching", "_get_first_stage_fine_matching", "_get_matches_from_scores", "_get_second_stage_fine_matching", "arange", "attentions", "auto_docstring", "batch_indices", "batch_size", "be", "can_return_tuple", "cat", "channels", "class", "coarse_embed_dim", "coarse_features", "coarse_features_0", "coarse_features_1", "coarse_height", "coarse_keypoints", "coarse_matched_indices", "coarse_matched_keypoints", "coarse_matching_border_removal", "coarse_matching_scores", "coarse_matching_skip_softmax", "coarse_matching_temperature", "coarse_matching_threshold", "coarse_scale", "coarse_width", "confidence", "config", "create_meshgrid", "def", "delta", "delta_0", "delta_1", "device", "dim", "dtype", "else", "embed_dim", "expand", "feature_maps", "features", "fine_confidence", "fine_coordinates", "fine_coordinates_normalized", "fine_embed_dim", "fine_features_0", "fine_features_1", "fine_height", "fine_indices", "fine_kernel_size", "fine_matches", "fine_matches_0", "fine_matches_1", "fine_matching_regress_temperature", "fine_matching_slice_dim", "fine_scale", "fine_window_size", "first_stage_fine_confidence", "float", "forward", "functional", "gather", "grid", "heatmap", "height", "height0", "height1", "hidden_size", "hidden_states", "if", "indices", "indices_0", "indices_1", "indices_1_i", "indices_1_j", "int", "is", "keepdim", "keypoints", "kwargs", "labels", "long", "mask", "mask_border", "masked_scores", "matched_indices", "matches", "matches_indices", "matching_indices", "matching_keypoints", "matching_scores", "matching_scores_0", "matching_scores_1", "max", "max_0", "max_1", "max_indices_0", "max_indices_1", "model_outputs", "nn", "no", "normalized_coordinates", "not", "num_keypoints", "permute", "pixel_values", "post_init", "provided", "r", "raise", "refinement_layer", "reshape", "residual_features", "return", "scores", "second_stage_fine_confidence", "self", "shape", "should", "similarity", "softmax", "spatial_expectation2d", "split", "split_fine_features_0", "split_fine_features_1", "squeeze", "stack", "super", "to", "torch", "torch_int", "trainable", "transpose", "tuple", "unsqueeze", "values", "view", "where", "width", "width0", "width1"], "timesfm/modeling_timesfm.py:TimesFmOutput": ["BaseModelOutput", "ModelOutput", "None", "Optional", "Tensor", "class", "loc", "r", "scale", "torch"], "timesfm/modeling_timesfm.py:TimesFmOutputForPrediction": ["BaseModelOutput", "ModelOutputForPrediction", "None", "Optional", "Tensor", "Union", "class", "float", "full_predictions", "loss", "mean_predictions", "r", "torch"], "timesfm/modeling_timesfm.py:TimesFmMLP": ["F", "LayerNorm", "Linear", "ModelConfig", "ModelMLP", "Module", "None", "__init__", "class", "config", "def", "down_proj", "eps", "forward", "gate", "gate_inp", "gate_proj", "hidden_size", "if", "intermediate_size", "is", "layer_norm", "nn", "normalized_shape", "not", "outputs", "paddings", "relu", "return", "self", "super", "x"], "timesfm/modeling_timesfm.py:TimesFmResidualBlock": ["Linear", "ModelResidualBlock", "Module", "SiLU", "__init__", "activation", "class", "def", "forward", "hidden", "hidden_dims", "input_dims", "input_layer", "nn", "output", "output_dims", "output_layer", "residual", "residual_layer", "return", "self", "super", "x"], "timesfm/modeling_timesfm.py:TimesFmRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "timesfm/modeling_timesfm.py:TimesFmPositionalEmbedding": ["Either", "F", "ModelConfig", "ModelPositionalEmbedding", "Module", "None", "ValueError", "__init__", "and", "arange", "be", "cat", "class", "config", "cos", "def", "device", "dim", "dimensional", "dtype", "elif", "embedding_dims", "exp", "f", "float", "float32", "forward", "got", "hidden_size", "if", "inv_Modelcales", "is", "log", "log_Modelcale_increment", "math", "max", "max_Modelcale", "min_Modelcale", "must", "ndim", "nn", "num_Modelcales", "or", "pad", "position", "provided", "raise", "register_buffer", "return", "scaled_time", "self", "seq_length", "shape", "signal", "sin", "super", "torch", "unsqueeze", "view"], "timesfm/modeling_timesfm.py:simple_eager_attention_forward": ["Model_eager_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key_states", "kwargs", "matmul", "module", "nn", "not", "p", "query_states", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value_states"], "timesfm/modeling_timesfm.py:TimesFmAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "F", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Parameter", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "_scale_query", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "class", "config", "contiguous", "def", "dropout", "eager", "else", "empty", "forward", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is_causal", "k_proj", "key_states", "kv_size", "kwargs", "layer_idx", "math", "mul", "nn", "not", "num_attention_heads", "num_heads", "o_proj", "q_proj", "q_size", "query", "query_states", "reshape", "return", "scale", "scaling", "self", "shape", "simple_eager_attention_forward", "softplus", "sqrt", "super", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "timesfm/modeling_timesfm.py:TimesFmDecoderLayer": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "Module", "Optional", "Tensor", "__init__", "attention_mask", "bool", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "layer_idx", "mlp", "nn", "output_attentions", "paddings", "residual", "return", "rms_norm_eps", "scores", "self", "self_attn", "super", "torch", "tuple"], "timesfm/modeling_timesfm.py:TimesFmPreTrainedModel": ["Model", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "class", "config", "def", "if", "init", "isinstance", "main_input_name", "module", "nn", "ones_", "past_values", "scaling", "self", "super"], "timesfm/modeling_timesfm.py:TimesFmModel": ["BoolTensor", "Embedding", "False", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelOutput", "ModelPositionalEmbedding", "ModelPreTrainedModel", "ModelResidualBlock", "ModuleList", "None", "Optional", "Tensor", "True", "_Model_masked_mean_std", "_Model_shift_padded_seq", "__init__", "_forward_transform", "_get_patch_index", "_prepare_4d_attention_mask", "abs", "all_attentions", "all_hidden_states", "any", "append", "arange", "argmax", "arr", "attention_mask", "attentions", "auto_docstring", "batch_size", "bidxs", "bool", "bsize", "can_return_tuple", "cat", "causal_mask", "class", "concat", "concat_inputs", "config", "def", "device", "diagonal", "dim", "dtype", "else", "embedding_dim", "expand", "f_emb", "feature_dim", "finfo", "for", "forward", "freq", "freq_emb", "freq_size", "gather", "hidden_dims", "hidden_size", "hidden_states", "idx_range", "if", "iinfo", "in", "indices", "input_dims", "input_ff_layer", "inputs", "int", "int32", "intermediate_size", "is", "is_causal", "is_floating_point", "last_hidden_state", "layer", "layer_idx", "layers", "loc", "mask", "masked_mean", "masked_squared_sum", "masked_std", "masked_sum", "masked_var", "min", "min_value", "minimum", "model_input", "mu", "new_mask", "nn", "not", "num_embeddings", "num_hidden_layers", "num_seq", "num_valid_elements", "ones", "output_attentions", "output_dims", "output_hidden_states", "outputs", "pad", "pad_sum", "pad_val", "padding", "paddings", "past_values", "past_values_padding", "patch_indices", "patch_length", "patched_inputs", "patched_padding", "patched_pads", "pos_emb", "position_emb", "post_init", "r", "range", "return", "row_sum", "scale", "scores", "self", "seq", "sequence_length", "shape", "shifted_idx", "shifted_seq", "sigma", "sqrt", "staticmethod", "stats", "sum", "super", "tensor", "to", "tolerance", "torch", "triu", "tuple", "use_positional_embedding", "view", "where"], "timesfm/modeling_timesfm.py:TimesFmModelForPrediction": ["Default", "F", "False", "Length", "ModelConfig", "ModelModel", "ModelModelForPrediction", "ModelOutputForPrediction", "ModelPreTrainedModel", "ModelResidualBlock", "No", "None", "Optional", "Sequence", "Tensor", "Union", "ValueError", "_", "_Model_moving_average", "__init__", "_postprocess_output", "_preprocess", "_quantile_loss", "and", "append", "arr", "arr_padded", "attentions", "auto_docstring", "axis", "b", "bool", "can_return_tuple", "cat", "class", "concatenate", "config", "constant", "context_len", "context_length", "conv1d", "current_padding", "decoder", "decoder_output", "def", "device", "dim", "dtype", "elif", "else", "enumerate", "errors", "extend", "f", "fcontext_len", "final_out", "for", "forecast_context_len", "forward", "fprop_outputs", "freq", "frequency", "full_outputs", "full_predictions", "future_values", "hidden_dims", "hidden_size", "hidden_states", "high", "horizon_ff_layer", "horizon_len", "horizon_length", "i", "if", "in", "info", "inp_freq", "inp_min", "input", "input_dims", "input_len", "input_padding", "input_ts", "inputs", "int", "int32", "intermediate_size", "is", "kernel", "last_hidden_state", "len", "length", "list", "loc", "logger", "loss", "losses", "match", "max", "maximum", "mean", "mean_outputs", "mean_predictions", "min", "model_output", "mse_loss", "mu", "must", "n", "new_freqs", "new_full_ts", "new_inputs", "new_ts", "not", "num_decode_patches", "num_front_pad", "of", "ones", "output_attentions", "output_dims", "output_hidden_states", "output_patch_len", "output_ts", "pad", "padding", "paddings", "past_values", "past_values_padding", "patch_length", "post_init", "predictions", "provided", "q", "quantile_loss", "quantiles", "r", "raise", "range", "reshape", "return", "return_forecast_on_context", "scale", "self", "shape", "sigma", "size", "smoothed_arr", "squeeze", "stack", "staticmethod", "stats", "step_index", "super", "targets", "tensor", "to", "torch", "truncate_negative", "ts", "tuple", "via", "view", "window_size", "zeros"], "depth_anything/modeling_depth_anything.py:DepthAnythingReassembleLayer": ["Conv2d", "ConvTranspose2d", "Identity", "ModelReassembleLayer", "Module", "__init__", "channels", "class", "config", "def", "elif", "factor", "forward", "hidden_state", "if", "in_channels", "int", "kernel_size", "nn", "out_channels", "padding", "projection", "reassemble_hidden_size", "resize", "return", "self", "stride", "super"], "depth_anything/modeling_depth_anything.py:DepthAnythingReassembleStage": ["ModelReassembleLayer", "ModelReassembleStage", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "append", "batch_size", "channels", "class", "config", "contiguous", "def", "enumerate", "factor", "for", "forward", "hidden_state", "hidden_states", "i", "in", "layers", "list", "neck_hidden_sizes", "nn", "num_channels", "out", "patch_height", "patch_width", "permute", "reassemble_factors", "reshape", "return", "self", "shape", "super", "torch", "zip"], "depth_anything/modeling_depth_anything.py:DepthAnythingPreActResidualLayer": ["Conv2d", "ModelPreActResidualLayer", "Module", "ReLU", "Tensor", "True", "__init__", "activation1", "activation2", "bias", "class", "config", "convolution1", "convolution2", "def", "forward", "fusion_hidden_size", "hidden_state", "kernel_size", "nn", "padding", "residual", "return", "self", "stride", "super", "torch"], "depth_anything/modeling_depth_anything.py:DepthAnythingFeatureFusionLayer": ["Conv2d", "False", "ModelFeatureFusionLayer", "ModelPreActResidualLayer", "Module", "None", "True", "__init__", "align_corners", "bias", "bilinear", "class", "config", "def", "else", "forward", "functional", "fusion_hidden_size", "hidden_state", "if", "interpolate", "is", "kernel_size", "mode", "modifier", "nn", "not", "projection", "residual", "residual_layer1", "residual_layer2", "return", "scale_factor", "self", "shape", "size", "super"], "depth_anything/modeling_depth_anything.py:DepthAnythingFeatureFusionStage": ["ModelConfig", "ModelFeatureFusionLayer", "ModelFeatureFusionStage", "Module", "ModuleList", "None", "_", "__init__", "append", "class", "config", "def", "else", "enumerate", "for", "forward", "fused_hidden_state", "fused_hidden_states", "hidden_state", "hidden_states", "idx", "if", "in", "is", "layer", "layers", "len", "neck_hidden_sizes", "nn", "range", "return", "self", "shape", "size", "super", "zip"], "depth_anything/modeling_depth_anything.py:DepthAnythingPreTrainedModel": ["Conv2d", "ConvTranspose2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "depth_anything/modeling_depth_anything.py:DepthAnythingNeck": ["Conv2d", "False", "ModelFeatureFusionStage", "ModelNeck", "ModelReassembleStage", "Module", "ModuleList", "None", "Tensor", "The", "TypeError", "ValueError", "__init__", "a", "append", "be", "bias", "channel", "class", "config", "convs", "def", "enumerate", "equal", "feature", "features", "for", "forward", "fusion_hidden_size", "fusion_stage", "hidden", "hidden_states", "i", "if", "in", "isinstance", "kernel_size", "len", "list", "neck", "neck_hidden_sizes", "nn", "not", "number", "of", "or", "output", "padding", "patch_height", "patch_width", "raise", "reassemble_stage", "return", "self", "should", "sizes", "states", "super", "tensors", "the", "to", "torch", "tuple"], "depth_anything/modeling_depth_anything.py:DepthAnythingDepthEstimationHead": ["Conv2d", "Model", "ModelModelEstimationHead", "Model_estimation_type", "Module", "ReLU", "Sigmoid", "Tensor", "True", "Unknown", "ValueError", "__init__", "activation1", "activation2", "align_corners", "bilinear", "class", "config", "conv1", "conv2", "conv3", "def", "dim", "elif", "else", "estimation", "f", "features", "forward", "functional", "fusion_hidden_size", "head_hidden_size", "head_in_index", "hidden_states", "if", "int", "interpolate", "kernel_size", "list", "max_Model", "metric", "mode", "nn", "padding", "patch_height", "patch_size", "patch_width", "predicted_Model", "raise", "relative", "return", "self", "squeeze", "stride", "super", "torch", "type"], "depth_anything/modeling_depth_anything.py:DepthAnythingForDepthEstimation": ["DPTViTEmbeddings", "FloatTensor", "LongTensor", "ModelEstimatorOutput", "ModelForModelEstimation", "ModelModelEstimationHead", "ModelNeck", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Tensor", "Training", "Union", "_", "__init__", "_no_split_modules", "attentions", "auto_docstring", "backbone", "bool", "class", "config", "def", "else", "feature_maps", "forward", "forward_with_filtered_kwargs", "head", "height", "hidden_states", "if", "implemented", "is", "labels", "load_backbone", "loss", "neck", "not", "output", "output_attentions", "output_hidden_states", "outputs", "patch_height", "patch_size", "patch_width", "pixel_values", "post_init", "predicted_Model", "r", "raise", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "use_return_dict", "width", "yet"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "glm4v_moe/modeling_glm4v_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "glm4v_moe/modeling_glm4v_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "glm4v_moe/modeling_glm4v_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "glm4v_moe/modeling_glm4v_moe.py:apply_multimodal_rotary_pos_emb": ["Model_multimodal_rotary_pos_emb", "cat", "cos", "def", "dim", "enumerate", "for", "i", "in", "k", "k_embed", "k_pass", "k_rot", "m", "mrope_section", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "split", "torch", "unsqueeze", "unsqueeze_dim"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelTextAttention", "ModelTextConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_multimodal_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "mrope_section", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "rope_scaling", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextTopkRouter": ["F", "False", "ModelTextConfig", "ModelTextTopkRouter", "Module", "Parameter", "True", "__init__", "bool", "class", "config", "def", "denominator", "dim", "dtype", "e_score_correction_bias", "empty", "expand", "float32", "forward", "gather", "get_topk_indices", "group_idx", "group_mask", "group_scores", "hidden_size", "hidden_states", "if", "k", "keepdim", "linear", "masked_fill", "n_group", "n_routed_experts", "nn", "no_grad", "norm_topk_prob", "num_experts_per_tok", "register_buffer", "reshape", "return", "routed_scaling_factor", "router_logits", "scatter_", "score_mask", "scores", "scores_for_choice", "self", "sigmoid", "sorted", "sum", "super", "top_k", "topk", "topk_group", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight", "zeros", "zeros_like"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextMoE": ["ModelTextConfig", "ModelTextMLP", "ModelTextMoE", "ModelTextTopkRouter", "Module", "ModuleList", "Tensor", "_", "__init__", "class", "config", "def", "dtype", "expert", "expert_idx", "expert_input", "expert_mask", "expert_output", "expert_weights", "experts", "final_hidden_states", "for", "forward", "functional", "gate", "hidden_states", "if", "in", "index_add_", "intermediate_size", "len", "mask", "moe", "moe_intermediate_size", "n_routed_experts", "n_shared_experts", "nn", "num_classes", "numel", "one_hot", "orig_shape", "permute", "r", "range", "residuals", "return", "self", "shape", "shared_experts", "super", "token_indices", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight_indices", "weighted_output", "where", "zeros_like"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextMLP": ["ACT2FN", "False", "Linear", "ModelTextMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "return", "self", "super", "up_proj", "x"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextRMSNorm": ["ModelTextRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextMLP", "ModelTextMoE", "ModelTextRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "first_k_dense_replace", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoePreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextDecoderLayer", "ModelTextTopkRouter", "ModelVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeisionMlp": ["ACT2FN", "False", "Linear", "ModelisionMlp", "Module", "__init__", "act_fn", "bias", "bool", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "nn", "out_hidden_size", "return", "self", "super", "up_proj"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionPatchEmbed": ["Conv3d", "ModelVisionConfig", "ModelVisionPatchEmbed", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dtype", "embed_dim", "forward", "hidden_size", "hidden_states", "in_channels", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionRotaryEmbedding": ["False", "ModelVisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionPatchMerger": ["ACT2FN", "False", "GELU", "LayerNorm", "Linear", "ModelVisionPatchMerger", "Module", "None", "Tensor", "__init__", "act1", "act_fn", "bias", "bool", "class", "context_dim", "def", "dim", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_state", "int", "nn", "post_projection_norm", "proj", "return", "self", "str", "super", "torch", "up_proj"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionEmbeddings": ["Embedding", "F", "False", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "__init__", "adapted_pos_embed", "adapted_pos_embed_fp32", "align_corners", "arange", "bicubic", "border", "cat", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "empty", "expand", "float32", "for", "forward", "grid", "grid_sample", "h_coords", "hidden_size", "i", "if", "image_shapes", "image_size", "in", "int", "interpolated_embed_fp32", "isinstance", "len", "lengths", "list", "long", "mode", "nn", "norm_h", "norm_w", "not", "num_patches", "num_positions", "orig_size", "orig_size_sq", "padding_mode", "patch_size", "permute", "persistent", "pos_embed_2d", "pos_embed_weight", "position_embedding", "position_ids", "range", "register_buffer", "repeat", "return", "self", "shape", "squeeze", "stack", "super", "target_h", "target_w", "tensor", "to", "torch", "total_seq", "unsqueeze", "view", "w_coords", "weight"], "glm4v_moe/modeling_glm4v_moe.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionBlock": ["False", "GradientCheckpointingLayer", "ModelRMSNorm", "ModelVisionAttention", "ModelVisionBlock", "ModelisionMlp", "None", "Optional", "Tensor", "__init__", "attn", "bias", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "norm1", "norm2", "position_embeddings", "return", "rms_norm_eps", "rotary_pos_emb", "self", "super", "torch", "tuple"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeModelOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeVisionModel": ["Conv2d", "F", "False", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "ModelVisionPatchEmbed", "ModelVisionPatchMerger", "ModelVisionRotaryEmbedding", "ModuleList", "None", "Tensor", "_", "__init__", "_no_split_modules", "append", "arange", "blk", "blocks", "cat", "class", "config", "context_dim", "cos", "cu_seqlens", "cumsum", "def", "depth", "dim", "downsample", "dtype", "else", "emb", "embeddings", "eps", "expand", "flatten", "for", "forward", "gradient_checkpointing", "grid_thw", "h", "head_dim", "hidden_act", "hidden_size", "hidden_states", "hpos_ids", "if", "image_type_ids", "in", "in_channels", "int32", "intermediate_size", "is_tracing", "jit", "kernel_size", "max", "max_grid_size", "merger", "nn", "num_heads", "out_channels", "out_hidden_size", "pad", "patch_embed", "patch_size", "permute", "pos_ids", "position_embeddings", "post_conv_layernorm", "post_init", "post_layernorm", "range", "repeat", "repeat_interleave", "reshape", "return", "rms_norm_eps", "rot_pos_emb", "rotary_pos_emb", "rotary_pos_emb_full", "self", "seqlens", "shape", "sin", "spatial_merge_size", "stack", "stride", "super", "t", "tolist", "torch", "unsqueeze", "value", "view", "w", "wpos_ids"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelRMSNorm", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "dim", "elif", "else", "embed_tokens", "eps", "exactly", "expand", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "tuple", "use_cache", "view", "vocab_size"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeModel": ["Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelTextDecoderLayer", "ModelTextModel", "ModelVisionBlock", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "_from_config", "_no_split_modules", "accepts_loss_kwargs", "add", "all", "and", "append", "arange", "attention_mask", "attention_mask_tensor", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "cache_position", "can_return_tuple", "cat", "class", "config", "cumsum", "decoder", "def", "delta", "device", "diagonal", "dict", "dim", "dim1", "dim2", "do", "dtype", "elif", "else", "end_idx", "end_index", "enumerate", "exactly", "expand", "expand_as", "f", "features", "finfo", "flatten", "flattened_video_grid_thw", "for", "forward", "full_attention", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_seq_length", "get_video_features", "grid_thw", "group", "groupby", "h", "h_index", "hidden_states", "i", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_index", "image_mask", "image_token_id", "in", "input_ids", "input_token_type", "input_tokens", "input_type_group", "inputs_embeds", "int", "is", "is_floating_point", "is_torchdynamo_compiling", "isinstance", "item", "itertools", "keepdim", "key", "kwargs", "lambda", "language_model", "last_hidden_state", "len", "list", "llm_grid_h", "llm_grid_t", "llm_grid_w", "llm_pos_ids_list", "llm_positions", "long", "masked_fill_", "masked_scatter", "match", "max", "max_position_ids", "min", "modality_type", "mrope_position_deltas", "must", "n_image_tokens", "n_video_tokens", "ndim", "not", "numel", "of", "one", "ones", "ones_like", "or", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prefill_compiled_stage", "prefill_noncompiled_stage", "prod", "r", "raise", "range", "repeat", "repeat_interleave", "repeated_row", "reshape", "return", "rope_deltas", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "special_video_mask", "specify", "split", "split_sizes", "st_idx", "stack", "start_idx", "start_index", "sum", "super", "t", "t_idx", "t_index", "temp_frames_hw", "tensor", "text", "text_config", "text_len", "to", "token", "tokens", "tolist", "torch", "total_input_ids", "tuple", "type", "unsqueeze", "value", "video", "video_check_flg", "video_embeds", "video_end_token_id", "video_features", "video_frame_num", "video_grid_thw", "video_group_index", "video_index", "video_mask", "video_start_token_id", "video_token_id", "view", "vision_config", "visual", "w", "w_index", "x", "zeros"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "glm4v_moe/modeling_glm4v_moe.py:Glm4vMoeForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "If", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_expand_dict_for_generation", "_expand_dict_for_generation_visual", "_expand_inputs_for_generation", "_get_image_nums_and_video_nums", "_repeat_interleave_samples", "_tied_weights_keys", "accepts_loss_kwargs", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "cumsum", "decoder", "def", "defined", "device", "dict", "dict_to_expand", "dim", "dtype", "elif", "else", "encoder_outputs", "expand_size", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "get_video_features", "hidden_size", "hidden_states", "if", "image_counts", "image_grid_thw", "image_nums", "image_start_token_id", "in", "input_ids", "inputs_embeds", "inside_video", "int", "is", "is_encoder_decoder", "is_image", "is_video_end", "is_video_start", "isinstance", "key", "kwargs", "labels", "language_model", "lengths", "list", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "make", "model", "model_inputs", "model_kwargs", "nn", "not", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "prod", "property", "r", "raise", "repeat", "repeat_args", "repeat_interleave", "repeat_times", "result", "return", "rope_deltas", "sample", "samples", "second_per_grid_ts", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "split", "standalone_images", "str", "sum", "super", "sure", "tensor", "text_config", "that", "torch", "tuple", "use_cache", "value", "video_counts", "video_end_token_id", "video_grid_thw", "video_level", "video_nums", "video_start_token_id", "visual", "visual_keys", "vocab_size", "weight", "x"], "timm_backbone/modeling_timm_backbone.py:TimmBackbone": ["BackboneMixin", "BackboneOutput", "Cannot", "False", "FloatTensor", "Model", "ModelConfig", "None", "Optional", "Please", "PreTrainedModel", "Tensor", "True", "Union", "ValueError", "__init__", "_all_layers", "_backbone", "_from_config", "_init_backbone", "_init_weights", "_return_layers", "a", "and", "at", "attentions", "backbone", "backbones", "be", "bool", "by", "class", "classmethod", "cls", "config", "create_model", "def", "else", "enumerate", "feature_info", "feature_maps", "features_only", "for", "forward", "freeze_batch_norm_2d", "from_pretrained", "get_dicts", "getattr", "hasattr", "hidden_states", "i", "if", "in", "in_chans", "index", "info", "instead", "is", "it", "kwargs", "layer", "main_input_name", "model", "model_args", "module", "moment", "must", "name", "not", "num_channels", "or", "out_features", "out_indices", "output", "output_attentions", "output_hidden_states", "pass", "pixel_values", "pop", "pretrained", "pretrained_model_name_or_path", "raise", "requires_backends", "return", "return_dict", "return_layers", "self", "set", "str", "super", "supported", "supports_gradient_checkpointing", "the", "to", "torch", "tuple", "unfreeze_batch_norm_2d", "use", "use_Model", "use_pretrained_backbone", "use_return_dict", "utils", "vision"], "dpt/modeling_dpt.py:BaseModelOutputWithIntermediateActivations": ["FloatTensor", "ModelModelOutputWithIntermediateActivations", "ModelOutput", "None", "Optional", "class", "intermediate_activations", "last_hidden_states", "r", "torch", "tuple"], "dpt/modeling_dpt.py:BaseModelOutputWithPoolingAndIntermediateActivations": ["FloatTensor", "ModelModelOutputWithPoolingAndIntermediateActivations", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "intermediate_activations", "last_hidden_state", "pooler_output", "r", "torch", "tuple"], "dpt/modeling_dpt.py:DPTViTHybridEmbeddings": ["BaseModelOutputWithIntermediateActivations", "Conv2d", "Expected", "False", "Input", "Iterable", "Make", "Model", "ModelConfig", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_resize_pos_embed", "abc", "backbone", "backbone_featmap_shape", "backbone_output", "batch_size", "bilinear", "bool", "cat", "channel", "channels", "class", "cls_token", "cls_tokens", "collections", "config", "configuration", "def", "dim", "dimension", "doesn", "else", "embeddings", "expand", "f", "feat_map_shape", "feature_dim", "feature_maps", "feature_size", "features", "flatten", "for", "forward", "functional", "got", "grid_size_height", "grid_size_width", "have", "height", "hidden_size", "if", "image", "image_size", "in", "index", "int", "intermediate_activations", "interpolate", "interpolate_pos_encoding", "is", "isinstance", "kernel_size", "last_hidden_states", "len", "load_backbone", "match", "mode", "model", "nn", "not", "num_channels", "num_patches", "of", "old_grid_size", "one", "or", "output", "output_hidden_states", "patch_size", "permute", "pixel", "pixel_values", "posemb", "posemb_grid", "posemb_tok", "position_embeddings", "projection", "raise", "reshape", "residual_feature_map_index", "return", "self", "set", "shape", "size", "start_index", "super", "sure", "t", "that", "the", "to", "torch", "torch_int", "transpose", "tuple", "values", "width", "with", "zeros"], "dpt/modeling_dpt.py:DPTViTEmbeddings": ["BaseModelOutputWithIntermediateActivations", "Dropout", "Model", "ModelViTPatchEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "_resize_pos_embed", "batch_size", "bilinear", "cat", "class", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "embeddings", "expand", "forward", "functional", "grid_size_height", "grid_size_width", "height", "hidden_dropout_prob", "hidden_size", "interpolate", "last_hidden_states", "mode", "nn", "num_channels", "num_patches", "old_grid_size", "patch_embeddings", "patch_size", "permute", "pixel_values", "posemb", "posemb_grid", "posemb_tok", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "start_index", "super", "torch", "torch_int", "width", "zeros"], "dpt/modeling_dpt.py:DPTViTPatchEmbeddings": ["Conv2d", "Iterable", "Make", "Model", "ModelConfig", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "flatten", "forward", "height", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "width", "with"], "dpt/modeling_dpt.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "dpt/modeling_dpt.py:DPTSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "dpt/modeling_dpt.py:DPTViTSelfOutput": ["Dropout", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "dpt/modeling_dpt.py:DPTViTAttention": ["Model", "ModelConfig", "ModelSelfAttention", "ModelViTSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "dpt/modeling_dpt.py:DPTViTIntermediate": ["ACT2FN", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "dpt/modeling_dpt.py:DPTViTOutput": ["Dropout", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "dpt/modeling_dpt.py:DPTViTLayer": ["GradientCheckpointingLayer", "LayerNorm", "Model", "ModelConfig", "ModelViTAttention", "ModelViTIntermediate", "ModelViTOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "dpt/modeling_dpt.py:DPTViTEncoder": ["BaseModelOutput", "False", "Model", "ModelConfig", "ModelViTLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_hidden_states", "append", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "output_hidden_states", "range", "return", "self", "super", "torch", "tuple"], "dpt/modeling_dpt.py:DPTReassembleStage": ["ACT2FN", "Hybrid", "Identity", "Linear", "Model", "ModelReassembleLayer", "Module", "ModuleList", "None", "Readout", "Sequential", "Tensor", "ValueError", "_", "__init__", "_get_backbone_hidden_size", "_init_reassemble_Model", "_init_reassemble_Model_hybrid", "add", "and", "append", "batch_size", "cat", "channels", "class", "cls_token", "config", "contiguous", "def", "elif", "else", "enumerate", "expand_as", "f", "factor", "feature_shape", "flatten", "for", "forward", "hidden_act", "hidden_size", "hidden_state", "hidden_states", "i", "if", "in", "is", "is_hybrid", "layers", "len", "list", "neck_hidden_sizes", "neck_ignore_stages", "nn", "not", "num_channels", "out", "patch_height", "patch_width", "permute", "project", "r", "raise", "range", "readout", "readout_projects", "readout_type", "reassemble_factors", "reshape", "return", "self", "sequence_length", "shape", "size", "super", "supported", "torch", "torch_int", "type", "unsqueeze", "zip"], "dpt/modeling_dpt.py:_get_backbone_hidden_size": ["False", "None", "_get_backbone_hidden_size", "and", "backbone_config", "config", "def", "else", "hidden_size", "if", "is", "is_hybrid", "not", "return"], "dpt/modeling_dpt.py:DPTReassembleLayer": ["Conv2d", "ConvTranspose2d", "Identity", "Model", "ModelConfig", "Module", "__init__", "_get_backbone_hidden_size", "channels", "class", "config", "def", "elif", "factor", "forward", "hidden_size", "hidden_state", "if", "in_channels", "int", "kernel_size", "nn", "out_channels", "padding", "projection", "resize", "return", "self", "stride", "super"], "dpt/modeling_dpt.py:DPTFeatureFusionStage": ["Model", "ModelConfig", "ModelFeatureFusionLayer", "Module", "ModuleList", "None", "_", "__init__", "append", "class", "config", "def", "else", "for", "forward", "fused_hidden_state", "fused_hidden_states", "hidden_state", "hidden_states", "if", "in", "is", "layer", "layers", "len", "neck_hidden_sizes", "nn", "range", "return", "self", "super", "zip"], "dpt/modeling_dpt.py:DPTPreActResidualLayer": ["BatchNorm2d", "Conv2d", "Model", "ModelConfig", "Module", "None", "ReLU", "Tensor", "__init__", "activation1", "activation2", "batch_norm1", "batch_norm2", "bias", "class", "config", "convolution1", "convolution2", "def", "else", "forward", "fusion_hidden_size", "hidden_state", "if", "is", "kernel_size", "nn", "not", "padding", "residual", "return", "self", "stride", "super", "torch", "use_batch_norm", "use_batch_norm_in_fusion_residual", "use_bias_in_fusion_residual"], "dpt/modeling_dpt.py:DPTFeatureFusionLayer": ["Conv2d", "False", "Model", "ModelConfig", "ModelPreActResidualLayer", "Module", "None", "Optional", "Tensor", "True", "__init__", "align_corners", "bias", "bilinear", "bool", "class", "config", "def", "forward", "functional", "fusion_hidden_size", "hidden_state", "if", "interpolate", "is", "kernel_size", "mode", "nn", "not", "projection", "residual", "residual_layer1", "residual_layer2", "return", "scale_factor", "self", "shape", "size", "super", "torch"], "dpt/modeling_dpt.py:DPTPreTrainedModel": ["BatchNorm2d", "Conv2d", "ConvTranspose2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelSelfAttention", "ModelViTEmbeddings", "ModelViTHybridEmbeddings", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "dpt/modeling_dpt.py:DPTModel": ["BaseModelOutput", "BaseModelOutputWithIntermediateActivations", "BaseModelOutputWithPoolingAndIntermediateActivations", "FloatTensor", "LayerNorm", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelViTEmbeddings", "ModelViTEncoder", "ModelViTHybridEmbeddings", "ModelViTPooler", "None", "Optional", "True", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "else", "embedding_last_hidden_states", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "intermediate_activations", "is", "is_hybrid", "items", "kwargs", "last_hidden_state", "last_hidden_states", "layer", "layer_norm_eps", "layernorm", "nn", "not", "num_hidden_layers", "output_hidden_states", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "return", "self", "sequence_output", "super", "torch"], "dpt/modeling_dpt.py:DPTViTPooler": ["ACT2FN", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "pooler_act", "pooler_output_size", "return", "self", "super", "torch"], "dpt/modeling_dpt.py:DPTNeck": ["Conv2d", "False", "Model", "ModelConfig", "ModelFeatureFusionStage", "ModelReassembleStage", "Module", "ModuleList", "None", "Optional", "Tensor", "The", "TypeError", "ValueError", "__init__", "a", "and", "append", "backbone_config", "be", "bias", "channel", "class", "config", "convs", "def", "else", "enumerate", "equal", "feature", "features", "for", "forward", "fusion_hidden_size", "fusion_stage", "hidden", "hidden_states", "i", "if", "in", "int", "is", "isinstance", "kernel_size", "len", "list", "model_type", "neck", "neck_hidden_sizes", "nn", "not", "number", "of", "or", "output", "padding", "patch_height", "patch_width", "raise", "reassemble_stage", "return", "self", "should", "sizes", "states", "super", "swinv2", "tensors", "the", "to", "torch", "tuple"], "dpt/modeling_dpt.py:DPTDepthEstimationHead": ["Conv2d", "Model", "ModelConfig", "Module", "None", "ReLU", "Sequential", "Tensor", "True", "Upsample", "__init__", "add_projection", "align_corners", "bilinear", "class", "config", "def", "dim", "features", "forward", "fusion_hidden_size", "head", "head_in_index", "hidden_states", "if", "is", "kernel_size", "list", "mode", "nn", "not", "padding", "predicted_depth", "projection", "return", "scale_factor", "self", "squeeze", "stride", "super", "torch"], "dpt/modeling_dpt.py:DPTForDepthEstimation": ["DepthEstimatorOutput", "False", "FloatTensor", "LongTensor", "Model", "ModelDepthEstimationHead", "ModelModel", "ModelNeck", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Training", "True", "_", "__init__", "add_pooling_layer", "and", "attentions", "auto_docstring", "backbone", "backbone_config", "backbone_hidden_states", "backbone_out_indices", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "extend", "feature", "feature_maps", "for", "forward", "forward_with_filtered_kwargs", "head", "head_mask", "height", "hidden_states", "idx", "if", "implemented", "in", "intermediate_activations", "is", "is_hybrid", "kwargs", "labels", "load_backbone", "loss", "neck", "not", "or", "output_hidden_states", "outputs", "patch_height", "patch_size", "patch_width", "pixel_values", "post_init", "predicted_depth", "r", "raise", "return", "self", "shape", "super", "torch", "width", "yet"], "dpt/modeling_dpt.py:DPTSemanticSegmentationHead": ["BatchNorm2d", "Conv2d", "Dropout", "False", "Model", "ModelConfig", "Module", "ReLU", "Sequential", "Tensor", "True", "Upsample", "__init__", "align_corners", "bias", "bilinear", "class", "config", "def", "features", "forward", "fusion_hidden_size", "head", "head_in_index", "hidden_states", "kernel_size", "list", "logits", "mode", "nn", "num_labels", "padding", "return", "scale_factor", "self", "semantic_classifier_dropout", "super", "torch"], "dpt/modeling_dpt.py:DPTAuxiliaryHead": ["BatchNorm2d", "Conv2d", "Dropout", "False", "Model", "ModelConfig", "Module", "ReLU", "Sequential", "Tensor", "__init__", "bias", "class", "config", "def", "features", "forward", "fusion_hidden_size", "head", "hidden_states", "kernel_size", "logits", "nn", "num_labels", "padding", "return", "self", "super", "torch"], "dpt/modeling_dpt.py:DPTForSemanticSegmentation": ["BaseModelOutputWithPoolingAndIntermediateActivations", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "Model", "ModelAuxiliaryHead", "ModelConfig", "ModelModel", "ModelNeck", "ModelPreTrainedModel", "ModelSemanticSegmentationHead", "None", "Optional", "SemanticSegmenterOutput", "The", "True", "ValueError", "__init__", "add_pooling_layer", "align_corners", "and", "attentions", "auto_docstring", "auxiliary_head", "auxiliary_logits", "auxiliary_loss", "auxiliary_loss_weight", "backbone_hidden_states", "backbone_out_indices", "be", "bilinear", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "extend", "feature", "for", "forward", "functional", "greater", "head", "head_mask", "hidden_states", "idx", "if", "ignore_index", "in", "intermediate_activations", "interpolate", "is", "is_hybrid", "kwargs", "labels", "logits", "loss", "loss_fct", "main_loss", "mode", "neck", "nn", "not", "num_labels", "number", "of", "one", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "return", "self", "semantic_loss_ignore_index", "shape", "should", "size", "super", "than", "torch", "upsampled_auxiliary_logits", "upsampled_logits", "use_auxiliary_head"], "gemma/modeling_gemma.py:GemmaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "gemma/modeling_gemma.py:GemmaMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "gemma/modeling_gemma.py:GemmaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gemma/modeling_gemma.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "gemma/modeling_gemma.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "gemma/modeling_gemma.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "gemma/modeling_gemma.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "gemma/modeling_gemma.py:GemmaAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "use_bidirectional_attention", "v_proj", "value_states", "version", "view"], "gemma/modeling_gemma.py:GemmaDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "gemma/modeling_gemma.py:GemmaPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "RMSNorm", "True", "__class__", "__name__", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "in", "model", "module", "past_key_values", "self", "super", "supports_gradient_checkpointing", "weight", "zero_"], "gemma/modeling_gemma.py:GemmaModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "dict", "dtype", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "normalizer", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "tensor", "torch", "unsqueeze", "use_cache", "vocab_size"], "gemma/modeling_gemma.py:GemmaForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "gemma/modeling_gemma.py:GemmaForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "gemma/modeling_gemma.py:GemmaForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "ones", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextFlexibleLinear": ["ModelFlexibleLinear", "Module", "None", "Parameter", "__init__", "class", "def", "else", "forward", "if", "index_select", "input_size", "is", "layer_idx", "matmul", "nn", "not", "num_layers", "output_size", "randn", "return", "selected_weights", "self", "squeeze", "super", "torch", "transpose", "weight", "x"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextPreTrainedModel": ["Embedding", "Linear", "MimiTransformerLayer", "ModelConfig", "ModelDecoderLayer", "ModelFlexibleLinear", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "input_ids", "is", "isinstance", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextConv1dPaddingCache": ["Expected", "ModelConv1dPaddingCache", "None", "NotImplementedError", "Tensor", "True", "ValueError", "__init__", "all", "and", "batch_size", "class", "constant", "convolutions", "current_cache", "def", "device", "dtype", "elif", "else", "empty", "f", "for", "from_args_num_layers", "hidden_states", "if", "in", "in_channels", "int", "is", "layer_idx", "len", "list", "mode", "not", "num_layers", "ones", "or", "other", "padding", "padding_cache", "padding_mode", "padding_states", "per_layer_in_channels", "per_layer_is_init", "per_layer_padding", "per_layer_padding_mode", "pop", "raise", "replicate", "return", "self", "shape", "str", "supported", "than", "torch", "update", "using", "values", "zeros"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextEmbeddings": ["Embedding", "False", "ModelEmbeddings", "Module", "__init__", "arange", "audio_pad_token_id", "audio_tokens_offsets", "class", "codebook_vocab_size", "config", "def", "dim", "embed_tokens", "forward", "functional", "hidden_size", "input_ids", "inputs_embeds", "nn", "num_codebooks", "pad", "padding_idx", "persistent", "register_buffer", "return", "self", "sum", "super", "torch", "vocab_size", "where"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextLinear": ["False", "Linear", "ModelFlexibleLinear", "ModelLinear", "Module", "None", "__init__", "bias", "class", "def", "else", "forward", "if", "input_dim", "layer_idx", "linear", "nn", "not", "num_codebooks", "num_layers", "output_dim", "return", "self", "super", "use_flexible_linear", "x"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextGatingMLP": ["ACT2FN", "False", "Linear", "ModelFlexibleLinear", "ModelGatingMLP", "Module", "None", "Optional", "Tensor", "_", "__init__", "activation_fn", "batch_size", "bias", "class", "config", "def", "else", "fc1", "fc2", "ffn_dim", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "int", "is", "layer_idx", "nn", "num_codebooks", "num_layers", "return", "self", "sequence_length", "shape", "super", "torch", "use_flexible_linear", "view"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextAttention": ["Cache", "False", "Instantiating", "LongTensor", "ModelAttention", "ModelConfig", "ModelLinear", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "else", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "lead", "logger", "make", "math", "matmul", "max_position_embeddings", "must", "new_name", "nn", "not", "num_attention_heads", "num_codebooks", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "repeat_kv", "return", "rope_theta", "rotary_emb", "scaling", "self", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "use_flexible_linear", "use_rope", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "StaticCache", "Tensor", "The", "ValueError", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "an", "and", "apply_rotary_pos_emb", "args", "at", "attention_dropout", "attention_mask", "attn_implementation", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache", "cache_kwargs", "cache_position", "cast", "casted", "class", "com", "compatible", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attention_2", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "github", "hasattr", "have", "head_dim", "hidden", "hidden_states", "https", "huggingface", "if", "implementation", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "isinstance", "issue", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "make", "mean", "might", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "o_proj", "open", "or", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "raise", "related", "reshape", "return", "rotary_emb", "sdpa", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "static", "super", "sure", "target_dtype", "the", "this", "time", "to", "torch", "training", "transformers", "transpose", "tuple", "type", "upcasted", "update", "use", "use_cache", "use_top_left_mask", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "with", "you"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "rotary_emb", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelGatingMLP", "ModelRMSNorm", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "sliding_window", "super", "torch", "tuple", "use_cache", "use_flexible_linear", "use_rope", "version"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextModel": ["Attention", "AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "False", "Flash", "FloatTensor", "LongTensor", "Make", "Model", "ModelConfig", "ModelDecoderLayer", "ModelEmbeddings", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all", "all_hidden_states", "all_self_attns", "and", "arange", "are", "attempting", "attention_mask", "attentions", "auto_docstring", "batch_size", "batched", "before", "behaviour", "bitwise_or_", "bool", "cache_position", "call", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal_attend_mask", "dim", "dtype", "else", "embed_tokens", "eps", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "generation", "get_max_cache_shape", "get_seq_length", "get_text_config", "getattr", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_padding_right", "is_sliding", "is_static_sliding_cache", "is_training", "isinstance", "item", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "lead", "left", "list", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "may", "min", "min_dtype", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "padding_side", "past_key_values", "past_key_values_length", "past_seen_tokens", "perform", "position_ids", "post_init", "raise", "range", "reshape", "return", "return_dict", "right", "rms_norm_eps", "sdpa", "self", "sequence_length", "shape", "size", "sliding_attend_mask", "sliding_window", "staticmethod", "sum", "super", "sure", "target_length", "text_config", "the", "this", "to", "tokenizer", "tokenizing", "torch", "training", "tuple", "type", "unexpected", "unsqueeze", "use_cache", "use_flexible_linear", "use_return_dict", "use_sliding_window", "using_static_cache", "v", "version", "vocab_size", "warning_once", "with", "xpu"], "kyutai_speech_to_text/modeling_kyutai_speech_to_text.py:KyutaiSpeechToTextForConditionalGeneration": ["AutoModel", "BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationConfig", "GenerationMixin", "Linear", "LongTensor", "MethodType", "ModelConv1dPaddingCache", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__func__", "__init__", "_from_model_config", "_get_cache", "_keep_in_fp32_modules_strict", "_mimiconv1d_layer_names", "_pp_plan", "_prepare_cache_for_generation", "_prepare_generation_config", "_prepare_model_inputs", "_supports_default_dynamic_cache", "_tied_weights_keys", "_tp_plan", "append", "args", "attention_mask", "attentions", "attr", "audio", "audio_bos_token_id", "audio_codes", "audio_tokens", "audio_window_size", "auto_docstring", "batch_size", "bias", "bool", "bos_token_id", "cache_implementation", "cache_methods", "cache_position", "can_return_tuple", "cat", "clamp", "class", "classmethod", "clone", "cls", "codec_", "codec_config", "codec_model", "codec_model_attrs", "codec_model_output", "colwise_rep", "config", "contiguous", "copy_", "current_audio_tokens", "current_audio_tokens_idxs", "current_input_values", "current_window", "def", "delattr", "device", "dict", "dim", "downsample", "dtype", "dynamic", "else", "encode", "encoder", "encoder_past_key_values", "end", "expand", "f", "for", "forward", "frame_size", "frames", "from_config", "from_model_config", "from_pretrained", "generate", "generation_config", "generation_mode", "get", "get_encoded_length", "get_submodule", "getattr", "greater", "hasattr", "hidden_size", "hidden_states", "if", "in", "in_channels", "input_ids", "input_name", "input_values", "input_values_end_idx", "input_values_start_idx", "inputs", "inputs_embeds", "int", "is", "isinstance", "item", "items", "kwargs", "labels", "lambda", "last_hidden_state", "layer_name", "len", "lm_head", "loading_info", "logger", "logits", "logits_to_keep", "long", "loss", "loss_function", "max_audio_frames", "max_cache_length", "max_new_tokens", "maximum", "method", "min", "model", "model_inputs", "model_kwargs", "new_audio_tokens", "nn", "no_grad", "not", "num_codebooks", "num_layers", "number", "of", "or", "output_loading_info", "outputs", "pad_mode", "padding_cache", "padding_mask", "padding_total", "past_key_values", "per_layer_in_channels", "per_layer_padding", "per_layer_padding_mode", "pop", "position_ids", "post_init", "prefix", "prefix_len", "prepare_inputs_for_generation", "r", "return", "save_pretrained", "self", "setattr", "shape", "slice", "slice_indices", "sliding_window", "start", "startswith", "str", "super", "temporary_model_kwargs", "tensor", "than", "the", "to", "to_diff_dict", "torch", "transformers_version", "transpose", "tuple", "types", "unsqueeze", "update", "use_cache", "value", "vars", "vocab_size", "warning", "weight", "with", "x", "zeros"], "metaclip_2/modeling_metaclip_2.py:MetaClip2TextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "metaclip_2/modeling_metaclip_2.py:MetaClip2VisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "metaclip_2/modeling_metaclip_2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "True", "attention_mask", "attn_output", "attn_weights", "bool", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "output_attentions", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "metaclip_2/modeling_metaclip_2.py:MetaClip2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "ModelTextConfig", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "metaclip_2/modeling_metaclip_2.py:MetaClip2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "metaclip_2/modeling_metaclip_2.py:MetaClip2PreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelForImageClassification", "ModelMLP", "ModelModel", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelTextModelWithProjection", "ModelVisionEmbeddings", "ModelVisionModelWithProjection", "None", "PreTrainedModel", "True", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "base_model_prefix", "bias", "class", "class_embedding", "classifier", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_embedding", "v_proj", "vision_config", "vision_embed_dim", "visual_projection", "weight", "zero_"], "metaclip_2/modeling_metaclip_2.py:MetaClip2EncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelMLP", "ModelTextConfig", "ModelVisionConfig", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "metaclip_2/modeling_metaclip_2.py:MetaClip2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "self", "super", "torch"], "metaclip_2/modeling_metaclip_2.py:MetaClip2TextTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "check_model_inputs", "class", "config", "def", "device", "dim", "dtype", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eos_token_id", "eps", "final_layer_norm", "flash_attention_2", "forward", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "pooled_output", "pooler_output", "position_ids", "return", "self", "shape", "size", "super", "to", "torch", "use_cache", "view"], "metaclip_2/modeling_metaclip_2.py:MetaClip2TextModel": ["BaseModelOutputWithPooling", "False", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "__init__", "_no_split_modules", "_supports_flash_attn", "attention_mask", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "value"], "metaclip_2/modeling_metaclip_2.py:MetaClip2TextModelOutput": ["FloatTensor", "ModelOutput", "ModelTextModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "metaclip_2/modeling_metaclip_2.py:MetaClip2TextModelWithProjection": ["BaseModelOutputWithPooling", "False", "Linear", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "ModelTextModelOutput", "ModelTextModelWithProjection", "Module", "None", "Optional", "Tensor", "__init__", "_from_config", "_no_split_modules", "_supports_flash_attn", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "input_ids", "last_hidden_state", "nn", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "return", "self", "set_input_embeddings", "super", "text_embeds", "text_model", "text_outputs", "text_projection", "token_embedding", "torch", "value"], "metaclip_2/modeling_metaclip_2.py:MetaClip2Output": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "metaclip_2/modeling_metaclip_2.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "metaclip_2/modeling_metaclip_2.py:metaclip_2_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "metaclip_2/modeling_metaclip_2.py:_get_vector_norm": ["Tensor", "True", "_get_vector_norm", "def", "dim", "keepdim", "normed_tensor", "pow", "return", "square_tensor", "sum", "sum_tensor", "tensor", "torch"], "metaclip_2/modeling_metaclip_2.py:MetaClip2Model": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelEncoderLayer", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "Model_loss", "None", "Optional", "Parameter", "Tensor", "TypeError", "__init__", "_from_config", "_get_vector_norm", "_no_split_modules", "_supports_flash_attn", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "can_return_tuple", "class", "config", "def", "device", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "not", "of", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "return", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "type", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "metaclip_2/modeling_metaclip_2.py:MetaClip2VisionTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "None", "Optional", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layrnorm", "raise", "return", "self", "specify", "super", "to", "torch"], "metaclip_2/modeling_metaclip_2.py:MetaClip2VisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "__init__", "_no_split_modules", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "self", "super", "torch", "vision_model"], "metaclip_2/modeling_metaclip_2.py:MetaClip2VisionModelOutput": ["FloatTensor", "ModelOutput", "ModelVisionModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "metaclip_2/modeling_metaclip_2.py:MetaClip2VisionModelWithProjection": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionModelOutput", "ModelVisionModelWithProjection", "Module", "None", "Optional", "__init__", "_from_config", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "image_embeds", "interpolate_pos_encoding", "last_hidden_state", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "pooled_output", "pooler_output", "post_init", "projection_dim", "r", "return", "self", "super", "torch", "vision_model", "vision_outputs", "visual_projection"], "metaclip_2/modeling_metaclip_2.py:MetaClip2ForImageClassification": ["BaseModelOutputWithPooling", "Identity", "ImageClassifierOutput", "Linear", "ModelConfig", "ModelForImageClassification", "ModelPreTrainedModel", "ModelVisionModel", "None", "Optional", "Tensor", "__init__", "_from_config", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "dim", "else", "forward", "hidden_size", "hidden_states", "if", "is", "labels", "last_hidden_state", "logits", "loss", "loss_function", "main_input_name", "mean", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch", "vision_config", "vision_model"], "granite/modeling_granite.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "granite/modeling_granite.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "granite/modeling_granite.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "granite/modeling_granite.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "granite/modeling_granite.py:GraniteAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_multiplier", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "granite/modeling_granite.py:GraniteRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "granite/modeling_granite.py:GraniteMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "granite/modeling_granite.py:GraniteDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "residual_multiplier", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "granite/modeling_granite.py:GranitePreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "granite/modeling_granite.py:GraniteRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "granite/modeling_granite.py:GraniteModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "checkpointing", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "embedding_multiplier", "eps", "exactly", "for", "forward", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "logger", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "training", "unsqueeze", "use_cache", "vocab_size", "warning_once", "with"], "granite/modeling_granite.py:GraniteForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logits", "logits_scaling", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "flava/modeling_flava.py:FlavaModelOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeddings", "image_output", "in", "k", "keys", "multimodal_embeddings", "multimodal_output", "not", "r", "return", "self", "text_embeddings", "text_output", "to_tuple", "torch", "tuple"], "flava/modeling_flava.py:FlavaLosses": ["False", "FloatTensor", "ModelLosses", "ModelOutput", "None", "Optional", "True", "all_none", "bool", "break", "class", "def", "for", "global_contrastive", "if", "in", "is", "itm", "mim", "mlm", "mmm_image", "mmm_text", "not", "r", "return", "self", "torch", "v", "values"], "flava/modeling_flava.py:FlavaForPreTrainingOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelForPreTrainingOutput", "ModelLosses", "ModelOutput", "None", "Optional", "class", "contrastive_logits_per_image", "contrastive_logits_per_text", "def", "else", "for", "getattr", "if", "image_embeddings", "image_masked_embeddings", "image_masked_output", "image_output", "in", "itm_logits", "k", "keys", "loss", "loss_info", "mim_logits", "mlm_logits", "mmm_image_logits", "mmm_text_logits", "multimodal_embeddings", "multimodal_masked_embeddings", "multimodal_masked_output", "multimodal_output", "not", "r", "return", "self", "text_embeddings", "text_masked_embeddings", "text_masked_output", "text_output", "to_tuple", "torch", "transformer_outputs", "tuple"], "flava/modeling_flava.py:FlavaImageEmbeddings": ["BoolTensor", "Dropout", "False", "ModelImageConfig", "ModelImageEmbeddings", "Module", "None", "Optional", "Parameter", "PatchEmbeddings", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "else", "embed_dim", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "image_size", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_mask_token", "view", "width", "zeros"], "flava/modeling_flava.py:PatchEmbeddings": ["Conv2d", "False", "Input", "Iterable", "ModelEmbeddings", "Model_size", "Module", "Tensor", "Union", "ValueError", "__init__", "abc", "batch_size", "bool", "class", "collections", "def", "doesn", "embed_dim", "f", "flatten", "forward", "height", "if", "image", "image_size", "int", "interpolate_pos_encoding", "isinstance", "kernel_size", "match", "model", "nn", "not", "num_Modeles", "num_channels", "or", "pixel_values", "projection", "raise", "return", "self", "shape", "size", "stride", "super", "t", "torch", "transpose", "tuple", "width", "x"], "flava/modeling_flava.py:FlavaTextEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "flava/modeling_flava.py:FlavaSelfAttention": ["Dropout", "False", "Linear", "ModelPossibleConfigs", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "Union", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "functional", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "view"], "flava/modeling_flava.py:FlavaSelfOutput": ["Dropout", "Linear", "ModelPossibleConfigs", "ModelSelfOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "flava/modeling_flava.py:FlavaAttention": ["False", "ModelAttention", "ModelPossibleConfigs", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "Union", "__init__", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "flava/modeling_flava.py:FlavaIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "ModelPossibleConfigs", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "flava/modeling_flava.py:FlavaOutput": ["Dropout", "Linear", "ModelOutput", "ModelPossibleConfigs", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "flava/modeling_flava.py:FlavaLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "ModelPossibleConfigs", "None", "Optional", "Tensor", "Union", "__init__", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "flava/modeling_flava.py:FlavaEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "flava/modeling_flava.py:FlavaPooler": ["Linear", "ModelPooler", "ModelPossibleConfigs", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "flava/modeling_flava.py:FlavaPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelImageEmbeddings", "ModelMaskedPredictionHead", "ModelModel", "ModelMultimodalModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "Union", "_init_weights", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "logit_scale", "logit_scale_init_value", "mask_token", "mean", "module", "nn", "normal_", "not", "padding_idx", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "use_cls_token", "weight", "zero_"], "flava/modeling_flava.py:FlavaImageModel": ["BaseModelOutputWithPooling", "BoolTensor", "LayerNorm", "Model", "ModelEncoder", "ModelImageConfig", "ModelImageEmbeddings", "ModelImageModel", "ModelPooler", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bool", "bool_masked_pos", "class", "config", "def", "dict", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "image_model", "in", "int", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "main_input_name", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "specify", "super", "to", "torch", "tuple", "use_return_dict", "value"], "flava/modeling_flava.py:FlavaTextModel": ["BaseModelOutputWithPooling", "LayerNorm", "Model", "ModelEncoder", "ModelPooler", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "Module", "None", "Optional", "PatchEmbeddings", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bool", "class", "config", "def", "device", "dict", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "input_ids", "input_shape", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "not", "num_hidden_layers", "ones", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "text_model", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "word_embeddings"], "flava/modeling_flava.py:FlavaMultimodalModel": ["BaseModelOutputWithPooling", "LayerNorm", "Model", "ModelEncoder", "ModelMultimodalConfig", "ModelMultimodalModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "_", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "bool", "cat", "class", "cls_token", "cls_tokens", "config", "def", "device", "dict", "dim", "else", "encoder", "encoder_outputs", "eps", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "main_input_name", "multimodal_model", "nn", "not", "num_hidden_layers", "ones", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "return", "return_dict", "self", "seq_length", "sequence_output", "size", "super", "torch", "tuple", "use_cls_token", "use_return_dict", "zeros"], "flava/modeling_flava.py:FlavaModel": ["BaseModelOutputWithPooling", "BoolTensor", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelImageConfig", "ModelImageModel", "ModelModel", "ModelModelOutput", "ModelMultimodalConfig", "ModelMultimodalModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "None", "Optional", "Parameter", "Please", "Tensor", "True", "TypeError", "Union", "ValueError", "_", "__init__", "and", "attention_mask", "attention_mask_image", "attention_multimodal", "auto_docstring", "batch_size", "be", "bool", "bool_masked_pos", "but", "cat", "class", "config", "def", "device", "dim", "else", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "head_mask", "hidden", "hidden_size", "if", "image_attention_mask", "image_config", "image_embeddings", "image_features", "image_hidden_size", "image_mm_projection", "image_model", "image_output", "image_outputs", "image_projection", "image_states", "image_to_mm_projection", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "last_hidden_state", "logit_scale", "logit_scale_init_value", "mm_hidden_size", "model", "multimodal_config", "multimodal_embeddings", "multimodal_input", "multimodal_model", "multimodal_output", "nn", "not", "of", "ones", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "position_ids", "post_init", "projection_dim", "r", "raise", "requires", "return", "return_dict", "self", "seq_len", "set", "shape", "skip_multimodal_encoder", "states", "super", "tensor", "text_config", "text_embeddings", "text_features", "text_hidden_size", "text_mm_projection", "text_model", "text_output", "text_outputs", "text_projection", "text_states", "text_to_mm_projection", "to", "token_type_ids", "torch", "tuple", "type", "use_cls_token", "work"], "flava/modeling_flava.py:FlavaImageCodebookResPath": ["Conv2d", "ModelImageCodebookResPath", "Module", "OrderedDict", "ReLU", "Sequential", "Tensor", "__init__", "class", "conv_1", "conv_2", "conv_3", "conv_4", "def", "forward", "hid_size", "in_size", "int", "kernel_size", "kwargs", "nn", "out_size", "padding", "path", "relu_1", "relu_2", "relu_3", "relu_4", "return", "self", "super", "torch", "x"], "flava/modeling_flava.py:FlavaImageCodebookBlock": ["Conv2d", "Identity", "ModelImageCodebookBlock", "ModelImageCodebookResPath", "Module", "Tensor", "__init__", "class", "def", "else", "forward", "id_path", "if", "in_size", "int", "kernel_size", "kwargs", "nn", "num_layers", "out_size", "padding", "post_gain", "res_path", "return", "self", "super", "torch", "x"], "flava/modeling_flava.py:FlavaImageCodebookLayerGroup": ["MaxPool2d", "ModelImageCodebookBlock", "ModelImageCodebookLayerGroup", "Module", "OrderedDict", "Sequential", "Tensor", "True", "__init__", "block_", "blocks", "bool", "class", "def", "else", "f", "for", "forward", "group", "i", "if", "in", "in_size", "int", "kernel_size", "nn", "num_blocks", "num_layers", "out_size", "pool", "range", "return", "self", "super", "torch", "use_pool", "x"], "flava/modeling_flava.py:FlavaImageCodebook": ["Any", "Conv2d", "False", "FloatTensor", "ModelImageCodebook", "ModelImageCodebookConfig", "ModelImageCodebookLayerGroup", "ModelPreTrainedModel", "OrderedDict", "ReLU", "Sequential", "Softmax", "Tensor", "ValueError", "__init__", "argmax", "axis", "base_model_prefix", "blocks", "built", "but", "channels", "class", "config", "conv", "def", "dim", "f", "for", "forward", "freeze", "get_codebook_indices", "get_codebook_probs", "group_1", "group_2", "group_3", "group_4", "has", "hidden_size", "if", "in", "input", "input_channels", "is", "kernel_size", "kwargs", "len", "main_input_name", "model", "nn", "not", "num_blocks_per_group", "num_groups", "num_layers", "output", "output_blocks", "padding", "param", "parameters", "pixel_values", "post_init", "raise", "relu", "requires_grad", "return", "self", "shape", "super", "supports_gradient_checkpointing", "torch", "use_pool", "vocab_size", "z_logits"], "flava/modeling_flava.py:FlavaPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "flava/modeling_flava.py:FlavaMaskedPredictionHead": ["False", "Linear", "ModelMaskedPredictionHead", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "if", "is", "nn", "not", "return", "self", "super", "torch", "transform", "vocab_size", "weight", "x", "zeros"], "flava/modeling_flava.py:FlavaITMHead": ["Linear", "ModelITMHead", "ModelPooler", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooler", "return", "self", "seq_relationship", "super", "x"], "flava/modeling_flava.py:FlavaGlobalContrastiveHead": ["ModelGlobalContrastiveHead", "Module", "_", "__init__", "all_gather", "arange", "cat", "class", "config", "def", "device", "distributed", "else", "exp", "for", "forward", "functional", "get_rank", "get_world_size", "global_backprop_contrastive", "if", "image_embeddings", "image_embeddings_all", "in", "is_available", "is_initialized", "labels", "local_batch_size", "logit_scale", "logits_per_image", "logits_per_text", "matmul", "nn", "not", "or", "range", "return", "self", "size", "super", "temperature", "text_embeddings", "text_embeddings_all", "torch", "transpose", "world_size", "zeros_like"], "flava/modeling_flava.py:FlavaForPreTraining": ["AutoProcessor", "Call", "FloatTensor", "LOGIT_SCALE_CLAMP_MAX", "LOGIT_SCALE_CLAMP_MIN", "LongTensor", "MLM", "Model", "ModelConfig", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelGlobalContrastiveHead", "ModelITMHead", "ModelImageCodebook", "ModelLosses", "ModelMaskedPredictionHead", "ModelModel", "ModelPreTrainedModel", "Model_losses", "Model_masked_output", "Model_output", "Module", "None", "OKAY", "Optional", "Please", "Reinstantiate", "RuntimeError", "Tensor", "This", "True", "Union", "ValueError", "__init__", "_resize_to_2d", "_tied_weights_keys", "all_none", "and", "any", "are", "attention_mask", "auto_docstring", "be", "been", "bias", "bool", "bool_masked_pos", "but", "calculated", "can", "ce_ignore_index", "clamp_", "class", "codebook", "codebook_pixel_value", "codebook_pixel_values", "config", "contrastive_logits_per_image", "contrastive_logits_per_text", "correctlySetting", "cross_entropy", "custom", "data", "decoder", "def", "dim", "doing", "else", "end_index", "expected", "for", "forward", "functional", "gc_labels", "gc_loss", "gc_loss_image", "gc_loss_text", "generate", "get_codebook_indices", "global_contrastive", "global_contrastive_head", "global_contrastive_weight", "have", "if", "image", "image_attention_mask", "image_codebook", "image_codebook_config", "image_config", "image_embedding", "image_embeddings", "image_masked_embeddings", "image_masked_output", "image_output", "image_projection", "image_vocab_size", "in", "inference", "init_codebook", "initialized", "input_ids", "input_ids_masked", "is", "isn", "it", "itm", "itm_head", "itm_labels", "itm_logits", "itm_loss", "itm_weight", "logger", "logit_scale", "logits_per_image", "logits_per_text", "loss", "loss_info", "masked_tokens", "means", "mim", "mim_head", "mim_labels", "mim_labels_filtered", "mim_logits", "mim_loss", "mim_weight", "mlm", "mlm_head", "mlm_labels", "mlm_labels_filtered", "mlm_logits", "mlm_loss", "mlm_weight", "mmm_image", "mmm_image_head", "mmm_image_logits", "mmm_image_loss", "mmm_image_weight", "mmm_text", "mmm_text_head", "mmm_text_logits", "mmm_text_loss", "mmm_text_weight", "model", "multimodal_embeddings", "multimodal_masked_embeddings", "multimodal_masked_output", "multimodal_output", "ne", "new", "nn", "no", "normalize", "not", "on", "or", "output", "output_attentions", "output_hidden_states", "pass", "passed", "pixel_values", "pos_mask", "pos_pairs", "position_ids", "post_init", "r", "raise", "required", "return", "return_codebook_pixels", "return_dict", "return_loss", "self", "sequence_for_image", "sequence_for_text", "set", "size", "skip_multimodal_encoder", "skip_unmasked_multimodal_encoder", "so", "sum", "super", "t", "text", "text_config", "text_embedding", "text_embeddings", "text_masked_embeddings", "text_masked_output", "text_output", "text_projection", "text_vocab_size", "that", "the", "this", "to", "to_tuple", "token_type_ids", "torch", "total_loss", "tuple", "unintentional", "unmasked", "use_return_dict", "usually", "values", "view", "vocab_size", "warning", "where", "which", "with", "won", "work", "x", "you", "your"], "smolvlm/modeling_smolvlm.py:SmolVLMRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "smolvlm/modeling_smolvlm.py:SmolVLMPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionAttention", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "get_text_config", "getattr", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "smolvlm/modeling_smolvlm.py:SmolVLMVisionEmbeddings": ["BoolTensor", "Conv2d", "Embedding", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "True", "_", "__init__", "arange", "batch_idx", "batch_size", "boundaries", "bucket_coords_h", "bucket_coords_w", "bucketize", "class", "config", "def", "device", "dtype", "embed_dim", "embeddings", "enumerate", "fill_value", "flatten", "for", "forward", "fractional_coords_h", "fractional_coords_w", "full", "h_indices", "hidden_size", "image_size", "in", "in_channels", "kernel_size", "max_im_h", "max_im_w", "max_nb_patches_h", "max_nb_patches_w", "nb_patches_h", "nb_patches_w", "nn", "num_channels", "num_patches", "num_patches_per_side", "num_positions", "out_channels", "p_attn_mask", "padding", "patch_attention_mask", "patch_embedding", "patch_embeds", "patch_size", "pixel_values", "pos_ids", "position_embedding", "position_ids", "return", "right", "self", "shape", "size", "stride", "sum", "super", "torch", "transpose", "valid", "view", "w_indices"], "smolvlm/modeling_smolvlm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "smolvlm/modeling_smolvlm.py:SmolVLMVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "smolvlm/modeling_smolvlm.py:SmolVLMVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "smolvlm/modeling_smolvlm.py:SmolVLMEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelEncoderLayer", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "smolvlm/modeling_smolvlm.py:SmolVLMEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "last_hidden_state", "layer_outputs", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "smolvlm/modeling_smolvlm.py:SmolVLMVisionTransformer": ["BaseModelOutput", "BoolTensor", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_attn_implementation", "_can_record_outputs", "_prepare_4d_attention_mask", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "any", "attention_mask", "attentions", "batch_size", "bool", "check_model_inputs", "class", "config", "def", "device", "dtype", "elif", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "flash_attention_2", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "ones", "patch_attention_mask", "patch_size", "pixel_values", "post_layernorm", "return", "self", "set_input_embeddings", "size", "super", "to", "torch", "tuple", "value", "view"], "smolvlm/modeling_smolvlm.py:SmolVLMBaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "smolvlm/modeling_smolvlm.py:SmolVLMSimpleMLP": ["False", "Linear", "ModelSimpleMLP", "Module", "__init__", "bias", "class", "config", "def", "forward", "hidden_size", "input_size", "nn", "output_size", "proj", "return", "scale_factor", "self", "super", "text_config", "vision_config", "x"], "smolvlm/modeling_smolvlm.py:SmolVLMConnector": ["ModelConnector", "ModelSimpleMLP", "Module", "__init__", "bsz", "class", "config", "def", "embed_dim", "forward", "height", "image_hidden_states", "int", "modality_projection", "nn", "permute", "pixel_shuffle", "reshape", "return", "scale_factor", "self", "seq", "size", "super", "view", "width", "x"], "smolvlm/modeling_smolvlm.py:SmolVLMModel": ["At", "AutoModel", "BoolTensor", "Cache", "DynamicCache", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelBaseModelOutputWithPast", "ModelConfig", "ModelConnector", "ModelModel", "ModelPreTrainedModel", "ModelVisionTransformer", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_from_config", "_text_require_grads_hook", "_vision_require_grads_hook", "all", "and", "any", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "block_idx", "block_offset", "blocks_per_sample", "bool", "both", "by", "cache_position", "can_return_tuple", "cannot", "checkpointing", "children", "chunk_idx", "class", "config", "connector", "contiguous", "cumsum", "custom_intro", "def", "device", "dim", "dimension", "disable_input_require_grads", "divisible", "dtype", "either", "elif", "else", "enable_input_require_grads", "for", "forward", "from_config", "functional", "get_image_features", "get_input_embeddings", "get_lowest_module", "gradient", "gradient_checkpointing", "has", "have", "height", "hidden_states", "i", "if", "image", "image_embeds", "image_hidden_states", "image_mask", "image_seq_len", "image_size", "image_token_id", "in", "incompatible", "input", "input_ids", "inputs_embeds", "inputs_merger", "int", "is", "kwargs", "last_hidden_state", "least", "len", "list", "local_idx", "logger", "long", "make_inputs_require_grads", "merged_embeds", "module", "nb_values_per_image", "nn", "not", "num_channels", "num_image_tokens", "num_images", "numel", "offsets", "one", "ones", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pad", "pad_token_id", "padding_idx", "past_key_values", "patch_attention_mask", "patch_size", "patches_subgrid", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "r", "raise", "real_images_inds", "register_forward_hook", "remove", "requires_grad_", "return", "return_dict", "row_cum", "same", "sample", "scale_factor", "self", "seq_length", "set_input_embeddings", "shape", "size", "specify", "step", "sum", "super", "tensor", "text_config", "text_model", "the", "time", "to", "tokens", "torch", "training", "tuple", "unfold", "unsqueeze", "use_cache", "use_return_dict", "value", "view", "vision_config", "vision_model", "vocab_size", "warning_once", "where", "width", "with", "zeros_like"], "smolvlm/modeling_smolvlm.py:SmolVLMCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "smolvlm/modeling_smolvlm.py:SmolVLMForConditionalGeneration": ["BoolTensor", "Cache", "False", "FloatTensor", "GenerationConfig", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_text_require_grads_hook", "_tied_weights_keys", "_vision_require_grads_hook", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "disable_input_require_grads", "else", "enable_input_require_grads", "forward", "from_model_config", "generation_config", "get_image_features", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_token_id", "input", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "make_inputs_require_grads", "model", "model_inputs", "module", "nn", "not", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "register_forward_hook", "remove", "requires_grad_", "return", "return_dict", "self", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "text_model", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_model", "vocab_size", "weight"], "rembert/modeling_rembert.py:RemBertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "arange", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "if", "input_embedding_size", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "rembert/modeling_rembert.py:RemBertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "rembert/modeling_rembert.py:RemBertSelfAttention": ["Cache", "Dropout", "EncoderDecoderCache", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "True", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bool", "cache_position", "class", "config", "context_layer", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "dropout", "else", "embedding_size", "encoder_hidden_states", "f", "forward", "functional", "get", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "key", "key_layer", "keys", "layer_idx", "layers", "math", "matmul", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "past_key_value", "past_key_values", "permute", "query", "query_layer", "raise", "return", "self", "self_attention_cache", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "update", "value", "value_layer", "values", "version", "view"], "rembert/modeling_rembert.py:RemBertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "rembert/modeling_rembert.py:RemBertAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "cache_position", "class", "config", "def", "dense", "dim", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "layer_idx", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "rembert/modeling_rembert.py:RemBertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "rembert/modeling_rembert.py:RemBertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "rembert/modeling_rembert.py:RemBertLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "bool", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_idx", "layer_output", "layers", "model", "not", "output", "output_attentions", "outputs", "passed", "past_key_values", "raise", "return", "self", "self_attention_outputs", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "rembert/modeling_rembert.py:RemBertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "Linear", "ModelEncoder", "ModelLayer", "Modeloved", "Module", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "You", "__init__", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "attention_mask", "attentions", "be", "bool", "cache_position", "checkpointing", "class", "config", "cross_attentions", "def", "deprecated", "e", "else", "embedding_hidden_mapping_in", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "from_legacy_cache", "g", "gradient", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_embedding_size", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_outputs", "logger", "nn", "not", "num_hidden_layers", "of", "output_attentions", "output_hidden_states", "pass", "past_key_values", "range", "return", "return_dict", "self", "should", "super", "torch", "training", "tuple", "use_cache", "v", "v4", "warning_once", "will", "with"], "rembert/modeling_rembert.py:RemBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "rembert/modeling_rembert.py:RemBertLMPredictionHead": ["ACT2FN", "LayerNorm", "Linear", "ModelLMPredictionHead", "Module", "Tensor", "__init__", "activation", "class", "config", "decoder", "def", "dense", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "layer_norm_eps", "nn", "output_embedding_size", "return", "self", "super", "torch", "vocab_size"], "rembert/modeling_rembert.py:RemBertOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "rembert/modeling_rembert.py:RemBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "rembert/modeling_rembert.py:RemBertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "False", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cache_position", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_key_values_length", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "rembert/modeling_rembert.py:RemBertForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "PAD", "The", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "assert", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bi", "bool", "can_generate", "cat", "class", "classmethod", "cls", "config", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "rembert/modeling_rembert.py:RemBertForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "Model", "ModelForCausalLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bool", "class", "cls", "config", "cross_attentions", "decoder", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "use_return_dict", "vocab_size", "want", "warning", "weight", "you"], "rembert/modeling_rembert.py:RemBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "rembert/modeling_rembert.py:RemBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "rembert/modeling_rembert.py:RemBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "rembert/modeling_rembert.py:RemBertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp_", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "granitemoeshared/modeling_granitemoeshared.py:GraniteFlashAttentionKwargs": ["False", "IntTensor", "LongTensor", "ModelFlashAttentionKwargs", "TypedDict", "class", "cu_seq_lens_k", "cu_seq_lens_q", "int", "max_length_k", "max_length_q", "seq_idx", "torch", "total"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedMLP": ["ACT2FN", "False", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "__init__", "activation", "bias", "chunk", "chunked_hidden_states", "class", "config", "def", "dim", "forward", "hidden_act", "hidden_size", "hidden_states", "input_linear", "input_size", "nn", "output_linear", "return", "self", "shared_intermediate_size", "super", "torch"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedParallelExperts": ["F", "ModelParallelExperts", "Module", "None", "Parameter", "__init__", "append", "cat", "class", "def", "dim", "empty", "expert_size", "for", "forward", "i", "in", "input_list", "input_size", "inputs", "int", "linear", "nn", "num_experts", "output_list", "output_size", "range", "results", "return", "self", "split", "super", "torch", "weight"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedTopKGating": ["False", "Linear", "ModelTopKGating", "Module", "_", "__init__", "batch_gates", "batch_index", "bias", "class", "def", "device", "dim", "div", "dtype", "expert_size", "flatten", "float", "forward", "gates", "hidden_states", "index_sorted_experts", "input_size", "int", "layer", "logits", "long", "nn", "num_experts", "return", "rounding_mode", "scatter", "self", "size", "softmax", "sort", "sum", "super", "tolist", "top_k", "top_k_experts", "top_k_gates", "top_k_indices", "top_k_logits", "topk", "torch", "trunc", "type_as", "zeros"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedMoE": ["ACT2FN", "ModelConfig", "ModelMoE", "ModelParallelExperts", "ModelTopKGating", "Module", "None", "_", "__init__", "activation", "batch_gates", "batch_index", "bsz", "chunk", "chunked_hidden_states", "class", "config", "def", "device", "dim", "dtype", "emb_size", "expert_inputs", "expert_outputs", "expert_size", "forward", "hidden_act", "hidden_size", "hidden_states", "index_add", "input_linear", "input_size", "intermediate_size", "layer_input", "layer_output", "length", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "output_linear", "reshape", "return", "router", "router_logits", "self", "size", "super", "top_k", "torch", "view", "zeros"], "granitemoeshared/modeling_granitemoeshared.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "granitemoeshared/modeling_granitemoeshared.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "granitemoeshared/modeling_granitemoeshared.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "granitemoeshared/modeling_granitemoeshared.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_multiplier", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "errors", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "return", "scaling", "self", "sin", "size", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFlashAttentionKwargs", "ModelMLP", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "block_sparse_moe", "bool", "cache_position", "class", "config", "def", "del", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is", "kwargs", "layer_idx", "moe_hidden_states", "new_name", "num_local_experts", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "residual_multiplier", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "shared_intermediate_size", "shared_mlp", "super", "torch", "tuple", "use_cache", "version"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedPreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelParallelExperts", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "data", "def", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "embedding_multiplier", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "head_dim", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "list", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "max_position_embeddings", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_attention_heads", "num_heads", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embedding_type", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "return_dict", "rms_norm_eps", "rope", "rope_theta", "rotary_emb", "router_logits", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "vocab_size", "warning_once", "with", "xpu"], "granitemoeshared/modeling_granitemoeshared.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "granitemoeshared/modeling_granitemoeshared.py:GraniteMoeSharedForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "class", "config", "def", "device", "else", "float", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "list", "lm_head", "load_balancing_loss_func", "logits", "logits_scaling", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyOutputWithPast": ["Cache", "FloatTensor", "ModelOutput", "ModelOutputWithPast", "None", "Optional", "attentions", "class", "encoder_hidden_states", "hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "musicgen_melody/modeling_musicgen_melody.py:shift_tokens_right": ["Make", "Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "attribute", "clone", "configuration", "decoder_start_token_id", "def", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "of", "pad_token_id", "raise", "return", "s", "set", "shape", "sure", "the", "to", "torch", "transpose"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodySinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "Tensor", "_", "__init__", "arange", "bsz", "cat", "class", "cos", "def", "detach", "device", "dim", "dtype", "emb", "emb_weights", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "index_select", "inputs_embeds", "int", "int64", "log", "make_weights", "math", "nn", "no_grad", "num_embeddings", "num_positions", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sin", "size", "staticmethod", "super", "to", "torch", "unsqueeze", "view", "weights", "zeros"], "musicgen_melody/modeling_musicgen_melody.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "class", "config", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "past_key_value", "past_key_values", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelDecoderConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bias", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "embed_dim", "fc1", "fc2", "ffn_dim", "final_layer_norm", "forward", "functional", "hidden_size", "hidden_states", "is_causal", "is_decoder", "layer_head_mask", "layer_idx", "new_name", "nn", "num_attention_heads", "num_heads", "output_attentions", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelAttention", "ModelDecoderConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_factor", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyDecoder": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelDecoder", "ModelDecoderConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Size", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_prepare_4d_causal_attention_mask", "_prepare_4d_causal_attention_mask_for_sdpa", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_attentions", "all_hidden_states", "an", "and", "at", "attention_mask", "attentions", "attn_implementation", "auto_docstring", "be", "bool", "both", "bsz", "but", "cache_position", "cannot", "cat", "checkpointing", "class", "codebook", "config", "continue", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "def", "deprecated", "device", "dim", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_dim", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "flash_attention_2", "flex_attention", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "int", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "make_flex_block_causal_mask", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "num_codebooks", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "r", "raise", "random", "range", "removed", "reshape", "return", "return_dict", "same", "scale_embedding", "sdpa", "self", "seq_len", "shape", "should", "size", "specified", "specify", "sqrt", "sum", "super", "the", "time", "to", "torch", "training", "tuple", "uniform", "use_cache", "use_return_dict", "v", "v4", "vocab_size", "warning_once", "will", "with"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyModel": ["BaseModelOutputWithPast", "Cache", "FloatTensor", "LongTensor", "ModelDecoder", "ModelDecoderConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "decoder", "decoder_outputs", "def", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "last_hidden_state", "not", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "value"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyForCausalLM": ["BaseStreamer", "Cache", "ClassifierFreeGuidanceLogitsProcessor", "CrossEntropyLoss", "Ensure", "False", "FloatTensor", "GREEDY_SEARCH", "GenerationConfig", "GenerationMixin", "GenerationMode", "Got", "Linear", "LogitsProcessorList", "LongTensor", "ModelDecoderConfig", "ModelForCausalLM", "ModelModel", "ModelOutputWithPast", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "SAMPLE", "StoppingCriteriaList", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_decoder_start_token_tensor", "_expand_inputs_for_generation", "_get_logits_processor", "_get_stopping_criteria", "_pad_token_tensor", "_prepare_attention_mask_for_generation", "_prepare_generated_length", "_prepare_model_inputs", "_prepare_special_tokens", "_sample", "_validate_model_kwargs", "activated", "and", "append", "apply_delay_pattern_mask", "attention_mask", "attentions", "audio_channels", "auto_docstring", "batch_size", "be", "beam", "bias", "bool", "bos_token_id", "bsz", "build_delay_pattern_mask", "by", "cache_position", "channel_codebooks", "class", "codebook", "codebook_labels", "codebook_logits", "concatenate", "config", "contiguous", "copy", "cpu", "de", "decoder", "decoder_pad_token_mask", "deepcopy", "def", "delay_pattern", "delay_pattern_mask", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "encoder_input_ids", "encoder_outputs", "expand_size", "first_codebook_ids", "first_start_id", "for", "forward", "generate", "generation", "generation_config", "generation_mode", "get", "get_decoder", "get_generation_mode", "get_input_embeddings", "get_output_embeddings", "greedy", "guidance_scale", "has_default_max_length", "has_default_min_length", "head", "head_mask", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_ids_length", "input_ids_seq_length", "input_ids_shifted", "inputs", "inputs_embeds", "inputs_tensor", "int", "is", "kwargs", "kwargs_has_attention_mask", "labels", "len", "lm_heads", "lm_logits", "logits", "logits_processor", "long", "loss", "loss_fct", "mask", "masked_fill", "max_length", "min", "min_length", "mode", "model", "model_input_name", "model_kwargs", "new_embeddings", "nn", "no_grad", "nonzero", "not", "num_beams", "num_codebooks", "num_return_sequences", "of", "one", "ones", "or", "output", "output_attentions", "output_hidden_states", "output_ids", "outputs", "pad_token_id", "past_key_values", "pattern_mask", "post_init", "prefix_allowed_tokens_fn", "prepare_inputs_for_generation", "put", "r", "raise", "range", "repeat", "repeat_interleave", "requires_attention_mask", "reshape", "return", "return_dict", "return_dict_in_generate", "sampling", "search", "self", "seq_len", "sequences", "set_decoder", "set_input_embeddings", "set_output_embeddings", "setting", "shape", "shift_tokens_right", "should", "stack", "start_ids", "staticmethod", "stopping_criteria", "streamer", "super", "synced_gpus", "that", "to", "torch", "tril", "triu", "tuple", "update", "use_cache", "use_return_dict", "validate", "value", "view", "vocab_size", "where", "zeros", "zeros_like"], "musicgen_melody/modeling_musicgen_melody.py:MusicgenMelodyForConditionalGeneration": ["Any", "AutoConfig", "AutoModel", "AutoModelForTextEncoding", "BaseStreamer", "BoolTensor", "Cache", "ClassifierFreeGuidanceLogitsProcessor", "Config", "Cross", "Decoder", "Either", "EncoderDecoderModel", "Ensure", "False", "FloatTensor", "GREEDY_SEARCH", "GenerationConfig", "GenerationMixin", "GenerationMode", "Got", "Head", "If", "In", "Initializing", "LM", "Linear", "LogitsProcessorList", "LongTensor", "Melody", "Model", "ModelConfig", "ModelForCausalLM", "ModelForConditionalGeneration", "ModelOutputWithPast", "None", "NotImplementedError", "Optional", "Please", "PreTrainedModel", "Resizing", "SAMPLE", "StoppingCriteriaList", "Tensor", "The", "True", "Union", "ValueError", "__init__", "_attn_implementation", "_bos_token_tensor", "_decoder_start_token_tensor", "_dynamic_tied_weights_keys", "_expand_inputs_for_generation", "_from_config", "_get_decoder_start_token_id", "_get_logits_processor", "_get_stopping_criteria", "_hf_hook", "_init_weights", "_maybe_initialize_input_ids_for_generation", "_modules", "_pad_token_tensor", "_prepare_attention_mask_for_generation", "_prepare_cache_for_generation", "_prepare_decoder_input_ids_for_generation", "_prepare_encoder_hidden_states_kwargs_for_generation", "_prepare_generated_length", "_prepare_model_inputs", "_prepare_special_tokens", "_requires_grad", "_sample", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_tie_encoder_decoder_weights", "_validate_generated_length", "_validate_model_kwargs", "a", "activated", "add_cross_attention", "added", "all", "allows", "an", "and", "any", "append", "apply_delay_pattern_mask", "architecture", "are", "args", "argument", "as", "attention", "attention_mask", "attentions", "attributes", "audio", "audio_channels", "audio_enc_to_dec_proj", "audio_encoder", "audio_encoder_", "audio_encoder_model", "audio_encoder_pretrained_model_name_or_path", "audio_hidden_states", "audio_scales", "audio_values", "auto_docstring", "base_model_prefix", "batch_size", "be", "beam", "bias", "bool", "bos_token_id", "break", "build_delay_pattern_mask", "by", "cat", "causal", "ceil", "chroma", "chroma_length", "class", "classmethod", "cls", "codec_outputs_left", "codec_outputs_right", "concatenate", "conditional", "config", "config_class", "configuration", "copy", "cpu", "cross", "data", "de", "decode", "decoder", "decoder_", "decoder_attention_mask", "decoder_base_model_prefix", "decoder_config", "decoder_delay_pattern_mask", "decoder_head_mask", "decoder_input_ids", "decoder_input_ids_start", "decoder_inputs_embeds", "decoder_model", "decoder_outputs", "decoder_pretrained_model_name_or_path", "decoder_start_token_id", "deepcopy", "def", "defined", "del", "device", "dict", "dim", "directly", "disabled", "do", "dtype", "duration", "elif", "else", "embedding", "enc_to_dec_proj", "encoder", "encoder_accepts_wildcard", "encoder_attention_mask", "encoder_config", "encoder_hidden_states", "encoder_input_ids", "encoder_kwargs", "encoder_outputs", "encoder_signature", "exceeds", "expand_size", "f", "for", "forward", "frames", "freeze_audio_encoder", "freeze_text_encoder", "from", "from_config", "from_pretrained", "from_sub_models_config", "from_sub_models_pretrained", "generate", "generation", "generation_config", "generation_mode", "get", "get_encoder", "get_generation_mode", "get_input_embeddings", "get_output_embeddings", "get_seq_length", "get_text_encoder", "greedy", "guidance_scale", "has", "has_default_max_length", "has_default_min_length", "hasattr", "have", "hidden_size", "hidden_states", "if", "in", "incompatible", "info", "initialize", "initialized", "initializer_factor", "input_features", "input_ids", "input_ids_length", "input_ids_seq_length", "inputs", "inputs_embeds", "inputs_tensor", "inspect", "int", "io_same_device", "irrelevant_prefix", "is", "is_decoder", "is_encoder_decoder", "isinstance", "item", "items", "key", "kwargs", "kwargs_audio_encoder", "kwargs_decoder", "kwargs_has_attention_mask", "kwargs_text_encoder", "labels", "last_hidden_state", "layers", "len", "length", "list", "logger", "logits", "logits_processor", "long", "loss", "main_input_name", "make", "mask", "math", "max_cache_length", "max_length", "maximum", "mean", "methods", "min_length", "mode", "model", "model_args", "model_input_name", "model_kwargs", "module", "n_repeat", "new_embeddings", "nn", "no", "no_grad", "normal_", "not", "null_audio_hidden_states", "num_beams", "num_chroma", "num_codebooks", "num_return_sequences", "objects", "of", "one", "ones", "ones_like", "or", "order", "output_attentions", "output_hidden_states", "output_ids", "output_values", "output_values_left", "output_values_right", "outputs", "p", "pad_token_id", "param", "parameters", "pass", "passed", "past_key_values", "past_length", "pop", "post_init", "prefix_allowed_tokens_fn", "prepare_decoder_input_ids_from_labels", "prepare_inputs_for_generation", "provided", "put", "r", "raise", "randomly", "remove_prefix_length", "repeat", "requires_attention_mask", "requires_grad", "reshape", "resize_token_embeddings", "respective", "return", "return_dict", "return_dict_in_generate", "return_unused_kwargs", "s", "sampling", "search", "self", "sequences", "set", "set_output_embeddings", "setting", "shape", "shift_tokens_right", "should", "signal", "signature", "startswith", "std", "stopping_criteria", "str", "streamer", "super", "supported", "supports_gradient_checkpointing", "sure", "synced_gpus", "text", "text_encoder", "text_encoder_", "text_encoder_model", "text_encoder_pretrained_model_name_or_path", "that", "the", "three", "tie_encoder_decoder", "tie_weights", "tied_weights", "to", "torch", "truncated", "tuple", "type", "update", "use", "use_cache", "use_return_dict", "validate", "value", "values", "via", "warning", "weight", "when", "which", "will", "without", "wrapped", "zero_", "zeros", "zeros_like"], "cvt/modeling_cvt.py:BaseModelOutputWithCLSToken": ["FloatTensor", "ModelModelOutputWithCLSToken", "ModelOutput", "None", "Optional", "class", "cls_token_value", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "cvt/modeling_cvt.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "cvt/modeling_cvt.py:CvtDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "cvt/modeling_cvt.py:CvtEmbeddings": ["Dropout", "ModelConvEmbeddings", "ModelEmbeddings", "Module", "__init__", "class", "convolution_embeddings", "def", "dropout", "dropout_rate", "embed_dim", "forward", "hidden_state", "nn", "num_channels", "padding", "patch_size", "pixel_values", "return", "self", "stride", "super"], "cvt/modeling_cvt.py:CvtConvEmbeddings": ["Conv2d", "Iterable", "LayerNorm", "ModelConvEmbeddings", "Module", "__init__", "abc", "batch_size", "class", "collections", "def", "else", "embed_dim", "forward", "height", "hidden_size", "if", "isinstance", "kernel_size", "nn", "normalization", "num_channels", "padding", "patch_size", "permute", "pixel_values", "projection", "return", "self", "shape", "stride", "super", "view", "width"], "cvt/modeling_cvt.py:CvtSelfAttentionConvProjection": ["BatchNorm2d", "Conv2d", "False", "ModelSelfAttentionConvProjection", "Module", "__init__", "bias", "class", "convolution", "def", "embed_dim", "forward", "groups", "hidden_state", "kernel_size", "nn", "normalization", "padding", "return", "self", "stride", "super"], "cvt/modeling_cvt.py:CvtSelfAttentionLinearProjection": ["ModelSelfAttentionLinearProjection", "Module", "batch_size", "class", "def", "forward", "height", "hidden_size", "hidden_state", "nn", "num_channels", "permute", "return", "self", "shape", "view", "width"], "cvt/modeling_cvt.py:CvtSelfAttentionProjection": ["ModelSelfAttentionConvProjection", "ModelSelfAttentionLinearProjection", "ModelSelfAttentionProjection", "Module", "__init__", "class", "convolution_projection", "def", "dw_bn", "embed_dim", "forward", "hidden_state", "if", "kernel_size", "linear_projection", "nn", "padding", "projection_method", "return", "self", "stride", "super"], "cvt/modeling_cvt.py:CvtSelfAttention": ["Dropout", "Linear", "ModelSelfAttention", "ModelSelfAttentionProjection", "Module", "True", "_", "__init__", "attention_drop_rate", "attention_probs", "attention_score", "avg", "batch_size", "bhlk", "bhlt", "bhlv", "bhtk", "bhtv", "bias", "cat", "class", "cls_token", "context", "contiguous", "convolution_projection_key", "convolution_projection_query", "convolution_projection_value", "def", "dim", "dropout", "einsum", "else", "embed_dim", "forward", "functional", "head_dim", "height", "hidden_size", "hidden_state", "if", "kernel_size", "key", "kwargs", "linear", "nn", "num_channels", "num_heads", "padding_kv", "padding_q", "permute", "projection_key", "projection_method", "projection_query", "projection_value", "qkv_bias", "qkv_projection_method", "query", "rearrange_for_multi_head_attention", "return", "scale", "self", "shape", "softmax", "split", "stride_kv", "stride_q", "super", "torch", "value", "view", "width", "with_cls_token"], "cvt/modeling_cvt.py:CvtSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "__init__", "class", "def", "dense", "drop_rate", "dropout", "embed_dim", "forward", "hidden_state", "input_tensor", "nn", "return", "self", "super"], "cvt/modeling_cvt.py:CvtAttention": ["ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "True", "__init__", "all_head_size", "attention", "attention_drop_rate", "attention_head_size", "attention_output", "class", "def", "dense", "dim", "drop_rate", "embed_dim", "find_pruneable_heads_and_indices", "forward", "heads", "height", "hidden_state", "if", "index", "kernel_size", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "padding_kv", "padding_q", "prune_heads", "prune_linear_layer", "pruned_heads", "qkv_bias", "qkv_projection_method", "query", "return", "self", "self_output", "set", "stride_kv", "stride_q", "super", "union", "value", "width", "with_cls_token"], "cvt/modeling_cvt.py:CvtIntermediate": ["GELU", "Linear", "ModelIntermediate", "Module", "__init__", "activation", "class", "def", "dense", "embed_dim", "forward", "hidden_state", "int", "mlp_ratio", "nn", "return", "self", "super"], "cvt/modeling_cvt.py:CvtOutput": ["Dropout", "Linear", "ModelOutput", "Module", "__init__", "class", "def", "dense", "drop_rate", "dropout", "embed_dim", "forward", "hidden_state", "input_tensor", "int", "mlp_ratio", "nn", "return", "self", "super"], "cvt/modeling_cvt.py:CvtLayer": ["Identity", "LayerNorm", "ModelAttention", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "Module", "True", "__init__", "attention", "attention_drop_rate", "attention_output", "class", "def", "drop_path", "drop_path_rate", "drop_prob", "drop_rate", "else", "embed_dim", "forward", "height", "hidden_state", "if", "intermediate", "kernel_size", "layer_output", "layernorm_after", "layernorm_before", "mlp_ratio", "nn", "num_heads", "output", "padding_kv", "padding_q", "qkv_bias", "qkv_projection_method", "return", "self", "self_attention_output", "stride_kv", "stride_q", "super", "width", "with_cls_token"], "cvt/modeling_cvt.py:CvtStage": ["ModelEmbeddings", "ModelLayer", "ModelStage", "Module", "None", "Parameter", "Sequential", "_", "__init__", "attention_drop_rate", "batch_size", "cat", "class", "cls_token", "config", "cpu", "def", "depth", "device", "dim", "drop_path_rate", "drop_path_rates", "drop_rate", "dropout_rate", "else", "embed_dim", "embedding", "expand", "for", "forward", "height", "hidden_state", "if", "in", "item", "kernel_qkv", "kernel_size", "layer", "layer_outputs", "layers", "linspace", "mlp_ratio", "nn", "num_channels", "num_heads", "padding", "padding_kv", "padding_q", "patch_padding", "patch_size", "patch_sizes", "patch_stride", "permute", "qkv_bias", "qkv_projection_method", "randn", "range", "return", "self", "shape", "split", "stage", "stride", "stride_kv", "stride_q", "super", "torch", "view", "width", "with_cls_token", "x"], "cvt/modeling_cvt.py:CvtEncoder": ["BaseModelOutputWithCLSToken", "False", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "append", "class", "cls_token", "cls_token_value", "config", "def", "depth", "else", "enumerate", "for", "forward", "hidden_state", "hidden_states", "if", "in", "is", "last_hidden_state", "len", "nn", "not", "output_hidden_states", "pixel_values", "range", "return", "return_dict", "self", "stage_idx", "stage_module", "stages", "super", "tuple", "v"], "cvt/modeling_cvt.py:CvtPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLayer", "ModelPreTrainedModel", "ModelStage", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "not", "pixel_values", "self", "stage", "std", "trunc_normal_", "weight", "zero_"], "cvt/modeling_cvt.py:CvtModel": ["BaseModelOutputWithCLSToken", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "class", "cls_token_value", "config", "def", "else", "encoder", "encoder_outputs", "for", "forward", "have", "heads", "heads_to_prune", "hidden_states", "if", "in", "is", "items", "last_hidden_state", "layer", "not", "output_hidden_states", "pixel_values", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "cvt/modeling_cvt.py:CvtForImageClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "Identity", "ImageClassifierOutputWithNoAttention", "LayerNorm", "Linear", "MSELoss", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "auto_docstring", "batch_size", "bool", "class", "classifier", "cls_token", "config", "def", "dim", "dtype", "elif", "else", "embed_dim", "forward", "height", "hidden_states", "if", "int", "is", "labels", "layernorm", "logits", "long", "loss", "loss_fct", "mean", "multi_label_classification", "nn", "not", "num_channels", "num_labels", "or", "output", "output_hidden_states", "outputs", "permute", "pixel_values", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "sequence_output_mean", "shape", "single_label_classification", "squeeze", "super", "torch", "tuple", "use_return_dict", "view", "width"], "dinat/modeling_dinat.py:DinatEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "reshaped_hidden_states", "torch", "tuple"], "dinat/modeling_dinat.py:DinatModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "reshaped_hidden_states", "torch", "tuple"], "dinat/modeling_dinat.py:DinatImageClassifierOutput": ["FloatTensor", "ModelImageClassifierOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "dinat/modeling_dinat.py:DinatEmbeddings": ["Dropout", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "Optional", "Tensor", "__init__", "class", "config", "def", "dropout", "embed_dim", "embeddings", "forward", "hidden_dropout_prob", "nn", "norm", "patch_embeddings", "pixel_values", "return", "self", "super", "torch", "tuple"], "dinat/modeling_dinat.py:DinatPatchEmbeddings": ["Conv2d", "FloatTensor", "Make", "Model", "ModelPatchEmbeddings", "Module", "Optional", "Sequential", "Tensor", "ValueError", "_", "__init__", "at", "channel", "class", "config", "configuration", "def", "dimension", "else", "embed_dim", "embeddings", "forward", "height", "hidden_size", "if", "in", "kernel_size", "match", "moment", "nn", "num_channels", "of", "one", "only", "padding", "pass", "patch", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "supports", "sure", "that", "the", "torch", "values", "width", "with"], "dinat/modeling_dinat.py:DinatDownsampler": ["Conv2d", "False", "LayerNorm", "ModelDownsampler", "Module", "None", "Tensor", "__init__", "bias", "class", "def", "dim", "forward", "input_feature", "int", "kernel_size", "nn", "norm", "norm_layer", "padding", "permute", "reduction", "return", "self", "stride", "super", "torch"], "dinat/modeling_dinat.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "dinat/modeling_dinat.py:DinatDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "dinat/modeling_dinat.py:NeighborhoodAttention": ["Dropout", "False", "Linear", "ModelAttention", "Module", "Optional", "Parameter", "Tensor", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dilation", "dim", "dropout", "else", "f", "forward", "functional", "heads", "hidden", "hidden_states", "if", "int", "is", "kernel_size", "key", "key_layer", "math", "multiple", "natten2dav", "natten2dqkrpb", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "rpb", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "view", "zeros"], "dinat/modeling_dinat.py:NeighborhoodAttentionOutput": ["Dropout", "Linear", "ModelAttentionOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "dinat/modeling_dinat.py:NeighborhoodAttentionModule": ["False", "ModelAttention", "ModelAttentionModule", "ModelAttentionOutput", "Module", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dilation", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_states", "if", "index", "kernel_size", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "dinat/modeling_dinat.py:DinatIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "dinat/modeling_dinat.py:DinatOutput": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "dinat/modeling_dinat.py:DinatLayer": ["False", "Identity", "LayerNorm", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "Module", "NeighborhoodAttentionModule", "None", "Optional", "Parameter", "Tensor", "True", "_", "__init__", "attention", "attention_output", "attention_outputs", "batch_size", "bool", "channels", "chunk_size_feed_forward", "class", "config", "contiguous", "def", "dilation", "dim", "drop_path", "drop_path_rate", "else", "eps", "forward", "functional", "height", "height_pad", "hidden_states", "if", "intermediate", "is", "kernel_size", "layer_norm_eps", "layer_output", "layer_outputs", "layer_scale_init_value", "layer_scale_parameters", "layernorm_after", "layernorm_before", "max", "maybe_pad", "nn", "not", "num_heads", "ones", "or", "output", "output_attentions", "pad", "pad_b", "pad_l", "pad_r", "pad_t", "pad_values", "requires_grad", "return", "self", "shape", "shortcut", "size", "super", "torch", "tuple", "was_padded", "width", "width_pad", "window_size"], "dinat/modeling_dinat.py:DinatStage": ["False", "LayerNorm", "ModelLayer", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "bool", "class", "config", "def", "depth", "dilation", "dilations", "dim", "downsample", "drop_path_rate", "else", "enumerate", "for", "forward", "height", "hidden_states", "hidden_states_before_downsampling", "i", "if", "in", "is", "layer_module", "layer_outputs", "layers", "nn", "norm_layer", "not", "num_heads", "output_attentions", "pointing", "range", "return", "self", "size", "stage_outputs", "super", "torch", "tuple", "width"], "dinat/modeling_dinat.py:DinatEncoder": ["False", "ModelDownsampler", "ModelEncoder", "ModelEncoderOutput", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "and", "attentions", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "dilations", "dim", "downsample", "dpr", "drop_path_rate", "elif", "else", "embed_dim", "enumerate", "for", "forward", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "int", "is", "item", "last_hidden_state", "layer_module", "layer_outputs", "len", "levels", "linspace", "nn", "not", "num_heads", "num_levels", "output_attentions", "output_hidden_states", "output_hidden_states_before_downsampling", "permute", "range", "reshaped_hidden_state", "reshaped_hidden_states", "return", "return_dict", "self", "sum", "super", "torch", "tuple", "v", "x"], "dinat/modeling_dinat.py:DinatPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "dinat/modeling_dinat.py:DinatModel": ["AdaptiveAvgPool1d", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flatten", "for", "forward", "get_input_embeddings", "have", "heads", "heads_to_prune", "hidden_states", "if", "in", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "len", "natten", "nn", "not", "num_features", "num_levels", "output", "output_attentions", "output_hidden_states", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "requires_backends", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "tuple", "use_return_dict"], "dinat/modeling_dinat.py:DinatForImageClassification": ["FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelImageClassifierOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "natten", "nn", "not", "num_features", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "post_init", "r", "requires_backends", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "dinat/modeling_dinat.py:DinatBackbone": ["BackboneMixin", "BackboneOutput", "LayerNorm", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "ModuleDict", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "_out_features", "attentions", "auto_docstring", "batch_size", "bool", "channels", "class", "config", "contiguous", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "get_input_embeddings", "height", "hidden_state", "hidden_states", "hidden_states_norms", "i", "if", "in", "int", "is", "len", "natten", "nn", "not", "num_channels", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "output_hidden_states_before_downsampling", "outputs", "patch_embeddings", "permute", "pixel_values", "post_init", "r", "range", "requires_backends", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "stage", "stage_names", "super", "torch", "use_return_dict", "view", "width", "zip"], "moonshine/modeling_moonshine.py:MoonshineEncoderMLP": ["ACT2FN", "Linear", "ModelEncoderMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "moonshine/modeling_moonshine.py:MoonshineDecoderMLP": ["ACT2FN", "Linear", "ModelDecoderMLP", "Module", "Tensor", "__init__", "activation_fn", "chunk", "class", "config", "def", "dim", "fc1", "fc2", "forward", "gate", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "moonshine/modeling_moonshine.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "moonshine/modeling_moonshine.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "moonshine/modeling_moonshine.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "moonshine/modeling_moonshine.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "repeat_interleave", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "moonshine/modeling_moonshine.py:MoonshineAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "cross_attention_cache", "current_states", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "functional", "get", "getattr", "head_dim", "head_dim_padding", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_updated", "k_proj", "key_states", "key_value_states", "keys", "kwargs", "layer_idx", "layers", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "pad", "pad_head_dim_to_multiple_of", "past_key_value", "past_key_values", "position_embeddings", "q_len", "q_proj", "query_states", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "sin", "super", "target_head_dim", "target_multiple", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "moonshine/modeling_moonshine.py:MoonshineRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "moonshine/modeling_moonshine.py:MoonshineEncoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelEncoderMLP", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bias", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "encoder_hidden_act", "encoder_num_attention_heads", "encoder_num_key_value_heads", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "is_causal", "kwargs", "layer_idx", "mlp", "new_name", "nn", "num_attention_heads", "num_key_value_heads", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "moonshine/modeling_moonshine.py:MoonshineDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelDecoderMLP", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "_", "__init__", "attention_mask", "bias", "bool", "cache_position", "class", "config", "decoder_hidden_act", "decoder_num_attention_heads", "decoder_num_key_value_heads", "def", "deprecate_kwarg", "encoder_attention_mask", "encoder_attn", "encoder_hidden_states", "encoder_position_embeddings", "encoder_position_ids", "final_layernorm", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is", "is_causal", "key_value_states", "kwargs", "layer_idx", "mlp", "new_name", "nn", "not", "num_attention_heads", "num_key_value_heads", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "moonshine/modeling_moonshine.py:MoonshinePreTrainedModel": ["LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_get_feat_extract_output_lengths", "_no_split_modules", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "def", "input_lengths", "input_values", "int", "main_input_name", "model", "output_conv1_length", "output_conv2_length", "output_conv3_length", "return", "self", "supports_gradient_checkpointing", "torch"], "moonshine/modeling_moonshine.py:MoonshineEncoder": ["BaseModelOutputWithPast", "Conv1d", "False", "FloatTensor", "GroupNorm", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelRotaryEmbedding", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "_can_record_outputs", "_get_feat_extract_output_lengths", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "any", "arange", "attention_mask", "attentions", "bias", "check_model_inputs", "class", "config", "conv1", "conv2", "conv3", "def", "device", "downsample_stride", "dtype", "elif", "else", "embed_dim", "encoder_layer", "encoder_num_hidden_layers", "eps", "flash_attention_2", "for", "forward", "functional", "gelu", "get_input_embeddings", "gradient_checkpointing", "groupnorm", "hidden_size", "hidden_states", "idx", "if", "in", "input_values", "is", "kernel_size", "kwargs", "last_hidden_state", "layer_norm", "layers", "main_input_name", "mask_len", "nn", "not", "num_channels", "num_groups", "permute", "position_embeddings", "position_ids", "post_init", "r", "range", "return", "rotary_emb", "sdpa", "self", "set_input_embeddings", "shape", "stride", "super", "tanh", "torch", "unsqueeze", "value"], "moonshine/modeling_moonshine.py:MoonshineDecoder": ["BaseModelOutputWithPast", "BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_can_record_outputs", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "and", "any", "arange", "attention_mask", "attentions", "bias", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "cross_attentions", "decoder_layer", "decoder_num_hidden_layers", "def", "device", "downsample_stride", "dtype", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attn", "encoder_hidden_states", "exactly", "flash_attention_2", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "idx", "if", "in", "index", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_name", "layers", "main_input_name", "mask_len", "must", "nn", "norm", "not", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rotary_emb", "sdpa", "self", "self_attn", "shape", "specify", "super", "torch", "tuple", "unsqueeze", "use_cache", "vocab_size"], "moonshine/modeling_moonshine.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "moonshine/modeling_moonshine.py:MoonshineModel": ["BaseModelOutput", "BaseModelOutputWithPastAndCrossAttentions", "EncoderDecoderCache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqModelOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_compute_mask_indices", "_freeze_parameters", "_mask_input_features", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_position_ids", "def", "device", "dtype", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "expand", "forward", "freeze_encoder", "get_encoder", "get_input_embeddings", "getattr", "hidden_size", "hidden_states", "if", "input_features", "input_ids", "input_values", "inputs_embeds", "is", "kwargs", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "min_masks", "not", "past_key_values", "position_ids", "post_init", "r", "return", "self", "sequence_length", "set_input_embeddings", "size", "super", "tensor", "torch", "training", "tuple", "use_cache", "value"], "moonshine/modeling_moonshine.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "moonshine/modeling_moonshine.py:MoonshineForConditionalGeneration": ["EncoderDecoderCache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Seq2SeqLMOutput", "Seq2SeqModelOutput", "TransformersKwargs", "Union", "Unpack", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_position_ids", "decoder_start_token_id", "def", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_encoder", "get_input_embeddings", "get_output_embeddings", "hidden_size", "if", "input_values", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "model", "new_embeddings", "nn", "not", "outputs", "pad_token_id", "past_key_values", "post_init", "proj_out", "r", "return", "self", "set_output_embeddings", "shift_tokens_right", "super", "torch", "tuple", "use_cache", "vocab_size", "weight"], "aya_vision/modeling_aya_vision.py:AyaVisionMultiModalProjector": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "True", "__init__", "act", "adapter_layer_norm_eps", "alignment_intermediate_size", "batch_size", "bias", "channels", "chunk", "class", "config", "def", "dim", "downsample_factor", "eps", "feature_dim", "forward", "gate", "getattr", "height", "hidden_size", "hidden_states", "image_features", "int", "layernorm", "linear_1", "linear_2", "nn", "permute", "pixel_shuffle", "reshape", "return", "self", "seq_length", "shape", "silu", "super", "text_config", "vision_config", "width", "x"], "aya_vision/modeling_aya_vision.py:AyaVisionPreTrainedModel": ["Attention", "DecoderLayer", "False", "ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "past_key_values", "supports_gradient_checkpointing"], "aya_vision/modeling_aya_vision.py:AyaVisionCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "aya_vision/modeling_aya_vision.py:AyaVisionModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "aya_vision/modeling_aya_vision.py:AyaVisionModel": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unexpected", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "cat", "check_model_inputs", "class", "config", "decoder", "def", "default", "device", "dim", "do", "dtype", "else", "exactly", "expand_as", "f", "feature", "features", "for", "forward", "from_config", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "hs", "hs_pool", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "k", "kwargs", "language_model", "last_hidden_state", "layer_idx", "list", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "raise", "return", "select", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "str", "strategy", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "v", "value", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower"], "aya_vision/modeling_aya_vision.py:AyaVisionForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "str", "super", "text_config", "torch", "tuple", "value", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "detr/modeling_detr.py:DetrDecoderOutput": ["BaseModelOutputWithCrossAttentions", "FloatTensor", "ModelDecoderOutput", "None", "Optional", "class", "intermediate_hidden_states", "r", "torch"], "detr/modeling_detr.py:DetrModelOutput": ["FloatTensor", "ModelModelOutput", "None", "Optional", "Seq2SeqModelOutput", "class", "intermediate_hidden_states", "r", "torch"], "detr/modeling_detr.py:DetrObjectDetectionOutput": ["FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "detr/modeling_detr.py:DetrSegmentationOutput": ["FloatTensor", "ModelOutput", "ModelSegmentationOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "pred_masks", "r", "torch", "tuple"], "detr/modeling_detr.py:DetrFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "detr/modeling_detr.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "detr/modeling_detr.py:DetrConvEncoder": ["Either", "False", "ModelConvEncoder", "Module", "None", "Tensor", "True", "ValueError", "__init__", "and", "append", "backbone", "backbone_config", "backbone_kwargs", "backbone_model_type", "be", "bool", "channels", "class", "config", "copy", "create_model", "def", "dilation", "elif", "else", "feature_info", "feature_map", "feature_maps", "features", "features_only", "float", "for", "forward", "functional", "get", "getattr", "if", "in", "in_chans", "intermediate_channel_sizes", "interpolate", "is", "kwargs", "layer2", "layer3", "layer4", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "num_channels", "or", "out", "out_indices", "output_stride", "parameter", "pixel_mask", "pixel_values", "pop", "pretrained", "provided", "raise", "replace_batch_norm", "requires_backends", "requires_grad_", "resnet", "return", "self", "shape", "should", "size", "stage", "super", "the", "timm", "to", "torch", "use_pretrained_backbone", "use_timm_backbone", "with"], "detr/modeling_detr.py:DetrConvModel": ["ModelConvModel", "Module", "__init__", "append", "class", "conv_encoder", "def", "dtype", "feature_map", "for", "forward", "in", "mask", "nn", "out", "pixel_mask", "pixel_values", "pos", "position_embedding", "return", "self", "super", "to"], "detr/modeling_detr.py:DetrSinePositionEmbedding": ["False", "ModelSinePositionEmbedding", "Module", "No", "None", "True", "ValueError", "__init__", "and", "arange", "be", "cat", "class", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "flatten", "float", "float32", "floor", "forward", "if", "int64", "is", "mask", "math", "nn", "normalize", "not", "passed", "permute", "pi", "pixel", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "provided", "raise", "return", "rounding_mode", "scale", "self", "should", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "detr/modeling_detr.py:DetrLearnedPositionEmbedding": ["Embedding", "ModelLearnedPositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "detr/modeling_detr.py:build_position_encoding": ["ModelLearnedPositionEmbedding", "ModelSinePositionEmbedding", "Model_position_encoding", "Not", "True", "ValueError", "config", "d_model", "def", "elif", "else", "f", "if", "learned", "n_steps", "normalize", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "detr/modeling_detr.py:DetrAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "is_cross_attention", "k_proj", "key_states", "key_value_states", "key_value_states_original", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "object_queries", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "spatial_position_embeddings", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "detr/modeling_detr.py:DetrEncoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "max", "min", "nn", "num_heads", "object_queries", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "detr/modeling_detr.py:DetrDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "key_value_states", "nn", "not", "num_heads", "object_queries", "output_attentions", "outputs", "p", "query_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_position_embeddings", "super", "torch", "training"], "detr/modeling_detr.py:DetrPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "Linear", "ModelConfig", "ModelConvEncoder", "ModelDecoderLayer", "ModelEncoderLayer", "ModelLearnedPositionEmbedding", "ModelMHAttentionMap", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "column_embeddings", "config", "data", "def", "elif", "gain", "if", "init", "init_std", "init_xavier_std", "is", "isinstance", "k_linear", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "q_linear", "r", "row_embeddings", "self", "std", "uniform_", "weight", "xavier_std", "xavier_uniform_", "zero_", "zeros_"], "detr/modeling_detr.py:DetrEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "True", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "for", "forward", "functional", "hidden_states", "i", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layers", "nn", "not", "object_queries", "output_attentions", "output_hidden_states", "p", "post_init", "r", "rand", "range", "return", "return_dict", "self", "super", "to_drop", "torch", "training", "tuple", "use_return_dict", "v"], "detr/modeling_detr.py:DetrDecoder": ["False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "_prepare_4d_attention_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attention_mask", "attentions", "auxiliary_loss", "class", "combined_attention_mask", "config", "continue", "cross_attentions", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "input_shape", "inputs_embeds", "intermediate", "intermediate_hidden_states", "is", "last_hidden_state", "layer_outputs", "layerdrop", "layernorm", "layers", "nn", "not", "object_queries", "output_attentions", "output_hidden_states", "post_init", "query_position_embeddings", "r", "rand", "range", "return", "return_dict", "self", "size", "stack", "super", "tgt_len", "torch", "training", "tuple", "use_return_dict", "v"], "detr/modeling_detr.py:DetrModel": ["Backbone", "BaseModelOutput", "Conv2d", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelConvEncoder", "ModelConvModel", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_size", "bool", "build_position_encoding", "class", "config", "conv_encoder", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "device", "does", "downsampled", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "feature_map", "features", "flatten", "flattened_features", "flattened_mask", "for", "forward", "freeze_backbone", "get_encoder", "height", "hidden_states", "if", "in", "input_projection", "inputs_embeds", "intermediate_channel_sizes", "intermediate_hidden_states", "is", "isinstance", "kernel_size", "last_hidden_state", "len", "mask", "model", "name", "named_parameters", "nn", "not", "num_channels", "num_queries", "object_queries", "object_queries_list", "ones", "output_attentions", "output_hidden_states", "param", "permute", "pixel", "pixel_mask", "pixel_values", "post_init", "projected_feature_map", "queries", "query_position_embeddings", "r", "raise", "repeat", "requires_grad_", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "unfreeze_backbone", "unsqueeze", "use_return_dict", "weight", "width", "zeros_like"], "detr/modeling_detr.py:DetrMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "detr/modeling_detr.py:DetrForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForObjectDetection", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "bbox_predictor", "bool", "class", "class_labels_classifier", "config", "cross_attentions", "d_model", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "def", "device", "dict", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "hidden_dim", "if", "input_dim", "inputs_embeds", "intermediate", "intermediate_hidden_states", "is", "labels", "last_hidden_state", "list", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_labels", "num_layers", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "outputs_class", "outputs_coord", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "r", "return", "return_dict", "self", "sequence_output", "sigmoid", "super", "torch", "tuple", "use_return_dict"], "detr/modeling_detr.py:DetrForSegmentation": ["BaseModelOutput", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelForObjectDetection", "ModelForSegmentation", "ModelMHAttentionMap", "ModelMaskHeadSmallConv", "ModelPreTrainedModel", "ModelSegmentationOutput", "None", "Optional", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "backbone", "batch_size", "bbox_attention", "bbox_mask", "bbox_predictor", "bool", "class", "class_labels_classifier", "config", "conv_encoder", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "device", "dict", "dropout", "elif", "else", "encoder", "encoder_attention_heads", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "feature_map", "features", "flatten", "flattened_features", "flattened_mask", "forward", "height", "hidden_size", "hidden_states", "if", "init_xavier_std", "input_projection", "inputs_embeds", "intermediate", "intermediate_channel_sizes", "intermediate_hidden_states", "is", "isinstance", "labels", "last_hidden_state", "len", "list", "logits", "loss", "loss_dict", "loss_function", "mask", "mask_head", "memory", "model", "not", "num_channels", "num_queries", "number_of_heads", "object_queries", "object_queries_list", "ones", "output", "output_attentions", "output_hidden_states", "outputs_class", "outputs_coord", "permute", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "pred_masks", "projected_feature_map", "queries", "query_position_embeddings", "r", "repeat", "return", "return_dict", "seg_masks", "self", "sequence_output", "shape", "sigmoid", "std", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "view", "weight", "width", "zeros_like"], "detr/modeling_detr.py:_expand": ["_expand", "def", "flatten", "int", "length", "repeat", "return", "tensor", "unsqueeze"], "detr/modeling_detr.py:DetrMaskHeadSmallConv": ["Conv2d", "GroupNorm", "ModelMaskHeadSmallConv", "Module", "Tensor", "The", "ValueError", "__init__", "_expand", "a", "adapter1", "adapter2", "adapter3", "as", "attention", "bbox_mask", "be", "bias", "by", "cat", "class", "constant_", "context_dim", "cur_fpn", "def", "dim", "divisible", "flatten", "for", "forward", "fpn_dims", "fpns", "functional", "gn1", "gn2", "gn3", "gn4", "gn5", "groups", "heads", "hidden_size", "if", "in", "init", "inter_dims", "interpolate", "is", "isinstance", "kaiming_uniform_", "lay1", "lay2", "lay3", "lay4", "lay5", "list", "m", "min", "mode", "modules", "must", "nearest", "nn", "number", "of", "out_lay", "padding", "raise", "relu", "return", "self", "set", "shape", "size", "super", "the", "to", "torch", "weight", "x"], "detr/modeling_detr.py:DetrMHAttentionMap": ["Dropout", "Linear", "ModelMHAttentionMap", "Module", "None", "Optional", "Tensor", "True", "__init__", "bias", "bnchw", "bqnc", "bqnhw", "class", "conv2d", "def", "dim", "dropout", "dtype", "einsum", "finfo", "flatten", "float", "forward", "functional", "hidden_dim", "if", "is", "k", "k_linear", "keys_per_head", "mask", "masked_fill", "min", "nn", "normalize_fact", "not", "num_heads", "q", "q_linear", "queries_per_head", "query_dim", "return", "self", "shape", "size", "softmax", "std", "super", "torch", "unsqueeze", "view", "weight", "weights"], "yoso/modeling_yoso.py:load_cuda_kernels": ["Model", "Model_cuda_kernels", "Path", "True", "__file__", "append_root", "cpp", "cu", "def", "fast_lsh_cumulation", "fast_lsh_cumulation_cuda", "fast_lsh_cumulation_torch", "file", "files", "for", "global", "in", "kernels", "lsh_cumulation", "parent", "resolve", "return", "src_files", "src_folder", "verbose"], "yoso/modeling_yoso.py:to_contiguous": ["append", "contiguous", "def", "else", "for", "if", "in", "input_tensors", "is_contiguous", "isinstance", "list", "not", "out", "return", "tensor", "to_contiguous"], "yoso/modeling_yoso.py:normalize": ["Model", "append", "def", "dim", "else", "for", "functional", "if", "in", "input_tensors", "isinstance", "list", "nn", "out", "p", "return", "tensor"], "yoso/modeling_yoso.py:hashing": ["Key", "Model", "Query", "ValueError", "arange", "def", "device", "dim", "has", "hash_len", "if", "incorrect", "int", "key", "key_binary", "key_projection", "len", "matmul", "num_hash", "query", "query_binary", "query_hash", "query_projection", "raise", "raise_pow", "randn", "reshape", "return", "rmat", "size", "sum", "torch"], "yoso/modeling_yoso.py:YosoCumulation": ["Function", "ModelCumulation", "None", "acos", "autograd", "backward", "class", "config", "ctx", "cumulation_value", "def", "expectation", "forward", "grad", "grad_key", "grad_query", "grad_value", "hash_code_len", "key", "key_mask", "math", "matmul", "pi", "query", "query_mask", "return", "save_for_backward", "saved_tensors", "staticmethod", "to_contiguous", "torch", "transpose", "value", "weighted_exp"], "yoso/modeling_yoso.py:YosoLSHCumulation": ["Function", "Key", "ModelLSHCumulation", "None", "Query", "Value", "ValueError", "acos", "and", "autograd", "backward", "class", "config", "ctx", "cumulation_value", "def", "differ", "dimension", "else", "expectation", "fast_hash", "forward", "grad", "grad_key", "grad_query", "grad_value", "hash_code_len", "hashing", "hashtable_capacity", "if", "in", "int", "is_cuda", "key", "key_hash_code", "key_mask", "lsh_backward", "lsh_cumulation", "lsh_weighted_cumulation", "mask", "math", "matmul", "num_hash", "pi", "query", "query_hash_code", "query_mask", "raise", "return", "save_for_backward", "saved_tensors", "size", "sizes", "staticmethod", "to_contiguous", "torch", "transpose", "use_cuda", "use_fast_hash", "value", "weighted_exp"], "yoso/modeling_yoso.py:YosoEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "yoso/modeling_yoso.py:YosoSelfAttention": ["Conv2d", "Could", "Dropout", "Exception", "False", "Linear", "ModelCumulation", "ModelLSHCumulation", "ModelSelfAttention", "Module", "None", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "apply", "as", "attention", "attention_head_size", "attention_mask", "attention_probs_dropout_prob", "batch_size", "bias", "cat", "class", "config", "context_layer", "contiguous", "conv", "conv_value_layer", "conv_window", "custom", "def", "deformable", "device", "dim", "dropout", "e", "else", "embedding_size", "except", "f", "for", "forward", "gpu_warp_size", "groups", "hasattr", "hash_code_len", "head_dim", "heads", "hidden", "hidden_size", "hidden_states", "if", "in_channels", "int", "is", "is_ninja_available", "is_torch_cuda_available", "kernel", "kernel_loaded", "kernel_size", "key", "key_layer", "load", "load_cuda_kernels", "logger", "lsh_backward", "lsh_config", "lsh_cumulation", "multi", "multiple", "new_context_layer_shape", "nn", "normalize", "not", "num_attention_heads", "num_hash", "num_heads", "number", "of", "or", "out_channels", "output_attentions", "outputs", "pad_size", "padding", "permute", "position_embedding_type", "query", "query_layer", "raise", "repeat_interleave", "reshape", "return", "scale", "self", "seq_len", "seq_length", "shape", "size", "super", "the", "torch", "training", "transpose", "try", "unsqueeze", "use_conv", "use_expectation", "use_fast_hash", "value", "value_layer", "view", "warning", "zeros"], "yoso/modeling_yoso.py:YosoSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "yoso/modeling_yoso.py:YosoAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "yoso/modeling_yoso.py:YosoIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "yoso/modeling_yoso.py:YosoOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "yoso/modeling_yoso.py:YosoLayer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "add_cross_attention", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "yoso/modeling_yoso.py:YosoEncoder": ["BaseModelOutputWithCrossAttentions", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "yoso/modeling_yoso.py:YosoPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "yoso/modeling_yoso.py:YosoLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "yoso/modeling_yoso.py:YosoOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "yoso/modeling_yoso.py:YosoPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "yoso/modeling_yoso.py:YosoModel": ["BaseModelOutputWithCrossAttentions", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "for", "forward", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "yoso/modeling_yoso.py:YosoForMaskedLM": ["CrossEntropyLoss", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "yoso/modeling_yoso.py:YosoClassificationHead": ["ACT2FN", "Dropout", "Linear", "ModelClassificationHead", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "x"], "yoso/modeling_yoso.py:YosoForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "yoso/modeling_yoso.py:YosoForMultipleChoice": ["CrossEntropyLoss", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "ReLU", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_state", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "pre_classifier", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "yoso/modeling_yoso.py:YosoForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "active_labels", "active_logits", "active_loss", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "ignore_index", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "tensor", "token_type_ids", "torch", "tuple", "type_as", "use_return_dict", "view", "where"], "yoso/modeling_yoso.py:YosoForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "dots1/modeling_dots1.py:Dots1RMSNorm": ["ModelRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "dots1/modeling_dots1.py:Dots1RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "dots1/modeling_dots1.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "dots1/modeling_dots1.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "dots1/modeling_dots1.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "dots1/modeling_dots1.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "dots1/modeling_dots1.py:Dots1Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "dots1/modeling_dots1.py:Dots1MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "return", "self", "super", "up_proj", "x"], "dots1/modeling_dots1.py:Dots1MoE": ["ModelMLP", "ModelMoE", "ModelTopkRouter", "Module", "ModuleList", "Tensor", "_", "__init__", "class", "config", "def", "dtype", "expert", "expert_idx", "expert_input", "expert_mask", "expert_output", "expert_weights", "experts", "final_hidden_states", "for", "forward", "functional", "gate", "hidden_states", "if", "in", "index_add_", "intermediate_size", "len", "mask", "moe", "moe_intermediate_size", "n_routed_experts", "n_shared_experts", "nn", "num_classes", "numel", "one_hot", "orig_shape", "permute", "r", "range", "residuals", "return", "self", "shape", "shared_experts", "super", "token_indices", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight_indices", "weighted_output", "where", "zeros_like"], "dots1/modeling_dots1.py:Dots1TopkRouter": ["F", "False", "ModelTopkRouter", "Module", "Parameter", "True", "__init__", "bool", "class", "config", "def", "denominator", "dim", "e_score_correction_bias", "empty", "expand", "float32", "forward", "gather", "get_topk_indices", "group_idx", "group_mask", "group_scores", "hidden_size", "hidden_states", "if", "k", "keepdim", "linear", "masked_fill", "n_group", "n_routed_experts", "nn", "no_grad", "norm_topk_prob", "num_experts_per_tok", "register_buffer", "reshape", "return", "routed_scaling_factor", "router_logits", "scatter_", "score_mask", "scores", "scores_for_choice", "self", "sigmoid", "sorted", "sum", "super", "top_k", "topk", "topk_group", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight", "zeros", "zeros_like"], "dots1/modeling_dots1.py:Dots1DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "first_k_dense_replace", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "dots1/modeling_dots1.py:Dots1PreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelTopkRouter", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "dots1/modeling_dots1.py:Dots1Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_types", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "dots1/modeling_dots1.py:Dots1ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaRotaryEmbedding": ["False", "ModelRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "autocast", "base", "cat", "class", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "if", "int64", "inv_freq", "inv_freq_expanded", "mps", "nn", "no_grad", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "self", "seq_len", "shape", "sin", "super", "tensor", "to", "torch", "transpose", "type", "with", "x"], "recurrent_gemma/modeling_recurrent_gemma.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "recurrent_gemma/modeling_recurrent_gemma.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "recurrent_gemma/modeling_recurrent_gemma.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaSdpaAttention": ["False", "Linear", "LongTensor", "ModelConfig", "ModelRotaryEmbedding", "ModelSdpaAttention", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_setup_cache", "_update_cache", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_mask", "attention_window_size", "attn_mask", "attn_output", "base", "batch_size", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "cache_shape", "cat", "causal_mask", "chunk", "clamp", "class", "config", "contiguous", "cos", "cumsum", "def", "device", "dim", "dropout_p", "dtype", "else", "float32", "forward", "functional", "get", "hasattr", "head_dim", "hidden_size", "hidden_states", "if", "indices", "int", "is", "k_out", "k_proj", "key_pass", "key_rot", "key_states", "long", "nn", "no_grad", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "ones", "partial_rotary_factor", "position_ids", "q_len", "q_proj", "query_pass", "query_rot", "query_states", "repeat_kv", "return", "rope_theta", "rotary_emb", "scale", "scaled_dot_product_attention", "self", "shape", "sin", "size", "slicing", "super", "to", "to_shift", "torch", "training", "transpose", "tuple", "use_cache", "v_out", "v_proj", "value_states", "view", "zeros"], "recurrent_gemma/modeling_recurrent_gemma.py:SqrtBoundDerivative": ["Function", "Model", "ModelBoundDerivative", "Tensor", "_MAX_Model_GRADIENT", "autograd", "backward", "class", "clip", "clipped_x_times_4", "ctx", "def", "forward", "grad_output", "min", "return", "save_for_backward", "saved_tensors", "staticmethod", "torch", "x"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaRglru": ["ModelRglru", "Model_gate", "Model_gate_bias", "Model_gate_weight", "Model_param", "Model_states", "Module", "None", "Parameter", "Proxy", "SqrtBoundDerivative", "Tensor", "Union", "__init__", "_rnn_scan", "a_square", "acc_dtype", "activations", "and", "apply", "baddbmm", "batch_size", "block_width", "class", "config", "contextualized_states", "def", "device", "dtype", "else", "empty", "exp", "float32", "for", "forward", "functional", "fx", "gated_inputs", "hidden_states", "if", "in", "input_gate", "input_gate_bias", "input_gate_weight", "is", "is_torchdynamo_compiling", "is_tracing", "isinstance", "jit", "log_Model_gate", "lru_width", "multiplier", "nn", "normalized_x", "not", "num_attention_heads", "or", "permute", "position_ids", "range", "res", "reset", "reshape", "reshape_act", "return", "self", "seq_len", "shape", "sigmoid", "softplus", "super", "t", "to", "torch", "tracing", "transpose", "tuple", "type", "zeros", "zeros_like"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaRecurrentBlock": ["ACT2FN", "Conv1d", "Linear", "ModelModelBlock", "ModelRglru", "Model_states", "Module", "None", "Tensor", "True", "_", "__init__", "_setup_cache", "act_fn", "attention_mask", "batch", "bias", "bool", "cache_position", "cat", "class", "config", "conv1d_state", "conv1d_width", "conv_1d", "conv_state", "def", "device", "dict", "dim", "dtype", "else", "float32", "forward", "functional", "groups", "hidden_activation", "hidden_size", "hidden_states", "if", "in_features", "input_states", "kernel_size", "linear_out", "linear_x", "linear_y", "lru_width", "nn", "out_features", "pad", "padding", "position_ids", "return", "rg_lru", "self", "seq_len", "shape", "str", "sum", "super", "torch", "transpose", "tuple", "unsqueeze", "use_cache", "weight", "x_branch", "y_branch", "zeros"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaMlp": ["ACT2FN", "Linear", "ModelMlp", "Module", "True", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate", "gate_proj", "hidden_activation", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "up_proj"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaDecoderLayer": ["GradientCheckpointingLayer", "ModelDecoderLayer", "ModelMlp", "ModelRMSNorm", "None", "Optional", "TEMPORAL_BLOCK_CLASSES", "Tensor", "__init__", "activations", "attention_mask", "bool", "cache_position", "channel_pre_norm", "class", "config", "def", "dict", "eps", "forward", "hidden_size", "hidden_states", "inputs_normalized", "layer_idx", "layers_block_type", "mlp_block", "position_ids", "raw_activations", "residual", "return", "rms_norm_eps", "self", "str", "super", "temporal_block", "temporal_pre_norm", "torch", "tuple", "use_cache"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaPreTrainedModel": ["Conv1d", "Embedding", "False", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelModelBlock", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRglru", "ModelSdpaAttention", "Model_gate_bias", "Model_gate_weight", "Model_param", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_setup_cache", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "batch", "bias", "cache", "class", "config", "conv1d_width", "data", "def", "device", "dtype", "elif", "exp_", "final_w_init_variance_scale", "for", "getattr", "hidden_size", "if", "in", "init", "input_gate_bias", "input_gate_weight", "is", "isinstance", "k_proj", "layer", "layers", "linear_out", "linear_x", "linear_y", "log_", "lru_width", "math", "mean", "model", "module", "mul_", "neg_", "nn", "normal_", "not", "num_attention_heads", "o_proj", "padding_idx", "pass", "q_proj", "reset_cache", "self", "sqrt", "std", "sub_", "supports_gradient_checkpointing", "temporal_block", "torch", "uniform_", "v_proj", "w_init_variance_scale", "weight", "zero_", "zeros_"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaModel": ["AttentionMaskConverter", "BaseModelOutputWithNoAttention", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_setup_cache", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "and", "arange", "attention_mask", "attention_window_size", "auto_docstring", "bfloat16", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "enumerate", "eps", "eq", "exactly", "expand", "fill_value", "final_norm", "finfo", "for", "forward", "full", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "is", "last_hidden_state", "layer_idx", "layers", "logger", "mask_length", "masked_fill", "max", "min", "min_dtype", "must", "nn", "normalizer", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "persistent", "position_ids", "post_init", "raise", "range", "register_buffer", "reshape", "residual_block", "return", "return_dict", "rms_norm_eps", "self", "sequence_length", "shape", "specify", "super", "target_length", "tensor", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "v", "vocab_size", "warning_once", "with", "xpu"], "recurrent_gemma/modeling_recurrent_gemma.py:RecurrentGemmaForCausalLM": ["CausalLMOutput", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "auto_docstring", "bias", "bool", "cache_position", "cap", "class", "config", "def", "else", "float", "forward", "functional", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "logits", "logits_soft_cap", "loss", "loss_function", "model", "nn", "not", "output", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "tanh", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "chameleon/modeling_chameleon.py:ChameleonRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "chameleon/modeling_chameleon.py:ChameleonRotaryEmbedding": ["False", "ModelRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "autocast", "base", "cat", "class", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "if", "int64", "inv_freq", "inv_freq_expanded", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "scaling_factor", "self", "shape", "sin", "super", "to", "torch", "transpose", "type", "with", "x"], "chameleon/modeling_chameleon.py:ChameleonLinearScalingRotaryEmbedding": ["ModelLinearScalingRotaryEmbedding", "ModelRotaryEmbedding", "class", "cos", "def", "float", "forward", "position_ids", "return", "scaling_factor", "self", "sin", "super", "x"], "chameleon/modeling_chameleon.py:ChameleonDynamicNTKScalingRotaryEmbedding": ["False", "ModelDynamicNTKScalingRotaryEmbedding", "ModelRotaryEmbedding", "arange", "base", "class", "cos", "def", "device", "dim", "dtype", "float", "forward", "if", "int64", "inv_freq", "max", "max_position_embeddings", "persistent", "position_ids", "register_buffer", "return", "scaling_factor", "self", "seq_len", "sin", "super", "to", "torch", "x"], "chameleon/modeling_chameleon.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "chameleon/modeling_chameleon.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "chameleon/modeling_chameleon.py:ChameleonMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "chameleon/modeling_chameleon.py:ChameleonLayerNorm": ["F", "LayerNorm", "ModelLayerNorm", "None", "__init__", "args", "bias", "class", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm", "nn", "normalized_shape", "return", "self", "super", "weight"], "chameleon/modeling_chameleon.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "chameleon/modeling_chameleon.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "chameleon/modeling_chameleon.py:ChameleonAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelDynamicNTKScalingRotaryEmbedding", "ModelLayerNorm", "ModelLinearScalingRotaryEmbedding", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "RoPE", "Tensor", "True", "Unknown", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "_init_rope", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "base", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "dynamic", "eager", "eager_attention_forward", "elif", "else", "errors", "f", "factor", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "lead", "linear", "logger", "make", "max_position_embeddings", "model_parallel_size", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "q_norm", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "rope_scaling", "rope_theta", "rotary_emb", "scaling", "scaling_factor", "scaling_type", "self", "sin", "size", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "chameleon/modeling_chameleon.py:ChameleonDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "chameleon/modeling_chameleon.py:ChameleonSwinDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelMLP", "ModelRMSNorm", "ModelSwinDecoderLayer", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "chameleon/modeling_chameleon.py:ChameleonVQVAEVectorQuantizer": ["Embedding", "ModelVQVAEVectorQuantizer", "Module", "Tensor", "True", "__init__", "argmin", "bd", "beta", "bn", "class", "config", "contiguous", "def", "detach", "dim", "distances", "dn", "einsum", "embed_dim", "embedding", "embedding_dim", "forward", "getattr", "hidden_state", "hidden_state_flattened", "hidden_state_quant", "keepdim", "loss", "mean", "min_encoding_indices", "nn", "num_embeddings", "permute", "return", "self", "shape", "sum", "super", "torch", "transpose", "view", "weight"], "chameleon/modeling_chameleon.py:ChameleonVQVAEEncoderConvDownsample": ["Conv2d", "F", "ModelVQVAEEncoderConvDownsample", "Module", "__init__", "class", "constant", "conv", "def", "forward", "hidden_states", "in_channels", "kernel_size", "mode", "nn", "pad", "padding", "return", "self", "stride", "super", "value"], "chameleon/modeling_chameleon.py:ChameleonVQVAEEncoderResnetBlock": ["Conv2d", "Dropout", "False", "GroupNorm", "ModelVQVAEEncoderResnetBlock", "Module", "None", "True", "__init__", "affine", "class", "config", "conv1", "conv2", "conv_shortcut", "def", "dropout", "else", "eps", "forward", "hidden_states", "if", "in_channels", "is", "kernel_size", "nin_shortcut", "nn", "norm1", "norm2", "num_channels", "num_groups", "out_channels", "padding", "residual", "return", "self", "sigmoid", "stride", "super", "torch", "use_conv_shortcut"], "chameleon/modeling_chameleon.py:ChameleonVQVAEEncoderAttnBlock": ["Conv2d", "F", "GroupNorm", "ModelVQVAEEncoderAttnBlock", "Module", "True", "__init__", "affine", "attn_output", "attn_weights", "batch_size", "bmm", "channels", "class", "def", "dim", "eps", "forward", "height", "hidden_states", "in_channels", "int", "k", "kernel_size", "key_states", "nn", "norm", "num_channels", "num_groups", "padding", "permute", "proj_out", "q", "query_states", "reshape", "residual", "return", "self", "shape", "softmax", "stride", "super", "torch", "v", "value_states", "width"], "chameleon/modeling_chameleon.py:ChameleonVQVAEEncoder": ["Conv2d", "GroupNorm", "Identity", "LongTensor", "ModelVQVAEEncoder", "ModelVQVAEEncoderAttnBlock", "ModelVQVAEEncoderConvDownsample", "ModelVQVAEEncoderResnetBlock", "Module", "ModuleList", "None", "True", "__init__", "affine", "and", "append", "attn", "attn_1", "attn_resolutions", "attn_type", "base_channels", "block", "block_1", "block_2", "block_in", "block_out", "channel_multiplier", "class", "config", "conv_in", "conv_out", "curr_res", "def", "double_latent", "down", "downsample", "else", "eps", "for", "forward", "hidden_state", "hidden_states", "i_block", "i_level", "if", "in", "in_channel_multiplier", "in_channels", "is", "kernel_size", "last_hidden_state", "latent_channels", "len", "mid", "nn", "norm_out", "not", "num_channels", "num_groups", "num_res_blocks", "num_resolutions", "out_channels", "padding", "pixel_values", "range", "resolution", "return", "self", "sigmoid", "stride", "super", "torch", "tuple", "vanilla"], "chameleon/modeling_chameleon.py:ChameleonImageVocabularyMapping": ["A", "IMGIMG", "ModelImageVocabularyMapping", "Tensor", "__init__", "bpe2img", "bpe2img_search_tensors", "c", "cached_property", "chr", "class", "convert_img2bpe", "cpu", "def", "device", "dtype", "for", "get", "i", "if", "image", "image_token_id", "image_tokens", "img2bpe", "img2bpe_mapping_tensor", "img_batch", "img_tkn_chr_mapping", "img_tokens", "in", "int", "items", "join", "k", "keys", "len", "mapping", "max", "name", "old_name", "ord", "range", "remap", "return", "self", "sorted", "startswith", "str", "tensor", "to", "tok", "torch", "v", "val", "val2name", "values", "vocab_map", "zeros"], "chameleon/modeling_chameleon.py:ChameleonPreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSwinDecoderLayer", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "base_model_prefix", "causal_mask", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "chameleon/modeling_chameleon.py:ChameleonVQVAE": ["Conv2d", "LongTensor", "ModelPreTrainedModel", "ModelVQVAE", "ModelVQVAEConfig", "ModelVQVAEEncoder", "ModelVQVAEEncoderAttnBlock", "ModelVQVAEEncoderResnetBlock", "ModelVQVAEVectorQuantizer", "__init__", "_no_split_modules", "class", "config", "def", "emb_loss", "embed_dim", "encode", "encoder", "eval", "hidden_states", "indices", "latent_channels", "nn", "pixel_values", "post_quant_conv", "quant", "quant_conv", "quantize", "return", "self", "super", "torch"], "chameleon/modeling_chameleon.py:ChameleonModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelImageVocabularyMapping", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelSwinDecoderLayer", "ModelVQVAE", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_from_config", "all", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "bpe_toks", "cache_position", "causal_mask", "checkpointing", "class", "config", "convert_img2bpe", "create_causal_mask", "decoder_layer", "def", "device", "do", "dtype", "else", "embed_tokens", "encode", "eps", "exactly", "expand_as", "f", "features", "for", "forward", "get_image_features", "get_image_tokens", "get_input_embeddings", "get_placeholder_mask", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "image", "image_embeds", "image_features", "image_token_id", "image_tokens", "image_toks", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "logger", "long", "masked_scatter", "match", "must", "n_image_features", "n_image_tokens", "nn", "norm", "not", "num_hidden_layers", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "pixel_values", "position_ids", "post_init", "raise", "range", "return", "return_dict", "rms_norm_eps", "self", "shape", "special_image_mask", "specify", "sum", "super", "swin_norm", "tensor", "to", "tokens", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "view", "vision_embeddings", "vocab_size", "vocabulary_map", "vocabulary_mapping", "vq_config", "vqmodel", "warning_once", "with"], "chameleon/modeling_chameleon.py:ChameleonForConditionalGeneration": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "dtype", "else", "finfo", "forward", "get_image_features", "get_image_tokens", "hidden_size", "hidden_states", "if", "image_tokens", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "logits", "loss", "loss_function", "min", "model", "model_inputs", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_cache", "vocab_size", "vocabulary_mapping", "weight"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextRMSNormGated": ["F", "ModelRMSNormGated", "Module", "None", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "gate", "hidden_size", "hidden_states", "input_dtype", "keepdim", "kwargs", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "silu", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextDynamicCache": ["Any", "False", "LongTensor", "ModelConfig", "ModelDynamicCache", "None", "Optional", "Tensor", "_", "__getitem__", "__init__", "__len__", "beam_idx", "cache_kwargs", "cache_position", "cat", "class", "config", "conv_states", "def", "device", "dict", "dim", "else", "for", "full_attention", "get_mask_sizes", "get_seq_length", "has_previous_state", "i", "if", "in", "index", "index_select", "int", "is", "is_compileable", "key_cache", "key_states", "kv_length", "kv_offset", "last_linear_layer", "layer_idx", "layer_types", "len", "linear_attention", "not", "num_hidden_layers", "or", "past_seen_tokens", "property", "query_length", "range", "recurrent_states", "reorder_cache", "return", "self", "shape", "str", "super", "to", "torch", "transformer_layers", "tuple", "update", "value_cache", "value_states"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "qwen3_next/modeling_qwen3_next.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen3_next/modeling_qwen3_next.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "qwen3_next/modeling_qwen3_next.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen3_next/modeling_qwen3_next.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "chunk", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dim", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "gate", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sigmoid", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_next/modeling_qwen3_next.py:apply_mask_to_padding_states": ["Model_mask_to_padding_states", "None", "and", "attention_mask", "def", "dtype", "hidden_states", "if", "is", "not", "return", "shape", "to"], "qwen3_next/modeling_qwen3_next.py:torch_causal_conv1d_update": ["F", "Model", "Model_causal_conv1d_update", "None", "_", "activation", "bias", "cat", "conv1d", "conv_state", "copy_", "def", "dim", "dtype", "groups", "hidden_size", "hidden_states", "hidden_states_new", "out", "padding", "return", "seq_len", "shape", "silu", "state_len", "to", "unsqueeze", "weight"], "qwen3_next/modeling_qwen3_next.py:l2norm": ["FloatTensor", "Model", "True", "def", "dim", "eps", "float", "int", "inv_norm", "keepdim", "return", "rsqrt", "sum", "torch", "x"], "qwen3_next/modeling_qwen3_next.py:torch_chunk_gated_delta_rule": ["F", "False", "Model", "Model_chunk_gated_delta_rule", "None", "attn", "attn_inter", "batch_size", "beta", "bool", "chunk_size", "clone", "contiguous", "core_attn_out", "cumsum", "decay_mask", "def", "device", "diagonal", "dim", "dtype", "else", "eps", "exp", "eye", "float", "float32", "for", "g", "i", "if", "in", "initial_dtype", "initial_state", "is", "k_beta", "k_cumdecay", "k_head_dim", "k_i", "key", "l2norm", "last_recurrent_state", "mask", "masked_fill", "masked_fill_", "not", "num_heads", "ones", "output_final_state", "pad", "pad_size", "q_i", "query", "range", "reshape", "return", "row", "scale", "sequence_length", "shape", "sub", "sum", "to", "total_sequence_length", "transpose", "tril", "triu", "unsqueeze", "use_qk_l2norm_in_kernel", "v_beta", "v_head_dim", "v_i", "v_new", "v_prime", "value", "x", "zeros", "zeros_like"], "qwen3_next/modeling_qwen3_next.py:torch_recurrent_gated_delta_rule": ["False", "Model", "Model_recurrent_gated_delta_rule", "None", "batch_size", "beta", "beta_t", "contiguous", "core_attn_out", "def", "delta", "dim", "dtype", "else", "eps", "exp", "float32", "for", "g", "g_t", "i", "if", "in", "initial_dtype", "initial_state", "is", "k_head_dim", "k_t", "key", "kv_mem", "l2norm", "last_recurrent_state", "not", "num_heads", "output_final_state", "q_t", "query", "range", "return", "scale", "sequence_length", "shape", "sum", "to", "transpose", "unsqueeze", "use_qk_l2norm_in_kernel", "v_head_dim", "v_t", "value", "x", "zeros"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextGatedDeltaNet": ["A", "ACT2FN", "AILab", "A_log", "Conv1d", "Dao", "F", "Falling", "False", "FusedRMSNormGated", "Linear", "LongTensor", "ModelConfig", "ModelDynamicCache", "ModelGatedDeltaNet", "ModelRMSNormGated", "Module", "None", "Optional", "Parameter", "Tensor", "The", "To", "True", "_", "__init__", "a", "act", "activation", "and", "apply_mask_to_padding_states", "attention", "attention_mask", "available", "b", "back", "batch_size", "because", "beta", "bias", "cache_params", "cache_position", "cat", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk_gated_delta_rule", "class", "com", "config", "conv1d", "conv_dim", "conv_kernel_size", "conv_state", "conv_states", "core_attn_out", "cuda", "current_device", "def", "device", "dim", "dt_bias", "dtype", "else", "empty", "eps", "exp", "fast", "fix_query_key_value_ordering", "fla", "flash", "float", "follow", "for", "forward", "fused_recurrent_gated_delta_rule", "g", "get_current_dtype", "github", "groups", "has_previous_state", "head_k_dim", "head_v_dim", "hidden_act", "hidden_size", "hidden_states", "https", "if", "implementation", "in", "in_channels", "in_proj_ba", "in_proj_qkvz", "initial_state", "install", "installed", "int", "is", "is_fast_path_available", "kernel_size", "key", "key_dim", "last_recurrent_state", "layer_idx", "layer_norm_epsilon", "library", "linear", "linear_conv_kernel_dim", "linear_key_head_dim", "linear_num_key_heads", "linear_num_value_heads", "linear_value_head_dim", "log", "logger", "mixed_ba", "mixed_qkv", "mixed_qkvz", "new_tensor_shape_ba", "new_tensor_shape_qkvz", "nn", "norm", "not", "num_k_heads", "num_v_heads", "of", "one", "ones", "or", "org", "out_channels", "out_proj", "output", "output_final_state", "pad", "padding", "path", "projected_states_ba", "projected_states_qkvz", "projection_size_ba", "projection_size_qkvz", "query", "recurrent_gated_delta_rule", "recurrent_state", "recurrent_states", "repeat_interleave", "required", "reshape", "return", "rms_norm_eps", "self", "seq_idx", "seq_len", "shape", "sigmoid", "silu", "size", "softplus", "split", "split_arg_list_ba", "split_arg_list_qkvz", "squeeze", "super", "the", "to", "torch", "torch_causal_conv1d_update", "torch_chunk_gated_delta_rule", "torch_recurrent_gated_delta_rule", "transpose", "uniform_", "use_precomputed_states", "use_qk_l2norm_in_kernel", "value", "value_dim", "view", "warning_once", "weight", "x", "z", "z_shape_og"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "not", "return", "self", "super", "up_proj", "x"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextSparseMoeBlock": ["F", "False", "Linear", "ModelMLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "keepdim", "moe_intermediate_size", "nn", "nonzero", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "shared_expert", "shared_expert_gate", "shared_expert_intermediate_size", "shared_expert_output", "sigmoid", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextDecoderLayer": ["Cache", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelGatedDeltaNet", "ModelMLP", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "and", "attention_mask", "cache_params", "cache_position", "class", "config", "decoder_sparse_step", "def", "deprecate_kwarg", "elif", "else", "eps", "forward", "full_attention", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "int", "intermediate_size", "isinstance", "kwargs", "layer_idx", "layer_type", "layer_types", "linear_attention", "linear_attn", "mlp", "mlp_only_layers", "new_name", "not", "num_experts", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextPreTrainedModel": ["A_log", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelGatedDeltaNet", "ModelPreTrainedModel", "ModelRMSNorm", "ModelSparseMoeBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_is_stateful", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn_2", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "dt_bias", "elif", "fill_", "hidden_states", "if", "index", "isinstance", "log_", "model", "module", "mtp", "past_key_values", "r", "router_logits", "self", "super", "supports_gradient_checkpointing", "uniform_", "weight", "zero_"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextModel": ["Cache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelDynamicCache", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_update_linear_attn_mask", "all", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layer_mask", "layer_type", "layers", "linear_attention", "linear_attn_mask", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "qwen3_next/modeling_qwen3_next.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextForCausalLM": ["False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDynamicCache", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "qwen3_next/modeling_qwen3_next.py:Qwen3NextForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "starcoder2/modeling_starcoder2.py:Starcoder2MLP": ["ACT2FN", "FloatTensor", "Linear", "ModelConfig", "ModelMLP", "Module", "Optional", "__init__", "act", "bias", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "functional", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "p", "residual_dropout", "return", "self", "super", "torch", "training", "tuple", "use_bias"], "starcoder2/modeling_starcoder2.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "starcoder2/modeling_starcoder2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "starcoder2/modeling_starcoder2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "starcoder2/modeling_starcoder2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "starcoder2/modeling_starcoder2.py:Starcoder2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "functional", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "p", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "residual_dropout", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "use_bias", "v_proj", "value_states", "version", "view"], "starcoder2/modeling_starcoder2.py:Starcoder2DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "nn", "norm_epsilon", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "starcoder2/modeling_starcoder2.py:Starcoder2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "starcoder2/modeling_starcoder2.py:Starcoder2PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "starcoder2/modeling_starcoder2.py:Starcoder2Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dropout", "else", "embed_tokens", "embedding_dropout", "eps", "exactly", "for", "forward", "functional", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "list", "mask_function", "must", "nn", "norm", "norm_epsilon", "not", "num_hidden_layers", "of", "one", "or", "p", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "training", "unsqueeze", "use_cache", "vocab_size"], "starcoder2/modeling_starcoder2.py:Starcoder2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "starcoder2/modeling_starcoder2.py:Starcoder2ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "starcoder2/modeling_starcoder2.py:Starcoder2ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "sam_hq/modeling_sam_hq.py:SamHQVisionEncoderOutput": ["FloatTensor", "ModelOutput", "ModelVisionEncoderOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "intermediate_embeddings", "last_hidden_state", "list", "r", "torch", "tuple"], "sam_hq/modeling_sam_hq.py:SamHQMMaskDecoderOutputs": ["FloatTensor", "ModelMMaskDecoderOutputs", "ModelOutput", "None", "Optional", "class", "iou_scores", "mask_decoder_attentions", "masks", "r", "torch"], "sam_hq/modeling_sam_hq.py:SamHQImageSegmentationOutput": ["FloatTensor", "ModelImageSegmentationOutput", "ModelOutput", "None", "Optional", "class", "iou_scores", "mask_decoder_attentions", "pred_masks", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states"], "sam_hq/modeling_sam_hq.py:SamHQVisionAttention": ["F", "Input", "Linear", "ModelVisionAttention", "Module", "None", "Parameter", "Tensor", "ValueError", "_", "__init__", "arange", "attention_dropout", "attn_output", "attn_probs", "attn_weights", "batch_size", "be", "bhwc", "bhwk", "bias", "class", "config", "decomposed_rel_pos", "def", "dim", "dropout", "dtype", "einsum", "else", "encoding", "float32", "forward", "functional", "get_decomposed_rel_pos", "get_rel_pos", "head_dim", "height", "hidden_size", "hidden_states", "hkc", "if", "image_size", "input_size", "int", "interpolate", "is", "k_coords", "k_size", "key", "key_height", "key_width", "linear", "long", "max", "max_rel_dist", "mode", "must", "nn", "num_attention_heads", "output_attentions", "p", "patch_size", "permute", "positional", "proj", "provided", "q_coords", "q_size", "qkv", "qkv_bias", "query", "query_height", "query_width", "raise", "rel_h", "rel_pos", "rel_pos_h", "rel_pos_resized", "rel_pos_w", "rel_w", "relative", "relative_coords", "relative_position_height", "relative_position_width", "reshape", "reshape_as", "reshaped_query", "return", "scale", "self", "shape", "size", "softmax", "super", "to", "torch", "training", "transpose", "tuple", "unbind", "use_rel_pos", "using", "value", "width", "window_size", "wkc", "zeros"], "sam_hq/modeling_sam_hq.py:SamHQMLPBlock": ["ACT2FN", "Linear", "ModelMLPBlock", "Module", "Tensor", "__init__", "act", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_states", "lin1", "lin2", "mlp_dim", "nn", "return", "self", "super", "torch"], "sam_hq/modeling_sam_hq.py:SamHQVisionSdpaAttention": ["Falling", "False", "ModelVisionAttention", "ModelVisionSdpaAttention", "None", "Tensor", "This", "Transformers", "True", "_", "__init__", "argument", "attention", "attn_bias", "attn_implementation", "attn_mask", "attn_output", "back", "batch_size", "be", "but", "can", "class", "config", "decomposed_rel_pos", "def", "does", "eager", "forward", "from", "functional", "get_decomposed_rel_pos", "height", "hidden_states", "if", "implementation", "is", "key", "loading", "logger", "manual", "model", "nn", "not", "num_attention_heads", "onwards", "output_attentions", "permute", "proj", "qkv", "query", "rel_pos_h", "rel_pos_w", "removed", "required", "reshape", "return", "scaled_dot_product_attention", "self", "shape", "specifying", "super", "support", "the", "to", "torch", "unbind", "use_rel_pos", "used", "using", "v5", "value", "version", "view", "warning", "warning_once", "when", "width", "will", "window_size"], "sam_hq/modeling_sam_hq.py:SamHQVisionLayer": ["F", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelMLPBlock", "ModelVisionLayer", "Model_VISION_ATTENTION_CLASSES", "Tensor", "__init__", "_attn_implementation", "attn", "attn_weights", "batch_size", "channel", "class", "config", "contiguous", "def", "eps", "forward", "height", "hidden_size", "hidden_states", "if", "int", "layer_norm1", "layer_norm2", "layer_norm_eps", "layernorm_output", "mlp", "nn", "original_shape", "pad", "pad_h", "pad_height", "pad_w", "pad_width", "padding_shape", "permute", "reshape", "residual", "return", "self", "shape", "super", "torch", "tuple", "width", "window_partition", "window_size", "window_unpartition", "windows"], "sam_hq/modeling_sam_hq.py:SamHQPreTrainedModel": ["Model", "ModelConfig", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionEncoder", "Module", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "class", "config", "data", "def", "elif", "if", "isinstance", "main_input_name", "module", "nn", "pixel_values", "pos_embed", "rel_pos_h", "rel_pos_w", "self", "super", "supports_gradient_checkpointing", "use_abs_pos", "use_rel_pos", "zero_"], "sam_hq/modeling_sam_hq.py:SamHQPatchEmbeddings": ["Conv2d", "Input", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "forward", "height", "hidden_size", "if", "image", "image_size", "in", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "of", "one", "or", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "values", "width", "with"], "sam_hq/modeling_sam_hq.py:SamHQVisionNeck": ["Conv2d", "False", "ModelLayerNorm", "ModelVisionConfig", "ModelVisionNeck", "Module", "__init__", "bias", "channels_first", "class", "config", "conv1", "conv2", "data_format", "def", "forward", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "nn", "output_channels", "padding", "permute", "return", "self", "super"], "sam_hq/modeling_sam_hq.py:SamHQVisionEncoder": ["False", "FloatTensor", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderOutput", "ModelVisionLayer", "ModelVisionNeck", "ModuleList", "None", "Optional", "Parameter", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "and", "append", "attentions", "check_model_inputs", "class", "config", "def", "else", "for", "forward", "get_input_embeddings", "global_attn_indexes", "gradient_checkpointing", "hasattr", "have", "hidden_size", "hidden_states", "i", "if", "image_size", "in", "intermediate_embeddings", "is", "kwargs", "last_hidden_state", "layer", "layer_module", "layers", "neck", "nn", "not", "num_hidden_layers", "patch_embed", "patch_size", "pixel_values", "pos_embed", "raise", "range", "return", "self", "specify", "super", "to", "torch", "tuple", "use_abs_pos", "window_size", "zeros"], "sam_hq/modeling_sam_hq.py:SamHQLayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "sam_hq/modeling_sam_hq.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "sam_hq/modeling_sam_hq.py:SamHQAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "__init__", "_attn_implementation", "_recombine_heads", "_separate_heads", "attention_downModelple_rate", "attention_interface", "attention_mask", "attention_similarity", "attn_output", "attn_weights", "batch", "c_per_head", "channel", "class", "config", "def", "divide", "downModelple_rate", "dropout", "eager", "eager_attention_forward", "else", "forward", "hidden_size", "hidden_states", "if", "int", "internal_dim", "is", "is_causal", "k_proj", "key", "kwargs", "must", "n_heads", "n_tokens", "nn", "num_attention_heads", "out_proj", "point_batch_size", "q_proj", "query", "raise", "reshape", "return", "scaling", "self", "shape", "super", "transpose", "v_proj", "value"], "sam_hq/modeling_sam_hq.py:SamHQTwoWayAttentionBlock": ["False", "LayerNorm", "ModelAttention", "ModelMLPBlock", "ModelTwoWayAttentionBlock", "Module", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_downModelple_rate", "attention_similarity", "attn_out", "bool", "class", "config", "cross_attn_image_to_token", "cross_attn_token_to_image", "def", "downModelple_rate", "else", "eps", "forward", "hidden_size", "if", "int", "key", "key_point_embedding", "keys", "kwargs", "layer_norm1", "layer_norm2", "layer_norm3", "layer_norm4", "layer_norm_eps", "mlp", "mlp_out", "nn", "queries", "query", "query_point_embedding", "return", "self", "self_attn", "skip_first_layer_pe", "super", "value"], "sam_hq/modeling_sam_hq.py:SamHQTwoWayTransformer": ["BaseModelOutput", "LayerNorm", "ModelAttention", "ModelMaskDecoderConfig", "ModelTwoWayAttentionBlock", "ModelTwoWayTransformer", "Module", "ModuleList", "None", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "_", "__init__", "an", "append", "attention_similarity", "attn_out", "class", "config", "def", "final_attn_token_to_image", "flatten", "for", "forward", "have", "hidden_size", "i", "if", "image_embedding", "image_embeddings", "image_positional_embeddings", "in", "is", "key", "key_point_embedding", "keys", "kwargs", "layer", "layer_norm_final_attn", "layers", "nn", "not", "num_hidden_layers", "permute", "point_embeddings", "queries", "query", "query_point_embedding", "raise", "range", "return", "self", "skip_first_layer_pe", "specify", "super", "target_embedding", "to", "tuple", "unsqueeze", "value"], "sam_hq/modeling_sam_hq.py:SamHQFeedForward": ["F", "False", "Linear", "ModelFeedForward", "Module", "ModuleList", "ReLU", "_", "__init__", "activation", "bool", "class", "def", "for", "forward", "hidden_dim", "hidden_states", "if", "in", "input_dim", "int", "layer", "layers", "nn", "num_layers", "output_dim", "proj_in", "proj_out", "range", "return", "self", "sigmoid", "sigmoid_output", "super"], "sam_hq/modeling_sam_hq.py:SamHQMaskDecoder": ["Conv2d", "ConvTranspose2d", "Embedding", "GELU", "ModelFeedForward", "ModelLayerNorm", "ModelMMaskDecoderOutputs", "ModelMaskDecoder", "ModelMaskDecoderConfig", "ModelTwoWayTransformer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "_", "__init__", "activation", "and", "attention_similarity", "batch_size", "bool", "cat", "channels_first", "class", "compress_vit_conv1", "compress_vit_conv2", "compress_vit_norm", "compressed_vit_features", "config", "contiguous", "current_mlp", "data_format", "def", "dense_prompt_embeddings", "descending", "dim", "dtype", "elif", "else", "embed_encode", "encoder_conv1", "encoder_conv2", "encoder_norm", "expand", "for", "forward", "gather", "has_intermediate", "height", "hidden_size", "hq_features", "hq_mask_mlp", "hq_token", "hq_token_only", "hyper_in", "hyper_in_list", "if", "image_embeddings", "image_positional_embeddings", "in", "intermediate_embeddings", "iou_head_depth", "iou_head_hidden_dim", "iou_pred", "iou_pred_sorted", "iou_prediction_head", "iou_token", "iou_token_out", "is", "kernel_size", "len", "list", "mask_conv1", "mask_conv2", "mask_norm", "mask_slice", "mask_token_index", "mask_tokens", "mask_tokens_out", "masks", "masks_Model", "masks_hq", "mlps_list", "multimask_output", "nn", "not", "num_channels", "num_mask_tokens", "num_multimask_outputs", "output_hypernetworks_mlps", "output_tokens", "padding", "permute", "point_batch_size", "point_embedding", "point_embeddings", "range", "repeat", "repeat_interleave", "reshape", "return", "self", "shape", "slice", "sort", "sort_indices", "sparse_prompt_embeddings", "stack", "stride", "super", "target_embedding", "to", "tokens", "torch", "transformer", "transpose", "upscale_conv1", "upscale_conv2", "upscale_layer_norm", "upscaled_embedding", "upscaled_embedding_hq", "vit_dim", "vit_features", "weight", "width"], "sam_hq/modeling_sam_hq.py:SamHQVisionModel": ["FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderOutput", "ModelVisionModel", "Module", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "__init__", "auto_docstring", "class", "config", "def", "forward", "get_input_embeddings", "kwargs", "main_input_name", "nn", "patch_embed", "pixel_values", "post_init", "return", "self", "super", "torch", "tuple", "vision_encoder"], "sam_hq/modeling_sam_hq.py:SamHQPositionalEmbedding": ["ModelPositionalEmbedding", "Module", "None", "__init__", "cat", "class", "clone", "config", "coordinates", "cos", "def", "dim", "dtype", "forward", "hidden_size", "if", "input_coords", "input_shape", "is", "nn", "not", "np", "num_pos_feats", "pi", "positional_embedding", "randn", "register_buffer", "return", "scale", "self", "sin", "super", "to", "torch"], "sam_hq/modeling_sam_hq.py:SamHQMaskEmbedding": ["ACT2FN", "Conv2d", "ModelLayerNorm", "ModelMaskEmbedding", "ModelPromptEncoderConfig", "Module", "__init__", "activation", "channels_first", "class", "config", "conv1", "conv2", "conv3", "data_format", "def", "dense_embeddings", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "layer_norm_eps", "mask_input_channels", "masks", "nn", "return", "self", "stride", "super"], "sam_hq/modeling_sam_hq.py:SamHQPromptEncoder": ["Embedding", "If", "ModelConfig", "ModelMaskEmbedding", "ModelPositionalEmbedding", "ModelPromptEncoder", "Module", "ModuleList", "None", "Optional", "Tensor", "ValueError", "__init__", "_embed_boxes", "_embed_points", "also", "are", "batch_size", "be", "bool", "box_embeddings", "boxes", "cat", "class", "config", "coords", "corner_embedding", "def", "dense_embeddings", "device", "dim", "else", "expand", "for", "forward", "hidden_size", "i", "if", "image_embedding_size", "image_size", "in", "input_boxes", "input_image_size", "input_labels", "input_masks", "input_points", "input_shape", "is", "labels", "mask_embed", "must", "nb_boxes", "nn", "no_mask_embed", "not", "not_a_point_embed", "num_point_embeddings", "ones", "pad", "padding_label", "padding_point", "point_embed", "point_embedding", "point_embeddings", "points", "prompt_encoder_config", "provided", "raise", "range", "reshape", "return", "self", "shape", "shared_embedding", "sparse_embeddings", "super", "target_labels_shape", "target_point_shape", "torch", "tuple", "vision_config", "weight", "where", "zeros", "zeros_like"], "sam_hq/modeling_sam_hq.py:SamHQModel": ["Either", "False", "FloatTensor", "Got", "LongTensor", "ModelImageSegmentationOutput", "ModelMaskDecoder", "ModelModel", "ModelPositionalEmbedding", "ModelPreTrainedModel", "ModelPromptEncoder", "ModelTwoWayAttentionBlock", "ModelVisionEncoder", "None", "Of", "Only", "Optional", "OutputRecorder", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_can_record_outputs", "_keys_to_ignore_on_load_missing", "_tie_weights", "_tied_weights_keys", "a", "and", "as", "attention_similarity", "attentions", "auto_docstring", "batch_size", "be", "bool", "bounding", "box", "box_batch_size", "boxes", "can", "check_model_inputs", "class", "config", "cumsum", "data", "def", "dense_embeddings", "dense_prompt_embeddings", "device", "dict", "dim", "dtype", "else", "f", "forward", "get_image_embeddings", "get_image_wide_positional_embeddings", "get_input_embeddings", "get_prompt_embeddings", "got", "grid", "hidden_states", "hq_token_only", "if", "image_embedding_size", "image_embeddings", "image_positional_embeddings", "index", "input", "input_boxes", "input_labels", "input_masks", "input_points", "int", "intermediate_embeddings", "iou_scores", "is", "kwargs", "last_hidden_state", "len", "list", "many", "mask_decoder", "mask_decoder_attentions", "mask_decoder_config", "mask_decoder_output", "multimask_output", "must", "nb_boxes", "nb_points_per_image", "no_grad", "not", "of", "one", "ones", "ones_like", "or", "per", "permute", "pixel_values", "point_batch_size", "points", "positional_embedding", "post_init", "pred_masks", "prompt_encoder", "prompt_encoder_config", "prompt_output", "provide", "provided", "r", "raise", "repeat", "return", "self", "shape", "shared_embedding", "shared_image_embedding", "should", "size", "sparse_embeddings", "sparse_prompt_embeddings", "stack", "str", "super", "target_device", "target_dtype", "target_embedding", "tensor", "torch", "unsqueeze", "vision_attentions", "vision_config", "vision_encoder", "vision_hidden_states", "vision_output", "vision_outputs", "x_embed", "y_embed"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertRotaryPositionalEmbedding": ["False", "ModelRotaryPositionalEmbedding", "Module", "None", "__init__", "and", "arange", "base", "cached_rotary_positional_embedding", "cached_sequence_length", "cat", "class", "config", "cos", "cos_embeddings", "def", "dim", "dtype", "einsum", "embeddings", "float", "forward", "freqs", "hidden_size", "hidden_states", "i", "if", "ij", "int64", "inv_freq", "is", "j", "nn", "not", "num_attention_heads", "persistent", "register_buffer", "return", "rotary_embedding_base", "self", "sequence_length", "shape", "sin", "sin_embeddings", "stack", "super", "time_stamps", "torch", "type_as"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertRelPositionalEmbedding": ["ModelRelPositionalEmbedding", "Module", "None", "Tensor", "__init__", "arange", "cat", "class", "config", "cos", "d_model", "def", "device", "dim", "div_term", "dtype", "end_idx", "exp", "expand", "extend_pe", "flip", "float", "forward", "hidden_size", "hidden_states", "if", "int64", "is", "log", "math", "max_len", "max_source_positions", "nn", "not", "or", "pe", "pe_negative", "pe_positive", "position", "relative_position_embeddings", "return", "self", "sin", "size", "start_idx", "super", "tensor", "to", "torch", "unsqueeze", "x", "zeros"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "def", "dropout", "eps", "feat_proj_dropout", "feature_projection_input_dim", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "None", "__init__", "act_fn", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "is", "isinstance", "nn", "not", "output_dense", "output_dropout", "return", "self", "str", "super"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertConvolutionModule": ["ACT2FN", "Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelConvolutionModule", "Module", "None", "SAME", "ValueError", "__init__", "a", "activation", "attention_mask", "be", "bias", "bool", "class", "config", "conformer_conv_dropout", "conv_depthwise_kernel_size", "def", "depthwise_conv", "depthwise_layer_norm", "dim", "dropout", "eps", "for", "forward", "functional", "glu", "groups", "hidden_act", "hidden_size", "hidden_states", "if", "is", "kernel_size", "layer_norm", "layer_norm_eps", "masked_fill", "nn", "not", "number", "odd", "pad", "padding", "pointwise_conv1", "pointwise_conv2", "raise", "return", "self", "should", "stride", "super", "torch", "transpose", "unsqueeze"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertSelfAttention": ["Dropout", "Embedding", "False", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_apply_relative_embeddings", "_apply_rotary_embedding", "arange", "attention_dropout", "attention_mask", "batch_size", "be", "bhld", "bhlr", "bias", "bool", "cat", "clamp", "class", "config", "cos", "def", "defined", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "else", "forward", "has", "head_size", "hidden_size", "hidden_states", "if", "is", "is_adapter_attention", "key", "key_length", "left_max_position_embeddings", "linear_k", "linear_out", "linear_pos", "linear_q", "linear_v", "long", "lrd", "math", "matmul", "ndim", "nn", "not", "num_attention_heads", "num_heads", "num_positions", "output_attentions", "output_hidden_size", "p", "pos_bias_u", "pos_bias_v", "position_embeddings_type", "position_ids_l", "position_ids_r", "positional_embedding", "probs", "proj_relative_position_embeddings", "q_with_bias_u", "q_with_bias_v", "query", "query_key_states", "query_length", "raise", "relative", "relative_key", "relative_position_attn_weights", "relative_position_embeddings", "reshape", "return", "right_max_position_embeddings", "rotary", "rotated_states", "rotated_states_begin", "rotated_states_end", "scores", "scores_ac", "scores_bd", "scores_bd_padded", "scores_bd_padded_shape", "self", "sequence_length", "shape", "sin", "size", "softmax", "sqrt", "super", "to", "torch", "transpose", "tuple", "value", "value_states", "view", "view_as", "when", "zero_pad", "zeros"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelConvolutionModule", "ModelEncoderLayer", "ModelFeedForward", "ModelSelfAttention", "None", "Optional", "Tensor", "__init__", "attention_dropout", "attention_mask", "attn_weigts", "bool", "class", "config", "conv_attention_mask", "conv_module", "def", "dropout", "embed_dim", "eps", "ffn1", "ffn1_layer_norm", "ffn2", "ffn2_layer_norm", "final_layer_norm", "forward", "hidden_size", "hidden_states", "layer_norm_eps", "nn", "output_attentions", "relative_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_dropout", "self_attn_layer_norm", "super", "torch"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertEncoder": ["BaseModelOutput", "Dropout", "False", "ModelEncoder", "ModelEncoderLayer", "ModelRelPositionalEmbedding", "ModelRotaryPositionalEmbedding", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "conv_attention_mask", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "embed_positions", "enumerate", "expand", "finfo", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_states", "i", "if", "in", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_outputs", "layerdrop", "layers", "masked_fill", "min", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "position_embeddings_type", "rand", "range", "relative", "relative_position_embeddings", "return", "return_dict", "rotary", "self", "shape", "skip_the_layer", "super", "synced_gpus", "to", "torch", "training", "tuple", "unsqueeze", "v"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertAdapter": ["LayerNorm", "Linear", "ModelAdapter", "ModelAdapterLayer", "Module", "ModuleList", "None", "_", "__init__", "_compute_sub_sample_lengths_from_attention_mask", "adapter_kernel_size", "adapter_stride", "and", "attention_mask", "class", "config", "def", "device", "else", "eps", "floor", "for", "forward", "hidden_size", "hidden_states", "if", "in", "int", "is", "kernel_size", "layer", "layer_norm_eps", "layerdrop", "layerdrop_prob", "layers", "nn", "not", "num_adapter_layers", "or", "output_hidden_size", "pad", "proj", "proj_layer_norm", "rand", "range", "return", "self", "seq_lens", "size", "stride", "sub_sampled_lengths", "sum", "super", "to", "torch", "training"], "wav2vec2_bert/modeling_wav2vec2_bert.py:_compute_new_attention_mask": ["Tensor", "_compute_new_attention_mask", "arange", "batch_size", "bool_mask", "def", "device", "expand", "hidden_states", "indices", "mask", "mask_seq_len", "masked_fill", "new_ones", "return", "seq_lens", "shape", "torch", "unsqueeze"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertAdapterLayer": ["Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelAdapterLayer", "ModelFeedForward", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "_compute_new_attention_mask", "_prepare_4d_attention_mask", "act_fn", "activation", "adapter_act", "adapter_kernel_size", "adapter_stride", "attention_mask", "attn_weights", "bool", "class", "config", "conformer_conv_dropout", "def", "dim", "dropout", "dtype", "embed_dim", "eps", "ffn", "ffn_layer_norm", "forward", "hidden_size", "hidden_states", "if", "is", "is_adapter_attention", "kernel_size", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_size", "padding", "residual", "residual_conv", "residual_layer_norm", "return", "self", "self_attn", "self_attn_conv", "self_attn_dropout", "self_attn_layer_norm", "seq_lens", "stride", "sub_sampled_lengths", "super", "torch", "transpose"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertPreTrainedModel": ["AMSoftmaxLoss", "Conv1d", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelFeatureProjection", "ModelForAudioFrameClassification", "ModelForSequenceClassification", "ModelForXVector", "ModelModel", "ModelPreTrainedModel", "ModelSelfAttention", "None", "Optional", "PreTrainedModel", "True", "Union", "_", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "a", "adapter_kernel_size", "adapter_stride", "add_adapter", "arange", "attention_mask", "b", "base_model_prefix", "batch_size", "bias", "bool", "class", "config", "cumsum", "data", "def", "device", "dim", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "groups", "hasattr", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_features", "input_length", "input_lengths", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "layer_weights", "long", "main_input_name", "masked_spec_embed", "math", "mean", "module", "nn", "non_padded_lengths", "normal_", "not", "num_adapter_layers", "num_hidden_layers", "output_lengths", "padding", "pos_bias_u", "pos_bias_v", "projection", "range", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "supports_gradient_checkpointing", "to", "torch", "uniform_", "weight", "xavier_uniform_", "zero_", "zeros"], "wav2vec2_bert/modeling_wav2vec2_bert.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertModel": ["FloatTensor", "LongTensor", "ModelAdapter", "ModelBaseModelOutput", "ModelConfig", "ModelEncoder", "ModelFeatureProjection", "ModelFeedForward", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_mask_hidden_states", "act_fn", "adapter", "add_adapter", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "dtype", "elif", "else", "encoder", "encoder_outputs", "expand", "expanded_hidden_states", "extract_features", "feature_projection", "forward", "getattr", "hidden_size", "hidden_states", "if", "input_features", "intermediate_ffn", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "relu", "return", "return_dict", "self", "sequence_length", "size", "super", "tensor", "to", "torch", "training", "tuple", "uniform_", "use_intermediate_ffn_before_adapter", "use_return_dict"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertForCTC": ["CausalLMOutput", "Dropout", "False", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_get_feat_extract_output_lengths", "a", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "define", "device", "dim", "does", "dropout", "dtype", "else", "enabled", "f", "final_dropout", "flags", "flattened_targets", "float32", "follows", "forward", "from_pretrained", "functional", "hasattr", "head", "hidden_size", "hidden_states", "if", "input_features", "input_lengths", "instantiate", "is", "labels", "labels_mask", "language", "lm_head", "log_probs", "log_softmax", "logits", "long", "loss", "masked_select", "max", "model", "must", "nn", "not", "of", "ones", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "post_init", "r", "raise", "reduction", "return", "return_dict", "s", "self", "shape", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "to", "torch", "transpose", "trying", "tuple", "use_return_dict", "values", "vocab_size", "vocabulary", "with", "your", "zero_infinity"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertForSequenceClassification": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Sequence", "SequenceClassifierOutput", "Tensor", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "dim", "does", "else", "expand_padding_mask", "for", "forward", "freeze_base_model", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_features", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "view"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertForAudioFrameClassification": ["Audio", "CrossEntropyLoss", "False", "Linear", "Model", "ModelForAudioFrameClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "TokenClassifierOutput", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "adapters", "add_adapter", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "axis", "bool", "class", "classification", "classifier", "config", "def", "dim", "does", "else", "for", "forward", "frame", "freeze_base_model", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "init_weights", "input_features", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "param", "parameters", "r", "raise", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "view"], "wav2vec2_bert/modeling_wav2vec2_bert.py:AMSoftmaxLoss": ["CrossEntropyLoss", "Model", "Module", "Parameter", "True", "__init__", "bool", "class", "cos_theta", "def", "dim", "flatten", "forward", "functional", "hidden_states", "input_dim", "labels", "logits", "loss", "margin", "mm", "nn", "normalize", "num_labels", "one_hot", "onehot", "psi", "randn", "requires_grad", "return", "scale", "self", "super", "torch", "weight", "where"], "wav2vec2_bert/modeling_wav2vec2_bert.py:TDNNLayer": ["Detected", "Linear", "LoRA", "LoraLayer", "Model", "Module", "ReLU", "Tensor", "You", "__init__", "activation", "applied", "be", "bias", "class", "config", "conv1d", "def", "dilation", "due", "else", "exclude", "forward", "from", "functional", "hidden_states", "if", "in_conv_dim", "is_peft_available", "isinstance", "kernel", "kernel_size", "layer_id", "modules", "nn", "on", "optimization", "out_conv_dim", "return", "s", "self", "should", "super", "t", "target", "tdnn_dilation", "tdnn_dim", "tdnn_kernel", "to", "torch", "transpose", "view", "warn", "warnings", "weight", "weights", "won"], "wav2vec2_bert/modeling_wav2vec2_bert.py:Wav2Vec2BertForXVector": ["AMSoftmaxLoss", "False", "Linear", "LongTensor", "Model", "ModelForXVector", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Parameter", "TDNNLayer", "Tensor", "True", "Union", "XVectorOutput", "_HIDDEN_STATES_START_POSITION", "__init__", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_tdnn_output_lengths", "append", "attention_mask", "attentions", "auto_docstring", "bool", "cat", "class", "classifier", "config", "def", "dim", "else", "embeddings", "enumerate", "feat_extract_output_lengths", "feature_extractor", "for", "forward", "freeze_base_model", "functional", "hidden_size", "hidden_states", "i", "if", "in", "init_weights", "input_features", "input_length", "input_lengths", "int", "is", "kernel_size", "labels", "layer_weights", "len", "length", "logits", "loss", "mean", "mean_features", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "objective", "ones", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "parameters", "projector", "r", "range", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "statistic_pooling", "std", "std_features", "stride", "sum", "super", "tdnn", "tdnn_dim", "tdnn_kernel", "tdnn_layer", "tdnn_layers", "tdnn_output_lengths", "torch", "tuple", "use_return_dict", "use_weighted_layer_sum", "view", "xvector_output_dim"], "trocr/modeling_trocr.py:TrOCRLearnedPositionalEmbedding": ["Embedding", "ModelLearnedPositionalEmbedding", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "else", "embedding_dim", "expand", "forward", "if", "input_ids", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "seq_len", "shape", "super", "torch", "unsqueeze", "weight"], "trocr/modeling_trocr.py:TrOCRScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "trocr/modeling_trocr.py:TrOCRSinusoidalPositionalEmbedding": ["FloatTensor", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "_float_tensor", "arange", "bsz", "cat", "class", "cos", "create_position_ids_from_input_ids", "cumsum", "def", "detach", "device", "dim", "dtype", "emb", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "if", "incremental_indices", "index_select", "input_ids", "int", "int64", "is", "log", "long", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "or", "padding_idx", "past_key_values_length", "position_ids", "register_buffer", "return", "self", "seq_len", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "x", "zeros"], "trocr/modeling_trocr.py:TrOCRAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "class", "config", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "else", "embed_dim", "f", "float", "for", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "kdim", "key_states", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "mask", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "should", "single", "size", "softmax", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "vdim", "version", "view", "weights"], "trocr/modeling_trocr.py:TrOCRDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attention_hidden_size", "cross_attn_layer_head_mask", "cross_attn_weights", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_size", "hidden_states", "if", "is", "is_cross_attention", "is_decoder", "kdim", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "vdim", "version"], "trocr/modeling_trocr.py:TrOCRPreTrainedModel": ["Conv1d", "Embedding", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "if", "init_std", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "trocr/modeling_trocr.py:TrOCRDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Passing", "Setting", "The", "Transformers", "True", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "at", "attention_mask", "attentions", "attn_mask", "be", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "math", "max_position_embeddings", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "shape", "should", "size", "specified", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_learned_position_embeddings", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "trocr/modeling_trocr.py:TrOCRDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "trocr/modeling_trocr.py:TrOCRForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "logits", "loss", "loss_fct", "model", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_projection", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "set_output_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "florence2/modeling_florence2.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "florence2/modeling_florence2.py:Florence2VisionDropPath": ["ModelVisionDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "florence2/modeling_florence2.py:Florence2VisionLearnedAbsolutePositionEmbedding2D": ["Embedding", "ModelConfig", "ModelVisionLearnedAbsolutePositionEmbedding2D", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "config", "def", "device", "dim", "embed_dim", "embedding_dim", "forward", "height", "height_values", "max_position_embeddings", "nn", "num_pos", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "vision_config", "width", "width_values", "x_emb", "y_emb"], "florence2/modeling_florence2.py:Florence2VisionPositionalEmbeddingCosine1D": ["Maximum", "ModelConfig", "ModelVisionPositionalEmbeddingCosine1D", "Module", "Tensor", "ValueError", "__init__", "arange", "class", "config", "cos", "cosine", "def", "dtype", "emb", "embed_dim", "empty", "exp", "f", "float", "forward", "get_sinusoid_embeddings", "got", "half_dim", "if", "int", "int64", "len_seq", "length", "log", "math", "max_positions", "max_seq_len", "max_temporal_embeddings", "nn", "pos_embeds", "pos_idx_to_embed", "raise", "register_buffer", "return", "self", "seq_embeds", "sequence", "sin", "sine", "size", "staticmethod", "super", "torch", "unsqueeze", "vision_config"], "florence2/modeling_florence2.py:Florence2VisionMLP": ["ACT2FN", "Linear", "ModelVisionConfig", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "activation_function", "class", "config", "def", "embed_dim", "fc1", "fc2", "forward", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "stage_idx", "super", "torch"], "florence2/modeling_florence2.py:Florence2VisionConvEmbed": ["Conv2d", "LayerNorm", "ModelVisionConfig", "ModelVisionConvEmbed", "Module", "Tensor", "__init__", "and", "class", "config", "conv", "def", "dim_norm", "else", "embed_dim", "forward", "hidden_states", "if", "in_channels", "int", "kernel_size", "nn", "norm", "not", "padding", "patch_padding", "patch_prenorm", "patch_size", "patch_stride", "permute", "pre_norm", "return", "self", "stage_idx", "stride", "super", "torch"], "florence2/modeling_florence2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "florence2/modeling_florence2.py:Florence2VisionChannelAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionChannelAttention", "ModelVisionConfig", "Module", "None", "Tensor", "_", "__init__", "_attn_implementation", "attention_interface", "attention_mask", "batch_size", "bias", "class", "config", "def", "dim", "eager", "eager_attention_forward", "embed_dim", "forward", "groups", "hidden_size", "hidden_states", "if", "int", "is_causal", "key", "nn", "num_groups", "num_tokens", "permute", "proj", "qkv", "qkv_bias", "query", "reshape", "return", "scale", "scaling", "self", "shape", "stage_idx", "super", "torch", "unbind", "value"], "florence2/modeling_florence2.py:Florence2VisionChannelBlock": ["Conv2d", "Identity", "LayerNorm", "ModelVisionChannelAttention", "ModelVisionChannelBlock", "ModelVisionConfig", "ModelVisionDropPath", "ModelVisionMLP", "Module", "Tensor", "__init__", "batch_size", "channel_attn", "class", "config", "conv1", "conv2", "def", "dim_in", "drop_path1", "drop_path2", "drop_path_rate", "else", "embed_dim", "ffn", "flatten", "float", "forward", "groups", "height", "hidden_states", "if", "int", "kernel_size", "nn", "norm1", "norm2", "padding", "residual", "return", "self", "shape", "stage_idx", "super", "torch", "transpose", "view", "width"], "florence2/modeling_florence2.py:Florence2VisionWindowAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "F", "False", "Linear", "ModelVisionConfig", "ModelVisionWindowAttention", "Module", "None", "Tensor", "_", "__init__", "_attn_implementation", "attention_interface", "attention_mask", "batch_size", "bias", "class", "config", "contiguous", "def", "dim", "eager", "eager_attention_forward", "embed_dim", "forward", "head_dim", "height", "hidden_states", "if", "int", "is_causal", "key", "nn", "num_heads", "num_tokens_per_window", "num_windows_per_batch", "pad", "pad_bottom", "pad_left", "pad_right", "pad_top", "padded_height", "padded_width", "permute", "proj", "qkv", "qkv_bias", "query", "reshape", "return", "scale", "scaling", "self", "shape", "stage_idx", "super", "torch", "unbind", "value", "view", "width", "window_size", "windowed_hidden_states"], "florence2/modeling_florence2.py:Florence2VisionSpatialBlock": ["Conv2d", "Identity", "LayerNorm", "ModelVisionConfig", "ModelVisionDropPath", "ModelVisionMLP", "ModelVisionSpatialBlock", "ModelVisionWindowAttention", "Module", "Tensor", "__init__", "batch_size", "class", "config", "conv1", "conv2", "def", "drop_path1", "drop_path2", "drop_path_rate", "else", "embed_dim", "ffn", "flatten", "float", "forward", "groups", "height", "hidden_states", "if", "int", "kernel_size", "nn", "norm1", "norm2", "padding", "residual", "return", "self", "shape", "stage_idx", "super", "torch", "transpose", "view", "width", "window_attn"], "florence2/modeling_florence2.py:Florence2VisionBlock": ["ModelVisionBlock", "ModelVisionChannelBlock", "ModelVisionConfig", "ModelVisionSpatialBlock", "Module", "Tensor", "__init__", "channel_block", "channel_drop_path_rate", "class", "config", "def", "drop_path_rate", "float", "forward", "hidden_states", "int", "nn", "return", "self", "spatial_block", "spatial_drop_path_rate", "stage_idx", "super", "torch"], "florence2/modeling_florence2.py:Florence2VisionPreTrainedModel": ["ModelVisionConfig", "ModelVisionPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "class", "config_class", "main_input_name", "pixel_values"], "florence2/modeling_florence2.py:Florence2VisionBackbone": ["Expected", "ModelVisionBackbone", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionConvEmbed", "ModelVisionPreTrainedModel", "ModuleList", "Tensor", "ValueError", "__init__", "append", "block", "block_idx", "blocks", "channel_drop_path_rate", "class", "config", "conv", "conv_embed", "convs", "cpu", "def", "depth_offset", "depths", "device", "dpr", "drop_path_rate", "embed_dim", "f", "for", "forward", "hidden_states", "if", "in", "item", "layer", "len", "linspace", "nn", "not", "num_groups", "num_heads", "num_stages", "post_init", "raise", "range", "return", "self", "spatial_drop_path_rate", "stage_idx", "sum", "super", "torch", "x", "zip"], "florence2/modeling_florence2.py:Florence2MultiModalProjector": ["False", "LayerNorm", "Linear", "ModelConfig", "ModelMultiModalProjector", "ModelVisionLearnedAbsolutePositionEmbedding2D", "ModelVisionPositionalEmbeddingCosine1D", "Module", "__init__", "bias", "cat", "class", "config", "def", "dim", "embed_dim", "flatten", "forward", "image_features", "image_position_embed", "image_proj_norm", "image_projection", "mean", "nn", "position_features", "projection_dim", "return", "self", "spatial_image_features", "super", "temporal_features", "temporal_image_features", "torch", "transpose", "unsqueeze", "vision_config", "vision_embedding_dim", "vision_projection_dim", "visual_temporal_embed", "visual_token_features"], "florence2/modeling_florence2.py:Florence2Seq2SeqModelOutput": ["FloatTensor", "ModelSeq2SeqModelOutput", "None", "Optional", "Seq2SeqModelOutput", "class", "image_hidden_states", "r", "torch"], "florence2/modeling_florence2.py:Florence2Seq2SeqLMOutput": ["FloatTensor", "ModelSeq2SeqLMOutput", "None", "Optional", "Seq2SeqLMOutput", "class", "image_hidden_states", "r", "torch", "tuple"], "florence2/modeling_florence2.py:Florence2PreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "config_class", "past_key_values", "supports_gradient_checkpointing"], "florence2/modeling_florence2.py:Florence2Model": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelMultiModalProjector", "ModelPreTrainedModel", "ModelSeq2SeqModelOutput", "ModelVisionBackbone", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "all", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "device", "do", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "exactly", "expand_as", "f", "features", "forward", "from_config", "get_decoder", "get_encoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "head_mask", "hidden_states", "if", "image", "image_embeds", "image_features", "image_hidden_states", "image_token_id", "input_ids", "inputs_embeds", "is", "kwargs", "language_model", "last_hidden_state", "list", "long", "masked_scatter", "match", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "pixel_values", "post_init", "raise", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "size", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_config", "vision_tower", "weight"], "florence2/modeling_florence2.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "florence2/modeling_florence2.py:Florence2ForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "ModelSeq2SeqLMOutput", "Module", "None", "Optional", "Tensor", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_prepare_encoder_decoder_kwargs_for_generation", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dict", "dtype", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "generation_config", "get_decoder", "get_encoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_placeholder_mask", "head_mask", "hidden_size", "hidden_states", "if", "image_features", "image_hidden_states", "input_ids", "inputs_embeds", "inputs_tensor", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logger", "logits", "logits_to_keep", "loss", "loss_function", "masked_scatter", "model", "model_input_name", "model_inputs", "model_kwargs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pixel_values", "pop", "post_init", "prepare_inputs_for_generation", "property", "provided", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shift_tokens_right", "since", "slice", "slice_indices", "special_image_mask", "str", "super", "text_config", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_tower", "vocab_size", "warning", "weight"], "mixtral/modeling_mixtral.py:MixtralBlockSparseTop2MLP": ["ACT2FN", "False", "Linear", "ModelBlockSparseTop2MLP", "ModelConfig", "Module", "__init__", "act_fn", "bias", "class", "config", "current_hidden_states", "def", "ffn_dim", "forward", "hidden_act", "hidden_dim", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "w1", "w2", "w3"], "mixtral/modeling_mixtral.py:MixtralSparseMoeBlock": ["F", "False", "Linear", "ModelBlockSparseTop2MLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "and", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "empty_like", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "ffn_dim", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "jitter_noise", "keepdim", "nn", "nonzero", "num_classes", "num_experts", "num_experts_per_tok", "num_local_experts", "one_hot", "permute", "range", "reshape", "return", "router_jitter_noise", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "training", "uniform_", "view", "where", "zeros"], "mixtral/modeling_mixtral.py:MixtralRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "mixtral/modeling_mixtral.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "mixtral/modeling_mixtral.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "mixtral/modeling_mixtral.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "mixtral/modeling_mixtral.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "mixtral/modeling_mixtral.py:MixtralAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "mixtral/modeling_mixtral.py:MixtralDecoderLayer": ["Cache", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "block_sparse_moe", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "mixtral/modeling_mixtral.py:MixtralRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "mixtral/modeling_mixtral.py:MixtralPreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSparseMoeBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "index", "model", "past_key_values", "router_logits", "supports_gradient_checkpointing"], "mixtral/modeling_mixtral.py:MixtralModel": ["Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "mixtral/modeling_mixtral.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "mixtral/modeling_mixtral.py:MixtralForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "mixtral/modeling_mixtral.py:MixtralForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "mixtral/modeling_mixtral.py:MixtralForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "mixtral/modeling_mixtral.py:MixtralForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "class", "pass"], "kosmos2_5/modeling_kosmos2_5.py:_expand_mask": ["None", "Optional", "Tensor", "_expand_mask", "bool", "bsz", "def", "dtype", "else", "expand", "expanded_mask", "finfo", "if", "int", "inverted_mask", "is", "mask", "masked_fill", "min", "not", "return", "size", "src_len", "tgt_len", "to", "torch"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5ModelOutput": ["Any", "BaseModelOutputWithPooling", "Cache", "FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "def", "else", "for", "getattr", "height", "hidden_states", "if", "image_embeds", "in", "k", "keys", "last_hidden_state", "not", "past_key_values", "projection_attentions", "return", "self", "to_tuple", "torch", "tuple", "vision_model_output", "width"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5ForConditionalGenerationModelOutput": ["Any", "BaseModelOutputWithPooling", "Cache", "FloatTensor", "ModelForConditionalGenerationModelOutput", "ModelOutput", "None", "Optional", "Union", "attentions", "class", "def", "else", "for", "getattr", "height", "hidden_states", "if", "image_embeds", "in", "k", "keys", "list", "logits", "loss", "not", "past_key_values", "projection_attentions", "return", "self", "to_tuple", "torch", "tuple", "vision_model_output", "width"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5LayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5VisionEmbeddings": ["Dropout", "Embedding", "False", "Linear", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "__init__", "class", "col_embeddings", "col_indices", "column_embedder", "config", "def", "device", "dropout", "dropout_rate", "embeddings", "flattened_patches", "forward", "hidden_size", "inplace", "long", "max_num_patches", "nn", "patch_embed_hidden_size", "patch_projection", "return", "row_embedder", "row_embeddings", "row_indices", "self", "super", "to", "torch"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5VisionMlp": ["ACT2FN", "Dropout", "False", "Linear", "ModelVisionConfig", "ModelVisionMlp", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_gelu", "hidden_linear", "hidden_size", "hidden_states", "if", "int8", "intermediate_size", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi_0", "wi_1", "wo"], "kosmos2_5/modeling_kosmos2_5.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Falling", "False", "Linear", "ModelVisionAttention", "Module", "None", "This", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "and", "argument", "attention", "attention_dropout", "attention_interface", "attention_mask", "attn_implementation", "attn_output", "attn_weights", "back", "be", "bias", "can", "class", "config", "def", "does", "dropout", "eager", "eager_attention_forward", "else", "forward", "functional", "get", "gradient_checkpointing", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "inner_dim", "input_shape", "is_causal", "key", "key_states", "kwargs", "loading", "logger", "model", "n_heads", "nn", "not", "num_attention_heads", "output", "output_attentions", "query", "query_states", "removed", "reshape", "return", "scaled_dot_product_attention", "scaling", "sdpa", "self", "shape", "super", "support", "the", "to", "torch", "training", "transpose", "using", "value", "value_states", "view", "warning", "warning_once", "when"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5VisionLayer": ["False", "GradientCheckpointingLayer", "ModelLayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionLayer", "ModelVisionMlp", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention", "attention_mask", "attention_output", "bool", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "if", "kwargs", "layer_norm_eps", "layer_output", "mlp", "output_attentions", "outputs", "pre_attention_layer_norm", "pre_mlp_layer_norm", "residual", "return", "self", "self_attn_weights", "super", "torch", "tuple"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5VisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_attn_implementation", "_expand_mask", "_prepare_attention_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "device", "dtype", "else", "enumerate", "expanded_attn_mask", "flash_attention_2", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "input_shape", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "self", "shape", "super", "tgt_len", "to", "torch"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextSinusoidalPositionalEmbedding": ["False", "ModelTextSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "contiguous", "cos", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "detach", "device", "dim", "dtype", "else", "emb", "emb_weights", "embedding_dim", "exp", "expand", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sequence_length", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextFFN": ["ACT2FN", "LayerNorm", "Linear", "ModelTextConfig", "ModelTextFFN", "Module", "__init__", "activation_dropout", "activation_fn", "activation_function", "class", "config", "def", "dropout", "embed_dim", "eps", "fc1", "fc2", "ffn_dim", "ffn_layernorm", "forward", "functional", "hidden_states", "layer_norm_eps", "nn", "p", "return", "self", "super", "training"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Falling", "False", "Linear", "LongTensor", "ModelTextAttention", "Module", "None", "Optional", "Tensor", "This", "TransformersKwargs", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "argument", "attention", "attention_interface", "attention_mask", "attn_implementation", "attn_output", "attn_weights", "back", "be", "bias", "bool", "by", "cache_kwargs", "cache_position", "can", "class", "config", "contiguous", "current_hidden_shape", "current_input_shape", "current_states", "def", "divisible", "does", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "encoder_hidden_states", "f", "float", "forward", "functional", "get", "got", "head_dim", "hidden_shape", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "kwargs", "layer_idx", "loading", "logger", "model", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "past_key_value", "q_proj", "query_states", "raise", "removed", "reshape", "return", "scaled_dot_product_attention", "scaling", "sdpa", "self", "shape", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "update", "using", "v_proj", "value_states", "view", "warning", "warning_once", "when"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextBlock": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelTextAttention", "ModelTextBlock", "ModelTextConfig", "ModelTextFFN", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "attention_dropout", "attention_heads", "attention_mask", "bool", "cache_position", "class", "config", "def", "dropout", "embed_dim", "eps", "ffn", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is_causal", "is_decoder", "kwargs", "layer_idx", "layer_norm_eps", "nn", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextTransformer": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BaseModelOutputWithPastAndCrossAttentions", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "LayerNorm", "LongTensor", "ModelTextBlock", "ModelTextConfig", "ModelTextSinusoidalPositionalEmbedding", "ModelTextTransformer", "Module", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attns", "and", "any", "arange", "at", "attention_mask", "attentions", "batch_size", "bool", "both", "bsz", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dropout", "dtype", "either", "else", "embed_dim", "embed_positions", "embed_scale", "embed_tokens", "embedding_dim", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "functional", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "logger", "long", "make_flex_block_causal_mask", "mask_length", "masked_fill", "math", "max_position_embeddings", "min", "min_dtype", "must", "ne", "nn", "not", "npu", "num_positions", "one", "output", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "padding_mask", "past_key_value", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_ids", "positions", "raise", "range", "reshape", "return", "same", "scale_embedding", "sdpa", "segment_emb", "segment_embeds", "self", "seq_len", "sequence_length", "shape", "size", "specify", "sqrt", "staticmethod", "super", "target_length", "the", "time", "to", "torch", "training", "triu", "type", "use_cache", "using_compilable_cache", "view", "vocab_size", "warning_once", "weight", "with", "xpu", "zero_emb", "zeros"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5ImageToTextProjection": ["False", "Linear", "ModelConfig", "ModelImageToTextProjection", "ModelTextAttention", "Module", "None", "Parameter", "__init__", "attention_dropout", "attention_heads", "attention_mask", "attn_weights", "cat", "class", "config", "def", "dense", "dim", "dropout", "embed_dim", "encoder_hidden_states", "expand", "features", "forward", "hidden_size", "hidden_states", "is_causal", "is_decoder", "key_value_states", "latent_query", "latent_query_num", "nn", "output_attentions", "past_key_value", "randn", "return", "self", "size", "super", "text_config", "torch", "unsqueeze", "vision_config", "x_attn"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5PreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelForConditionalGeneration", "ModelImageToTextProjection", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "ModelTextBlock", "ModelTextForCausalLM", "ModelTextModel", "ModelVisionLayer", "ModelVisionModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_cache_class", "_supports_flash_attn_2", "_supports_sdpa", "bias", "class", "config", "config_class", "data", "def", "elif", "fill_", "getattr", "if", "init_factor", "init_std", "initializer_factor", "initializer_range", "is", "isinstance", "latent_query", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5VisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "ModelLayerNorm", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "attention", "attention_mask", "attentions", "bool", "class", "config", "config_class", "def", "dict", "dim", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flattened_patches", "float", "for", "forward", "get_input_embeddings", "have", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "int", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "not", "output_attentions", "output_hidden_states", "patch_projection", "post_init", "prune_heads", "raise", "return", "self", "sequence_output", "specify", "sum", "super", "to", "torch"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Model_TEXT_INPUTS_DOCSTRING", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "add_start_docstrings_to_model_forward", "attention_mask", "bool", "cache_position", "class", "config", "config_class", "def", "embed_tokens", "forward", "get_input_embeddings", "image_embeds", "image_embeds_position_mask", "input_ids", "inputs_embeds", "kwargs", "list", "model", "nn", "output_attentions", "output_hidden_states", "output_type", "past_key_values", "position_ids", "post_init", "r", "replace_return_docstrings", "return", "self", "set_input_embeddings", "super", "torch", "use_cache", "value"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5Model": ["Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelImageToTextProjection", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionModel", "Model_INPUTS_DOCSTRING", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_CONFIG_FOR_DOC", "__init__", "_from_config", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "bool", "cache_position", "can_return_tuple", "class", "config", "config_class", "def", "dim", "else", "embed_tokens", "flattened_patches", "forward", "functional", "get_input_embeddings", "height", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "image_to_text_projection", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "list", "model", "nn", "normalize", "not", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "position_ids", "post_init", "projection_attentions", "r", "replace_return_docstrings", "return", "self", "set_input_embeddings", "super", "text_config", "text_model", "torch", "use_cache", "value", "vision_config", "vision_model", "vision_model_output", "width"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5TextForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "Linear", "LongTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextForCausalLM", "ModelTextSinusoidalPositionalEmbedding", "ModelTextTransformer", "Model_TEXT_INPUTS_DOCSTRING", "Module", "None", "Optional", "Tensor", "The", "TransformersKwargs", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "argument", "attention_mask", "attentions", "batch_size", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "config_class", "create_position_ids_from_input_ids", "def", "device", "dim", "dtype", "elif", "embed_dim", "embed_tokens", "for", "forward", "get_input_embeddings", "get_output_embeddings", "get_seq_length", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "in", "in_features", "input_ids", "input_shape", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "lm_logits", "lm_loss", "logger", "logits", "loss", "loss_function", "mask_len", "model", "model_inputs", "model_kwargs", "new_embeddings", "new_ones", "nn", "not", "out_features", "output_attentions", "output_hidden_states", "output_type", "outputs", "pad_token_id", "padding_idx", "past_key_values", "past_key_values_length", "position_ids", "post_init", "prepare_inputs_for_generation", "provided", "r", "replace_return_docstrings", "return", "self", "seq_len", "set_input_embeddings", "set_output_embeddings", "shape", "since", "size", "super", "to", "torch", "use_cache", "value", "vocab_size", "warning", "weight", "zeros"], "kosmos2_5/modeling_kosmos2_5.py:Kosmos2_5ForConditionalGeneration": ["Cache", "FloatTensor", "GenerationMixin", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelForConditionalGenerationModelOutput", "ModelImageToTextProjection", "ModelPreTrainedModel", "ModelTextForCausalLM", "ModelVisionModel", "Model_INPUTS_DOCSTRING", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_CONFIG_FOR_DOC", "__init__", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "bool", "cache_position", "can_return_tuple", "class", "config", "config_class", "def", "dim", "else", "embed_tokens", "flattened_patches", "forward", "functional", "get_input_embeddings", "get_output_embeddings", "height", "hidden_states", "if", "image_embeds", "image_embeds_position_mask", "image_to_text_projection", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "lm_outputs", "logits", "loss", "model", "model_inputs", "model_kwargs", "new_embeddings", "nn", "normalize", "not", "output_attentions", "output_hidden_states", "output_type", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "projection_attentions", "r", "replace_return_docstrings", "return", "self", "set_input_embeddings", "set_output_embeddings", "super", "text_config", "text_model", "torch", "use_cache", "value", "vision_config", "vision_model", "vision_model_output", "weight", "width"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attention_mask", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "qwen2_audio/modeling_qwen2_audio.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "and", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "ndim", "nn", "not", "p", "query", "return", "scaling", "shape", "size", "softmax", "torch", "training", "transpose", "value", "view"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "_shape", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "caching", "call", "class", "config", "contiguous", "creating", "decoder", "def", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "k_proj", "key_states", "kwargs", "layer_head_mask", "layer_idx", "logger", "make", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "provide", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "seq_len", "size", "super", "sure", "tensor", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "used", "v_proj", "value_states", "view", "warning_once", "when", "will", "without"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioPreTrainedModel": ["Conv1d", "Embedding", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "audio_config", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioEncoder": ["AvgPool1d", "BaseModelOutput", "Conv1d", "Embedding", "False", "LayerNorm", "LongTensor", "Make", "Model", "ModelEncoder", "ModelEncoderConfig", "ModelEncoderLayer", "ModelPreTrainedModel", "Module", "ModuleList", "None", "The", "True", "ValueError", "_", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "_no_split_modules", "_requires_grad", "all_attentions", "assert", "attention_mask", "attentions", "avg_pooler", "be", "but", "class", "config", "conv1", "conv2", "d_model", "def", "device", "dropout", "dropout_probability", "dtype", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "expected_seq_length", "expects", "f", "features", "for", "forward", "found", "functional", "gelu", "get_input_embeddings", "gradient_checkpointing", "head_mask", "hidden_states", "idx", "if", "in", "input", "input_features", "input_lengths", "inputs_embeds", "is", "it", "kernel_size", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "length", "main_input_name", "math", "max_source_positions", "mel", "nn", "not", "num_mel_bins", "of", "output_attentions", "output_hidden_states", "output_lengths", "p", "pad", "pad_token_id", "padding", "padding_idx", "param", "parameters", "permute", "post_init", "r", "raise", "rand", "range", "requires_grad", "requires_grad_", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shape", "should", "size", "specified", "sqrt", "stride", "super", "sure", "the", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "value", "weight"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioMultiModalProjector": ["Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "True", "__init__", "audio_config", "audio_features", "bias", "class", "config", "d_model", "def", "forward", "hidden_size", "hidden_states", "linear", "nn", "return", "self", "super", "text_config"], "qwen2_audio/modeling_qwen2_audio.py:Qwen2AudioForConditionalGeneration": ["Audio", "AutoModel", "AutoModelForCausalLM", "Cache", "CrossEntropyLoss", "Expanding", "False", "FloatTensor", "GenerationMixin", "LongTensor", "Model", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "The", "This", "True", "Union", "ValueError", "_", "__init__", "_get_feat_extract_output_lengths", "_left_padding", "_merge_input_ids_with_audio_features", "_padding_side", "_right_padding", "_tied_weights_keys", "and", "any", "arange", "are", "args", "attention_mask", "attentions", "audio", "audio_attention_mask", "audio_attention_mask_", "audio_config", "audio_feat_lengths", "audio_features", "audio_features_mask", "audio_output_lengths", "audio_outputs", "audio_to_overwrite", "audio_token_id", "audio_tokens", "audio_tower", "auto_docstring", "batch", "batch_indices", "batch_size", "be", "bool", "both", "breaks", "cache_position", "class", "config", "contiguous", "conv1", "correct", "cumsum", "decoder", "def", "device", "dim", "do", "done", "dtype", "elif", "else", "embed_dim", "expand", "expand_as", "f", "feature_attention_mask", "features", "final_attention_mask", "final_embedding", "final_input_ids", "final_labels", "float", "for", "forward", "from_config", "full", "full_like", "generation", "get", "get_decoder", "get_input_embeddings", "get_output_embeddings", "given", "has", "hidden_states", "if", "ignore_index", "in", "indexing", "inf", "input", "input_features", "input_ids", "inputs", "inputs_embeds", "invalid", "is", "item", "k", "kwargs", "labels", "language_model", "last_hidden_state", "left", "left_padding", "legacy_processing", "lengths_expand", "logger", "logits", "long", "loss", "loss_fct", "masked_audio_features", "masked_fill_", "masked_scatter", "match", "max", "max_audio_tokens", "max_mel_seq_len", "max_seq_len", "max_token_num", "model", "model_inputs", "multi_modal_projector", "n_audio_features", "n_audio_tokens", "nb_audio_pad", "new_embeddings", "new_token_positions", "nn", "non_audio_indices", "not", "num_audio_tokens", "num_audios", "num_special_audio_tokens", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "padding_mask", "padding_side", "past_key_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "prevents", "processing", "property", "provided", "r", "raise", "reshape", "return", "return_dict", "right", "selected_audio_feature", "self", "seq_indices", "seq_range", "sequence_length", "set_decoder", "set_input_embeddings", "set_output_embeddings", "setter", "shape", "shift_attention_mask", "shift_labels", "shift_logits", "should", "side", "size", "special_audio_mask", "special_audio_token_mask", "str", "sum", "super", "target_device", "text_config", "text_to_overwrite", "the", "to", "token_placeholder_num", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "val", "value", "view", "vocab_size", "warning_once", "weight", "where", "while", "wrong", "zero", "zeros", "zeros_like"], "emu3/modeling_emu3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "emu3/modeling_emu3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "emu3/modeling_emu3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "emu3/modeling_emu3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "emu3/modeling_emu3.py:Emu3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "emu3/modeling_emu3.py:Emu3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "emu3/modeling_emu3.py:Emu3MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "emu3/modeling_emu3.py:Emu3DecoderLayer": ["Cache", "Dropout", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "nn", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "emu3/modeling_emu3.py:Emu3VQVAEVectorQuantizer": ["Embedding", "ModelVQVAEConfig", "ModelVQVAEVectorQuantizer", "Module", "Tensor", "True", "__init__", "argmin", "batch_size", "channels", "class", "codebook_size", "config", "contiguous", "data", "def", "dim", "distances", "embed_dim", "embedding", "embedding_sum", "forward", "height", "hidden_state", "hidden_state_flattened", "hidden_state_sum", "keepdim", "matmul", "min_encoding_indices", "nn", "permute", "return", "self", "shape", "sum", "super", "temporal", "torch", "transpose", "uniform_", "view", "weight", "width"], "emu3/modeling_emu3.py:Emu3VQVAEEncoderConvDownsample": ["Conv2d", "F", "ModelVQVAEEncoderConvDownsample", "Module", "__init__", "class", "constant", "conv", "def", "forward", "hidden_states", "in_channels", "kernel_size", "mode", "nn", "pad", "padding", "return", "self", "stride", "super", "value"], "emu3/modeling_emu3.py:Emu3VQVAEEncoderConvUpsample": ["Conv2d", "F", "ModelVQVAEEncoderConvUpsample", "Module", "__init__", "class", "conv", "def", "forward", "hidden_states", "in_channels", "interpolate", "kernel_size", "mode", "nearest", "nn", "padding", "return", "scale_factor", "self", "stride", "super"], "emu3/modeling_emu3.py:Emu3VQVAEConv3d": ["Conv3d", "F", "ModelVQVAEConv3d", "Module", "Tensor", "__init__", "class", "conv", "def", "for", "forward", "hidden_states", "in", "in_channel", "int", "kernel_size", "nn", "one_kernel", "one_stride", "out_channel", "pad", "pad_size", "padding", "padding_sizes", "return", "self", "stride", "super", "torch", "tuple", "zip"], "emu3/modeling_emu3.py:Emu3VQVAESpatialNorm": ["Conv2d", "F", "GroupNorm", "ModelVQVAESpatialNorm", "Module", "Tensor", "True", "__init__", "affine", "class", "conv_b", "conv_y", "def", "eps", "forward", "hidden_states", "in_channels", "int", "interpolate", "kernel_size", "mode", "nearest", "nn", "norm_layer", "num_channels", "num_groups", "out_channels", "padding", "quant_states", "return", "self", "shape", "size", "stride", "super", "torch"], "emu3/modeling_emu3.py:Emu3VQVAETemporalUpsample": ["F", "ModelVQVAEConv3d", "ModelVQVAETemporalUpsample", "Module", "Tensor", "__init__", "batch_size", "channels", "class", "contiguous", "conv", "def", "forward", "height", "hidden_states", "in_channel", "int", "interpolate", "kernel_size", "mode", "nearest", "nn", "out_channel", "permute", "return", "scale_factor", "self", "shape", "stride", "super", "temporal", "torch", "view", "width"], "emu3/modeling_emu3.py:Emu3VQVAETemporalDownsample": ["ModelVQVAEConv3d", "ModelVQVAETemporalDownsample", "Module", "Tensor", "__init__", "class", "conv", "def", "forward", "hidden_states", "in_channel", "int", "kernel_size", "nn", "out_channel", "return", "self", "stride", "super", "torch"], "emu3/modeling_emu3.py:Emu3VQVAETemporalResnetBlock": ["BatchNorm3d", "Conv3d", "ModelVQVAEConv3d", "ModelVQVAETemporalResnetBlock", "Module", "None", "__init__", "class", "conv1", "conv2", "def", "else", "forward", "hidden_states", "if", "in_channels", "is", "kernel_size", "nin_shortcut", "nn", "norm1", "norm2", "out_channels", "padding", "residual", "return", "self", "sigmoid", "stride", "super", "torch"], "emu3/modeling_emu3.py:Emu3VQVAEResnetBlock": ["Conv2d", "GroupNorm", "ModelVQVAEResnetBlock", "ModelVQVAESpatialNorm", "Module", "None", "Optional", "Tensor", "True", "__init__", "affine", "class", "conv1", "conv2", "def", "else", "eps", "forward", "hidden_states", "if", "in_channels", "int", "is", "kernel_size", "nin_shortcut", "nn", "norm1", "norm2", "norm_args", "num_channels", "num_groups", "out_channels", "padding", "quant_channels", "residual", "return", "self", "sigmoid", "stride", "super", "torch"], "emu3/modeling_emu3.py:Emu3VQVAEAttentionBlock": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVQVAEAttentionBlock", "ModelVQVAEConfig", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "emu3/modeling_emu3.py:Emu3VQVAEGroupNorm": ["F", "GroupNorm", "ModelVQVAEGroupNorm", "None", "__init__", "bias", "class", "def", "eps", "forward", "group_norm", "input", "kwargs", "nn", "num_groups", "quant_states", "return", "self", "super", "weight"], "emu3/modeling_emu3.py:Emu3VQVAEMiddleBlock": ["FloatTensor", "ModelVQVAEAttentionBlock", "ModelVQVAEGroupNorm", "ModelVQVAEMiddleBlock", "ModelVQVAEResnetBlock", "ModelVQVAESpatialNorm", "Module", "None", "Optional", "True", "__init__", "affine", "attn_1", "attn_norm", "batch_size", "block_1", "block_2", "channels", "class", "config", "def", "else", "eps", "forward", "height", "hidden_states", "if", "in_channels", "is", "nn", "num_channels", "num_groups", "out_channels", "permute", "quant_channels", "quant_states", "reshape", "residual", "return", "self", "shape", "super", "torch", "transpose", "view", "width"], "emu3/modeling_emu3.py:Emu3VQVAEDownBlock": ["FloatTensor", "GroupNorm", "ModelVQVAEAttentionBlock", "ModelVQVAEDownBlock", "ModelVQVAEEncoderConvDownsample", "ModelVQVAEResnetBlock", "Module", "ModuleList", "None", "True", "__init__", "affine", "and", "append", "attn", "attn_norms", "attn_resolutions", "base_channels", "batch_size", "block", "block_in", "block_out", "blocks", "channel_multiplier", "channels", "class", "config", "def", "down", "downsample", "enumerate", "eps", "for", "forward", "height", "hidden_states", "i_block", "i_level", "if", "in", "in_channel_multiplier", "in_channels", "is", "len", "nn", "not", "num_channels", "num_groups", "num_res_blocks", "num_resolutions", "out_channels", "permute", "range", "reshape", "residual", "return", "self", "shape", "super", "torch", "transpose", "tuple", "view", "width"], "emu3/modeling_emu3.py:Emu3VQVAEUpBlock": ["FloatTensor", "ModelVQVAEAttentionBlock", "ModelVQVAEEncoderConvUpsample", "ModelVQVAEResnetBlock", "ModelVQVAESpatialNorm", "ModelVQVAEUpBlock", "Module", "ModuleList", "__init__", "append", "attn", "attn_norms", "attn_resolutions", "base_channels", "batch_size", "block", "block_in", "block_out", "blocks", "channel_multiplier", "channels", "class", "config", "def", "embed_dim", "enumerate", "for", "forward", "height", "hidden_states", "i_block", "i_level", "if", "in", "in_channels", "insert", "len", "nn", "num_res_blocks", "num_resolutions", "out_channels", "permute", "quant_channels", "quant_states", "range", "reshape", "residual", "return", "reversed", "self", "shape", "super", "torch", "transpose", "up", "upsample", "view", "width"], "emu3/modeling_emu3.py:Emu3VQVAEEncoder": ["Conv2d", "GroupNorm", "LongTensor", "ModelVQVAEDownBlock", "ModelVQVAEEncoder", "ModelVQVAEMiddleBlock", "ModelVQVAETemporalDownsample", "ModelVQVAETemporalResnetBlock", "Module", "ModuleList", "True", "_", "__init__", "affine", "append", "base_channels", "block_in", "channel_multiplier", "class", "config", "conv", "conv_in", "conv_out", "def", "double_latent", "down_block", "else", "eps", "for", "forward", "hidden_states", "i", "if", "in", "in_channels", "int", "kernel_size", "latent_channels", "layer", "log2", "math", "middle_block", "nn", "norm_out", "num_channels", "num_groups", "num_res_blocks", "out_channels", "padding", "permute", "pixel_values", "range", "reshape", "return", "self", "shape", "sigmoid", "stride", "super", "temporal_dim", "temporal_down_blocks", "temporal_downsample_factor", "time_conv", "time_res_conv", "time_res_stack", "torch"], "emu3/modeling_emu3.py:Emu3VQVAEDecoder": ["Conv2d", "ModelVQVAEConfig", "ModelVQVAEDecoder", "ModelVQVAEMiddleBlock", "ModelVQVAESpatialNorm", "ModelVQVAETemporalResnetBlock", "ModelVQVAETemporalUpsample", "ModelVQVAEUpBlock", "Module", "ModuleList", "Tensor", "_", "__init__", "append", "base_channels", "block_in", "cat", "channel_multiplier", "chunk", "class", "config", "conv", "conv_in", "conv_out", "def", "dim", "embed_dim", "for", "forward", "hidden_quant_states", "hidden_states", "i", "in", "in_channels", "int", "kernel_size", "latent_channels", "layer", "log2", "math", "middle_block", "nn", "norm_out", "num_res_blocks", "out_channels", "padding", "permute", "quant_channels", "quant_states", "range", "reshape", "return", "self", "shape", "sigmoid", "stride", "super", "temp_upsample_block_num", "temporal_downsample_factor", "time_conv", "time_res_conv", "time_res_stack", "torch", "up_block"], "emu3/modeling_emu3.py:Emu3VQVAE": ["BatchNorm2d", "BatchNorm3d", "Conv2d", "Conv3d", "Embedding", "GroupNorm", "Linear", "ModelVQVAE", "ModelVQVAEAttentionBlock", "ModelVQVAEConfig", "ModelVQVAEConv3d", "ModelVQVAEDecoder", "ModelVQVAEEncoder", "ModelVQVAEResnetBlock", "ModelVQVAETemporalResnetBlock", "ModelVQVAEVectorQuantizer", "Modelvideovq", "None", "PreTrainedModel", "Tensor", "True", "_", "__init__", "_calculate_fan_in_and_fan_out", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "base_model_prefix", "batch_size", "bias", "bound", "channel_multiplier", "channels", "class", "codes", "config", "constant_", "contiguous", "data", "decode", "decoder", "def", "elif", "else", "embed_dim", "embedding", "encode", "encoder", "eval", "fan_in", "fan_out", "flatten", "for", "height", "hidden_states", "if", "image_sizes", "image_tokens", "in", "init", "int", "is", "is_image", "isinstance", "kaiming_normal_", "kaiming_uniform_", "kernel_size", "latent_channels", "len", "main_input_name", "math", "mode", "module", "ndim", "nn", "nonlinearity", "normal_", "not", "out_channels", "padding_idx", "permute", "pixel_values", "post_init", "post_quant", "post_quant_conv", "quant", "quant_conv", "quantize", "relu", "repeat", "reshape", "return", "self", "shape", "single_image", "size", "spatial_scale_factor", "sqrt", "squeeze", "stride", "super", "temporal", "temporal_downsample_factor", "torch", "uniform_", "unsqueeze", "video", "view", "vision_spatial_factor", "weight", "width", "zero_", "zip"], "emu3/modeling_emu3.py:Emu3ImageVocabularyMapping": ["ModelImageVocabularyMapping", "Tensor", "__init__", "bpe2img", "bpe2img_mapping_tensor", "cached_property", "cat", "class", "convert_bpe2img", "convert_img2bpe", "cpu", "def", "device", "dim", "dtype", "eol_row", "eol_token_id", "extra_200", "for", "get", "if", "image", "image_token_id", "image_tokens", "image_tokens_str", "img2bpe", "img2bpe_mapping_tensor", "img_batch", "img_tokens", "in", "int", "items", "k", "keys", "list", "mapping", "max", "name", "ones", "return", "self", "shape", "sorted", "startswith", "to", "token", "torch", "v", "val", "visual", "vocab_map", "zeros"], "emu3/modeling_emu3.py:Emu3PreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "base_model_prefix", "causal_mask", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "emu3/modeling_emu3.py:Emu3RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "emu3/modeling_emu3.py:Emu3TextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTextModel", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "emu3/modeling_emu3.py:Emu3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "emu3/modeling_emu3.py:Emu3Model": ["Cache", "CausalLMOutputWithPast", "FloatTensor", "Image", "LongTensor", "ModelImageVocabularyMapping", "ModelModel", "ModelPreTrainedModel", "ModelTextModel", "ModelVQVAE", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "_from_config", "all", "and", "at", "attention_mask", "auto_docstring", "bool", "both", "bpe_tokens", "bpe_tokens_list", "cache_position", "can_return_tuple", "cannot", "cat", "class", "config", "convert_bpe2img", "convert_img2bpe", "decode", "decode_image_tokens", "decoder", "def", "device", "dim", "do", "dtype", "either", "else", "encode", "expand_as", "f", "features", "flatten", "for", "forward", "get_decoder", "get_image_features", "get_image_tokens", "get_input_embeddings", "get_placeholder_mask", "height", "if", "image", "image_embeds", "image_features", "image_sizes", "image_token_id", "image_tokens", "image_tokens_list", "in", "input_ids", "inputs_embeds", "int", "is", "kwargs", "long", "masked_scatter", "match", "model", "must", "n_image_features", "n_image_tokens", "no_grad", "not", "numel", "one", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "r", "raise", "return", "same", "self", "sequences", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "split", "split_sizes", "sum", "super", "tensor", "text_config", "text_model", "the", "time", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "value", "view", "vision_spatial_factor", "vocabulary_map", "vocabulary_mapping", "vq_config", "vqmodel", "width"], "emu3/modeling_emu3.py:Emu3ForConditionalGeneration": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decode_image_tokens", "decoder", "def", "else", "forward", "get_decoder", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "text_model", "torch", "tuple", "use_cache", "value", "vocab_size", "vocabulary_mapping", "vqmodel", "weight"], "colpali/modeling_colpali.py:ColPaliPreTrainedModel": ["Conv2d", "Embedding", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "text_config", "vlm_config", "weight", "zero_"], "colpali/modeling_colpali.py:ColPaliForRetrievalOutput": ["Cache", "FloatTensor", "ModelForRetrievalOutput", "ModelOutput", "None", "Optional", "Tensor", "attentions", "class", "embeddings", "hidden_states", "image_hidden_states", "loss", "past_key_values", "r", "torch", "tuple"], "colpali/modeling_colpali.py:ColPaliForRetrieval": ["AutoModelForImageTextToText", "Embedding", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForRetrieval", "ModelForRetrievalOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "dim", "dtype", "else", "embedding_dim", "embedding_proj_layer", "embeddings", "f", "for", "forward", "from_config", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "in", "input_ids", "int", "is", "k", "keepdim", "kwargs", "language_model", "last_hidden_states", "lm_head", "mean_resizing", "model", "model_embeds", "multi_modal_projector", "new_embeddings", "new_num_tokens", "nn", "norm", "not", "num_embeddings", "or", "output_attentions", "output_hidden_states", "pad_to_multiple_of", "past_key_values", "pixel_values", "post_init", "proj_dtype", "resize_token_embeddings", "return", "return_dict", "self", "set_input_embeddings", "set_output_embeddings", "super", "text_config", "tie_weights", "to", "torch", "unsqueeze", "use_return_dict", "value", "vision_tower", "vlm", "vlm_config", "vlm_hidden_states", "vlm_image_hidden_states", "vlm_output", "vocab_size", "weight"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "phi4_multimodal/modeling_phi4_multimodal.py:simple_eager_attention_forward": ["Model_eager_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key_states", "kwargs", "matmul", "module", "nn", "not", "p", "query_states", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value_states"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "class", "config", "def", "dropout", "eager", "else", "embed_dim", "forward", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "is_causal", "k_proj", "key_states", "kwargs", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "simple_eager_attention_forward", "super", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoderLayer", "ModelVisionMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "phi4_multimodal/modeling_phi4_multimodal.py:_trunc_normal_": ["The", "_trunc_normal_", "a", "add_", "b", "be", "clamp_", "def", "distribution", "erf", "erfinv_", "from", "if", "in", "incorrect", "init", "is", "l", "math", "max", "may", "mean", "min", "more", "mul_", "nn", "norm_cdf", "of", "or", "return", "sqrt", "stacklevel", "std", "tensor", "than", "trunc_normal_", "u", "uniform_", "values", "warn", "warnings", "x"], "phi4_multimodal/modeling_phi4_multimodal.py:trunc_normal_tf_": ["Model_normal_tf_", "Tensor", "_Model_normal_", "a", "add_", "b", "def", "float", "mean", "mul_", "no_grad", "std", "tensor", "torch", "with"], "phi4_multimodal/modeling_phi4_multimodal.py:variance_scaling_": ["Model", "Model_scaling_", "ValueError", "_calculate_fan_in_and_fan_out", "bound", "def", "denom", "distribution", "elif", "else", "f", "fan_avg", "fan_in", "fan_out", "if", "invalid", "math", "mode", "no_grad", "normal", "normal_", "raise", "scale", "sqrt", "std", "tensor", "torch", "trunc_normal_tf_", "truncated_normal", "uniform", "uniform_", "with"], "phi4_multimodal/modeling_phi4_multimodal.py:lecun_normal_": ["Model_normal_", "def", "distribution", "fan_in", "mode", "tensor", "truncated_normal", "variance_scaling_"], "phi4_multimodal/modeling_phi4_multimodal.py:default_flax_embed_init": ["Model_flax_embed_init", "def", "distribution", "fan_in", "mode", "normal", "tensor", "variance_scaling_"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoderLayer", "ModelVisionMLP", "ModelVisionMultiheadAttentionPoolingHead", "ModelVisionPreTrainedModel", "Model_vision", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "default_flax_embed_init", "elif", "else", "fc1", "fc2", "fill_", "hidden_size", "hidden_states", "if", "in_proj_bias", "in_proj_weight", "init", "is", "isinstance", "k_proj", "lecun_normal_", "module", "nn", "normal_", "not", "np", "out_proj", "position_embedding", "probe", "q_proj", "self", "sqrt", "std", "supports_gradient_checkpointing", "v_proj", "weight", "width", "zero_", "zeros_"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionEmbeddings": ["BoolTensor", "Conv2d", "Embedding", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "True", "__init__", "align_corners", "and", "arange", "batch_idx", "batch_size", "bicubic", "boundaries", "bucket_coords_h", "bucket_coords_w", "bucketize", "class", "config", "cpu", "def", "device", "dim", "embeddings", "enumerate", "fill_value", "flatten", "for", "forward", "fractional_coords_h", "fractional_coords_w", "full", "functional", "height", "hidden_size", "if", "image_size", "in", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "max_im_h", "max_im_w", "max_nb_patches_h", "max_nb_patches_w", "mode", "nb_patches_h", "nb_patches_w", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_patches_per_side", "num_positions", "out_channels", "p_attn_mask", "padding", "patch_attention_mask", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "pixel_values", "pos_ids", "position_embedding", "position_ids", "reshape", "return", "right", "self", "shape", "size", "sqrt_num_positions", "stride", "sum", "super", "to", "torch", "torch_int", "transpose", "unsqueeze", "valid", "view", "weight", "width"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionMultiheadAttentionPoolingHead": ["LayerNorm", "ModelVisionConfig", "ModelVisionMLP", "ModelVisionMultiheadAttentionPoolingHead", "Module", "MultiheadAttention", "Parameter", "True", "__init__", "attention", "attention_mask", "batch_first", "batch_size", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_state", "key", "key_padding_mask", "layer_norm_eps", "layernorm", "mlp", "nn", "num_attention_heads", "probe", "query", "randn", "repeat", "residual", "return", "self", "shape", "super", "torch", "value"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalVisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "BoolTensor", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionModel", "ModelVisionMultiheadAttentionPoolingHead", "ModelVisionPreTrainedModel", "Module", "None", "Optional", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "any", "attention_mask", "batch_size", "bool", "check_model_inputs", "class", "config", "def", "device", "dtype", "else", "embeddings", "encoder", "encoder_outputs", "eps", "flash_attention_2", "forward", "get_input_embeddings", "head", "hidden_size", "hidden_state", "hidden_states", "if", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "main_input_name", "nn", "not", "ones", "patch_attention_mask", "patch_embedding", "patch_size", "pixel_values", "pooled_output", "pooler_output", "post_init", "post_layernorm", "return", "self", "size", "super", "torch", "view"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalImageEmbedding": ["AvgPool2d", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelImageEmbedding", "ModelVisionModel", "Module", "None", "Optional", "Parameter", "ReflectionPad2d", "Tensor", "True", "__init__", "_from_config", "accumulate", "append", "area_ratio", "as_tuple", "attention_mask", "autocast", "base_feat_size", "batch_size", "bias", "bool", "cat", "class", "config", "contiguous", "crop_size", "def", "device", "device_type", "dim", "drop", "dtype", "else", "embd_pdrop", "embeddings", "enabled", "feature_layer", "flatten", "for", "forward", "functional", "gelu", "get_img_features", "getattr", "global_img", "global_img_feature_extensor", "height", "height_ratio", "hidden_size", "hidden_states", "idx", "if", "image_attention_mask", "image_dim_out", "image_embeds", "image_pixel_values", "image_size", "image_sizes", "image_token_compression", "image_token_id", "img_embeds", "img_feature", "img_feature_proj", "img_features", "img_processor", "img_processor_output", "img_processor_padding", "img_projection_down", "img_projection_up", "img_set_tensor", "in", "index_put", "indices", "input_ids", "inputs_embeds", "int", "is", "item", "kernel_size", "layer_idx", "math", "merged_img_set_tensor", "n_patches", "nn", "no_grad", "nonzero", "not", "np", "num_img_tokens", "output_hidden_states", "output_img", "output_imgs", "patch_attention_mask", "patch_embedding", "patch_feature", "patch_size", "permute", "positions_tuple", "range", "repeat", "reshape", "reshaped_image_attention_mask", "return", "self", "shape", "size", "sqrt", "squeeze", "stride", "sub_img", "sub_img_feature_extensor", "sum", "super", "target_device", "target_dtype", "temporary_extensor", "to", "torch", "transpose", "type", "useful_height", "useful_width", "values", "view", "vision_config", "weight", "width", "width_ratio", "with", "zeros"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioMLP": ["ACT2FN", "Dropout", "LayerNorm", "Linear", "ModelAudioConfig", "ModelAudioMLP", "Module", "__init__", "act_fn", "activation", "chunk", "class", "config", "def", "dim", "down_proj", "dropout", "dropout_rate", "forward", "gate", "gate_up_proj", "hidden_size", "hidden_states", "intermediate_size", "layer_norm", "nn", "out", "return", "self", "super", "up_states"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Linear", "ModelAudioAttention", "ModelAudioConfig", "Module", "Tensor", "True", "_", "__init__", "_attn_implementation", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "bias", "class", "config", "contiguous", "def", "dropout", "dropout_rate", "eager", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "is_causal", "k_proj", "key_states", "kwargs", "nn", "not", "num_attention_heads", "o_proj", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "simple_eager_attention_forward", "super", "torch", "training", "transpose", "v_proj", "value_states", "view"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioDepthWiseSeparableConv1d": ["Conv1d", "ModelAudioConfig", "ModelAudioDepthWiseSeparableConv1d", "Module", "__init__", "class", "config", "def", "depthwise_multiplier", "depthwise_separable_out_channel", "dw_conv", "forward", "groups", "hidden_size", "hidden_states", "int", "kernel_size", "nn", "padding", "pw_conv", "return", "self", "super"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioGluPointWiseConv": ["ACT2FN", "Conv1d", "ModelAudioConfig", "ModelAudioGluPointWiseConv", "Module", "Parameter", "__init__", "b1", "b2", "class", "config", "conv_glu_type", "def", "ext_pw_conv_1d", "ext_pw_out_channel", "forward", "glu_act", "hidden_size", "hidden_states", "kernel_size", "nn", "out", "output_dim", "permute", "return", "self", "stride", "super", "torch", "zeros"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioConvModule": ["ACT2FN", "Conv1d", "Dropout", "LayerNorm", "ModelAudioConfig", "ModelAudioConvModule", "ModelAudioDepthWiseSeparableConv1d", "ModelAudioGluPointWiseConv", "Module", "Tensor", "__init__", "act", "class", "config", "conv_activation", "def", "dropout", "dropout_rate", "dw_sep_conv_1d", "ext_pw_conv_1d", "ext_pw_out_channel", "forward", "glu", "hidden_size", "hidden_states", "if", "kernel_size", "layer_norm", "nn", "out", "padding", "permute", "return", "self", "stride", "super", "torch"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioConformerEncoderLayer": ["LayerNorm", "ModelAudioAttention", "ModelAudioConfig", "ModelAudioConformerEncoderLayer", "ModelAudioConvModule", "ModelAudioMLP", "Module", "Tensor", "__init__", "attention_mask", "class", "config", "conv", "def", "feed_forward_in", "feed_forward_out", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_att", "nn", "out", "residual", "return", "self", "self_attn", "super", "torch"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioNemoConvSubsampling": ["ACT2FN", "Conv2d", "Linear", "ModelAudioConfig", "ModelAudioNemoConvSubsampling", "Module", "None", "Optional", "Sequential", "Tensor", "_", "__init__", "act_fn", "arange", "arange_", "b", "ceil", "class", "config", "conv", "conv_channels", "def", "device", "expand", "extend", "feature_lens", "for", "forward", "groups", "hidden_size", "hidden_states", "if", "in", "int", "is", "kernel_size", "layers", "log", "mask", "math", "max_audio_length", "nemo_activation", "nemo_conv_channels", "nemo_final_size", "nn", "out", "pad_mask", "padding", "padding_length", "range", "reshape", "return", "sampling_num", "self", "shape", "size", "stride", "subsampling_factor", "sum", "super", "t", "time_reduction", "torch", "transpose", "unsqueeze"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioRelativeAttentionBias": ["Embedding", "ModelAudioConfig", "ModelAudioRelativeAttentionBias", "Module", "None", "__init__", "abs", "arange", "att_bias", "bias_idx", "bias_max_distance", "bias_symmetric", "bias_values", "class", "config", "context_position", "def", "device", "dtype", "else", "forward", "if", "long", "masked_fill", "max_distance", "max_pos", "memory_position", "nn", "not", "num_attention_heads", "num_buckets", "permute", "relative_position", "return", "self", "size", "super", "symmetric", "torch", "unsqueeze", "x"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioMeanVarianceNormLayer": ["ModelAudioConfig", "ModelAudioMeanVarianceNormLayer", "Module", "__init__", "class", "config", "def", "forward", "global_invstd", "global_mean", "input_size", "nn", "ones", "register_buffer", "return", "self", "super", "torch", "x", "zeros"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioPreTrainedModel": ["ModelAudioConfig", "ModelAudioConformerEncoderLayer", "ModelAudioGluPointWiseConv", "ModelAudioPreTrainedModel", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "b1", "b2", "class", "config", "data", "def", "if", "isinstance", "module", "self", "super", "supports_gradient_checkpointing", "zero_"], "phi4_multimodal/modeling_phi4_multimodal.py:unfold_tensor": ["D", "F", "Model", "Model_tensor", "None", "_", "contiguous", "def", "kernel_size", "max_seq_len", "new_bsz", "permute", "return", "shape", "slen", "stride", "tensor", "transpose", "view"], "phi4_multimodal/modeling_phi4_multimodal.py:adaptive_enc_mask": ["Model_enc_mask", "Tensor", "arange", "boundary_left", "boundary_right", "chunk_start_idx", "def", "end_pad", "expand", "functional", "idx", "idx_left", "idx_right", "left_window", "len", "long", "mask_left", "mask_right", "nn", "nonzero", "pad", "return", "right_window", "seq_range", "seq_range_expand", "start_pad", "torch", "unsqueeze", "value", "x_len"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioModel": ["F", "False", "ModelAudioConfig", "ModelAudioConformerEncoderLayer", "ModelAudioMeanVarianceNormLayer", "ModelAudioModel", "ModelAudioNemoConvSubsampling", "ModelAudioPreTrainedModel", "ModelAudioRelativeAttentionBias", "ModuleList", "None", "Optional", "Tensor", "The", "True", "ValueError", "Your", "_", "__init__", "_streaming_mask", "adaptive_enc_mask", "after", "and", "arange", "attention_mask", "batch_size", "bool", "bs", "calculate_hs_mask", "ceil", "chunk_pad_size", "chunk_size", "chunk_start_idx", "class", "config", "constant", "def", "device", "elif", "else", "embed", "embed_dim", "enc_streaming_mask", "encoder_embedding", "encoders", "expand", "extra_padded_subsamlped_pad_mask", "f", "feature", "feature_lens", "float", "for", "forward", "forward_embeddings", "gradient_checkpointing", "hidden_states", "hidden_states_pad", "hs_mask", "if", "in", "input", "insert", "invalid", "is", "layer", "left_chunk", "left_window", "length", "mask", "masks", "masks_unfold", "math", "max_audio_length", "max_seq_len", "nn", "not", "np", "num_blocks", "pad", "pad_mask", "padding_length", "post_init", "raise", "rand", "random", "range", "reduction", "relative_attention_bias", "relative_attention_bias_layer", "reshape", "return", "self", "seq_len", "sequence", "shape", "short", "size", "squeeze", "streaming_mask", "subsampled_pad_mask", "sum", "super", "time", "time_reduction", "to", "too", "torch", "training", "unfold_tensor", "unfolded", "unsqueeze"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAudioEmbedding": ["Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelAudioEmbedding", "ModelAudioModel", "ModelConfig", "Module", "None", "Tensor", "True", "__init__", "_from_config", "accumulate", "as_tuple", "audio_attention_mask", "audio_config", "audio_embed_sizes", "audio_embeds", "audio_encoder_hidden_states", "audio_input_features", "audio_projection_mode", "audio_token_id", "autocast", "bias", "cat", "class", "config", "def", "device", "device_type", "dim", "down_proj", "down_proj_for_speech", "down_proj_for_vision_speech", "downsample_rate", "drop", "dtype", "else", "embd_pdrop", "enabled", "encoder", "feature_layer", "for", "forward", "functional", "gelu", "hidden_size", "i", "if", "in", "index_put", "indices", "input_ids", "inputs_embeds", "layer_idx", "len", "merged_audio_embeds", "nn", "no_grad", "nonzero", "positions_tuple", "range", "return", "self", "speech", "super", "target_device", "target_dtype", "to", "torch", "type", "up_proj", "up_proj_for_speech", "up_proj_for_vision_speech", "values", "with"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalMLP": ["ACT2FN", "False", "FloatTensor", "Linear", "ModelMLP", "Module", "__init__", "activation_fn", "bias", "chunk", "class", "config", "def", "dim", "down_proj", "forward", "gate", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch", "up_states"], "phi4_multimodal/modeling_phi4_multimodal.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "phi4_multimodal/modeling_phi4_multimodal.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "phi4_multimodal/modeling_phi4_multimodal.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "phi4_multimodal/modeling_phi4_multimodal.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "op_size", "past_key_value", "past_key_values", "position_embeddings", "qkv", "qkv_proj", "query_pos", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "value_states", "version", "view"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalDecoderLayer": ["Cache", "Dropout", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "nn", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "resid_attn_dropout", "resid_mlp_dropout", "resid_pdrop", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalFeatureEmbedding": ["FloatTensor", "LongTensor", "ModelAudioEmbedding", "ModelConfig", "ModelFeatureEmbedding", "ModelImageEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "and", "any", "audio_attention_mask", "audio_config", "audio_embed", "audio_embed_sizes", "audio_embeds", "audio_input_features", "audio_projection_mode", "audio_token_id", "class", "config", "def", "elif", "else", "forward", "if", "image_attention_mask", "image_embed", "image_embeds", "image_pixel_values", "image_position_mask", "image_sizes", "image_token_id", "input_ids", "inputs_embeds", "is", "nn", "no_grad", "non_image_position_mask", "not", "return", "self", "speech", "super", "torch", "unsqueeze", "vision", "vision_config", "with"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelImageEmbedding", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_version", "attentions", "base_model_prefix", "class", "config", "data", "def", "global_img_feature_extensor", "hidden_states", "if", "isinstance", "model", "module", "past_key_values", "self", "sub_img_feature_extensor", "super", "supports_gradient_checkpointing", "zero_"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalModel": ["BaseModelOutputWithPast", "Cache", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelFeatureEmbedding", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "audio_attention_mask", "audio_embed_sizes", "audio_input_features", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "else", "embd_pdrop", "embed_dropout", "embed_tokens", "embed_tokens_extend", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "image_attention_mask", "image_pixel_values", "image_sizes", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "phi4_multimodal/modeling_phi4_multimodal.py:Phi4MultimodalForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "and", "attention_mask", "attentions", "audio_attention_mask", "audio_embed_sizes", "audio_input_features", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "image_attention_mask", "image_pixel_values", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "nn", "not", "original_max_position_embeddings", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "past_length", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "rope_scaling", "self", "shape", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "vitmatte/modeling_vitmatte.py:ImageMattingOutput": ["FloatTensor", "ModelMattingOutput", "ModelOutput", "None", "Optional", "alphas", "attentions", "class", "hidden_states", "loss", "r", "torch", "tuple"], "vitmatte/modeling_vitmatte.py:VitMattePreTrainedModel": ["BatchNorm2d", "Conv2d", "ModelConfig", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "bias", "class", "config", "data", "def", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "vitmatte/modeling_vitmatte.py:VitMatteBasicConv3x3": ["BatchNorm2d", "Conv2d", "False", "ModelBasicConv3x3", "Module", "ReLU", "__init__", "batch_norm", "batch_norm_eps", "bias", "class", "config", "conv", "def", "eps", "forward", "hidden_state", "in_channels", "kernel_size", "nn", "out_channels", "padding", "relu", "return", "self", "stride", "super"], "vitmatte/modeling_vitmatte.py:VitMatteConvStream": ["ModelBasicConv3x3", "ModelConvStream", "Module", "ModuleList", "None", "__init__", "append", "backbone_config", "class", "config", "conv_chans", "convs", "convstream_hidden_sizes", "def", "detailed_feature_map_", "detailed_feature_map_0", "embeddings", "for", "forward", "i", "if", "in", "in_chan_", "in_channels", "is", "len", "name_", "nn", "not", "num_channels", "out_chan_", "out_channels", "out_dict", "pixel_values", "range", "return", "self", "str", "super"], "vitmatte/modeling_vitmatte.py:VitMatteFusionBlock": ["False", "ModelBasicConv3x3", "ModelFusionBlock", "Module", "__init__", "align_corners", "bilinear", "cat", "class", "config", "conv", "def", "detailed_feature_map", "dim", "features", "forward", "functional", "in_channels", "interpolate", "mode", "nn", "out", "out_channels", "padding", "return", "scale_factor", "self", "stride", "super", "torch", "upscaled_features"], "vitmatte/modeling_vitmatte.py:VitMatteHead": ["BatchNorm2d", "Conv2d", "ModelHead", "Module", "ReLU", "Sequential", "True", "__init__", "class", "config", "def", "forward", "fusion_hidden_sizes", "hidden_state", "in_channels", "kernel_size", "matting_convs", "mid_channels", "nn", "padding", "return", "self", "stride", "super"], "vitmatte/modeling_vitmatte.py:VitMatteDetailCaptureModule": ["ModelConvStream", "ModelDetailCaptureModule", "ModelFusionBlock", "ModelHead", "Module", "ModuleList", "The", "ValueError", "__init__", "alphas", "append", "be", "class", "config", "conv_chans", "convstream", "convstream_hidden_sizes", "def", "detail_features", "detailed_feature_map_", "detailed_feature_map_name", "equal", "features", "for", "forward", "fusion_blocks", "fusion_channels", "fusion_hidden_sizes", "hidden_size", "i", "if", "in", "in_channels", "len", "length", "matting_head", "nn", "of", "out_channels", "pixel_values", "raise", "range", "return", "self", "should", "sigmoid", "str", "super", "the", "to", "torch"], "vitmatte/modeling_vitmatte.py:VitMatteForImageMatting": ["ImageMattingOutput", "ModelDetailCaptureModule", "ModelForImageMatting", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Tensor", "Training", "__init__", "alphas", "attentions", "auto_docstring", "backbone", "bool", "class", "config", "decoder", "def", "else", "feature_maps", "features", "forward", "forward_with_filtered_kwargs", "hidden_states", "if", "is", "labels", "load_backbone", "loss", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "self", "super", "supported", "torch", "use_return_dict", "yet"], "voxtral/modeling_voxtral.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "and", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "ndim", "nn", "not", "p", "query", "return", "scaling", "shape", "size", "softmax", "torch", "training", "transpose", "value", "view"], "voxtral/modeling_voxtral.py:VoxtralAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "_shape", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "caching", "call", "class", "config", "contiguous", "creating", "decoder", "def", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "k_proj", "key_states", "kwargs", "layer_head_mask", "layer_idx", "logger", "make", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "provide", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "seq_len", "size", "super", "sure", "tensor", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "used", "v_proj", "value_states", "view", "warning_once", "when", "will", "without"], "voxtral/modeling_voxtral.py:VoxtralEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "voxtral/modeling_voxtral.py:VoxtralPreTrainedModel": ["Conv1d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_cache_class", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "audio_config", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "voxtral/modeling_voxtral.py:VoxtralEncoder": ["AvgPool1d", "BaseModelOutput", "Conv1d", "Embedding", "False", "LayerNorm", "LongTensor", "Make", "ModelAttention", "ModelEncoder", "ModelEncoderConfig", "ModelEncoderLayer", "ModelPreTrainedModel", "Module", "ModuleList", "None", "Qwen2Audio", "TransformersKwargs", "Unpack", "ValueError", "_", "__init__", "_can_record_outputs", "_freeze_parameters", "_get_feat_extract_output_lengths", "_no_split_modules", "_requires_grad", "attention_mask", "attentions", "avg_pooler", "be", "but", "check_model_inputs", "class", "config", "conv1", "conv2", "d_model", "def", "device", "dropout", "dtype", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "encoder_layer", "encoder_layerdrop", "encoder_layers", "enumerate", "expected_seq_length", "expects", "f", "features", "for", "forward", "found", "functional", "gelu", "get_input_embeddings", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "input", "input_features", "input_lengths", "inputs_embeds", "kernel_size", "kwargs", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "length", "main_input_name", "math", "max_source_positions", "mel", "nn", "num_mel_bins", "of", "output_lengths", "p", "pad", "pad_token_id", "padding", "padding_idx", "param", "parameters", "permute", "post_init", "r", "raise", "range", "requires_grad", "requires_grad_", "return", "scale_embedding", "self", "set_input_embeddings", "shape", "sqrt", "stride", "super", "sure", "the", "to", "torch", "training", "value", "weight"], "voxtral/modeling_voxtral.py:VoxtralMultiModalProjector": ["ACT2FN", "False", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "audio_config", "audio_features", "bias", "class", "config", "def", "forward", "hidden_size", "hidden_states", "intermediate_size", "linear_1", "linear_2", "nn", "projector_hidden_act", "return", "self", "super", "text_config"], "voxtral/modeling_voxtral.py:VoxtralForConditionalGeneration": ["AutoModel", "AutoModelForCausalLM", "BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "FloatTensor", "FutureWarning", "GenerationMixin", "LongTensor", "ModelForConditionalGeneration", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Union", "Unpack", "__init__", "_keep_in_fp32_modules_strict", "_pp_plan", "_tied_weights_keys", "_tp_plan", "and", "args", "attention_mask", "audio_config", "audio_embeds", "audio_hidden_states", "audio_outputs", "audio_token_id", "audio_token_mask", "audio_tower", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "decoder", "def", "deprecated", "device", "embed_positions", "forward", "from_config", "get", "get_audio_embeds", "get_audio_features", "get_decoder", "get_input_embeddings", "get_output_embeddings", "hidden_states", "if", "input_features", "input_ids", "inputs_embeds", "instead", "int", "intermediate_size", "is", "kwargs", "labels", "language_model", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "masked_scatter", "method", "model_inputs", "multi_modal_projector", "new_embeddings", "not", "outputs", "past_key_values", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "reshape", "return", "self", "set_decoder", "set_input_embeddings", "set_output_embeddings", "super", "text_config", "to", "torch", "unsqueeze", "use", "use_cache", "value", "vocab_size", "warn", "warnings", "weight"], "deepseek_vl/modeling_deepseek_vl.py:DeepseekVLBaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "deepseek_vl/modeling_deepseek_vl.py:DeepseekVLCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "deepseek_vl/modeling_deepseek_vl.py:DeepseekVLAligner": ["GELU", "Linear", "ModelAligner", "Module", "Tensor", "__init__", "activation", "class", "config", "def", "forward", "hidden_size", "in_features", "linear1", "linear2", "nn", "out_features", "return", "self", "super", "text_config", "torch", "vision_config", "vision_encodings", "x"], "deepseek_vl/modeling_deepseek_vl.py:DeepseekVLPreTrainedModel": ["False", "Linear", "LlamaDecoderLayer", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "base_model_prefix", "bias", "causal_mask", "class", "config", "data", "def", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "past_key_values", "self", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "deepseek_vl/modeling_deepseek_vl.py:DeepseekVLModel": ["AutoModel", "Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelAligner", "ModelBaseModelOutputWithPast", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "aligner", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cache_position", "can_return_tuple", "cannot", "class", "config", "def", "device", "do", "dtype", "either", "else", "expand_as", "f", "features", "forward", "from_config", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "gradient_checkpointing", "hidden_states", "if", "image", "image_attention_mask", "image_embeds", "image_features", "image_hidden_states", "image_token_id", "input_ids", "inputs_embeds", "int", "is", "kwargs", "language_model", "last_hidden_state", "lm_output", "logits_to_keep", "long", "masked_scatter", "match", "must", "n_image_features", "n_image_tokens", "not", "numel", "one", "past_key_values", "pixel_values", "position_ids", "post_init", "raise", "reshape", "return", "same", "self", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "the", "time", "to", "tokens", "torch", "unsqueeze", "use_cache", "value", "vision_config", "vision_model"], "deepseek_vl/modeling_deepseek_vl.py:DeepseekVLForConditionalGeneration": ["AttributeError", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Not", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "else", "embed_tokens", "for", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "needed", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_embeddings_for_image_generation", "prepare_inputs_for_generation", "r", "raise", "return", "self", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "use_cache", "value", "vocab_size", "weight"], "marian/modeling_marian.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "marian/modeling_marian.py:MarianSinusoidalPositionalEmbedding": ["Embedding", "False", "FloatTensor", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_init_weight", "arange", "array", "bsz", "class", "cos", "def", "device", "dim", "dtype", "else", "embedding_dim", "empty", "for", "forward", "if", "in", "input_ids_shape", "int", "is", "j", "long", "n_pos", "nn", "no_grad", "np", "num_positions", "out", "padding_idx", "past_key_values_length", "pos", "position_enc", "position_ids", "power", "range", "requires_grad", "return", "self", "sentinel", "seq_len", "shape", "sin", "super", "torch", "weight"], "marian/modeling_marian.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "marian/modeling_marian.py:MarianAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "marian/modeling_marian.py:MarianEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "None", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "int", "isinf", "isnan", "layer_head_mask", "layer_idx", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "marian/modeling_marian.py:MarianDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "marian/modeling_marian.py:MarianPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weight", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "decoder_input_ids", "def", "device", "diagonal", "dim", "dtype", "dummy_inputs", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "ne", "nn", "normal_", "not", "npu", "ones", "pad_token", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "property", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tensor", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "marian/modeling_marian.py:MarianEncoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "_", "__init__", "_update_full_mask", "all_attentions", "and", "assert", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "marian/modeling_marian.py:MarianDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "assert", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "but", "cache_position", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "decoder_vocab_size", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "exactly", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "must", "nn", "not", "of", "one", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "warning_once", "will", "with", "zip"], "marian/modeling_marian.py:MarianModel": ["BaseModelOutput", "Cache", "Embedding", "FloatTensor", "In", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Seq2SeqModelOutput", "Tensor", "True", "Union", "ValueError", "__init__", "_get_resized_embeddings", "_tied_weights_keys", "and", "appropriate", "are", "attention_mask", "attentions", "auto_docstring", "be", "bool", "by", "cache_position", "called", "calling", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_embed_tokens", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_vocab_size", "deepcopy", "def", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_embed_tokens", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_decoder_input_embeddings", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input", "input_ids", "inputs_embeds", "instead", "int", "is", "isinstance", "last_hidden_state", "len", "meaning", "model_embeds", "new_embeddings", "new_num_tokens", "nn", "not", "old_embeddings", "order", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "raise", "resize_decoder_token_embeddings", "resize_token_embeddings", "return", "return_dict", "self", "set", "set_decoder_input_embeddings", "set_input_embeddings", "share_encoder_decoder_embeddings", "shared", "should", "simply", "super", "the", "tie_weights", "to", "torch", "tuple", "use", "use_cache", "use_return_dict", "value", "vocab_size", "weight", "with", "you"], "marian/modeling_marian.py:MarianMTModel": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelMTModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "ValueError", "__init__", "_dynamic_tied_weights_keys", "_get_resized_embeddings", "_get_resized_lm_head", "_keys_to_ignore_on_load_missing", "_keys_to_ignore_on_save", "_resize_final_logits_bias", "_resize_token_embeddings", "_tie_encoder_decoder_weights", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "args", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "be", "bias", "bool", "cache_position", "called", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "decoder_vocab_size", "def", "device", "dim", "else", "embed_positions", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "for", "forward", "get_decoder", "get_decoder_input_embeddings", "get_encoder", "get_input_embeddings", "get_output_embeddings", "getattr", "hasattr", "head_mask", "if", "in", "input_ids", "inputs_embeds", "instead", "int", "is", "is_encoder_decoder", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "model_embeds", "module", "modules", "new_bias", "new_embeddings", "new_lm_head", "new_num_tokens", "nn", "not", "old_embeddings", "old_lm_head", "old_num_tokens", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "raise", "register_buffer", "resize_decoder_token_embeddings", "resize_token_embeddings", "return", "return_dict", "self", "set_decoder_input_embeddings", "set_input_embeddings", "set_output_embeddings", "shape", "share_encoder_decoder_embeddings", "shift_tokens_right", "should", "since", "super", "target_vocab_size", "tie_encoder_decoder", "tie_weights", "tie_word_embeddings", "tied_weights", "to", "torch", "tuple", "use", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "word_embeddings", "zeros"], "marian/modeling_marian.py:MarianDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "marian/modeling_marian.py:MarianForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "olmoe/modeling_olmoe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "olmoe/modeling_olmoe.py:OlmoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "olmoe/modeling_olmoe.py:OlmoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "olmoe/modeling_olmoe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "olmoe/modeling_olmoe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "olmoe/modeling_olmoe.py:OlmoeMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "olmoe/modeling_olmoe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "olmoe/modeling_olmoe.py:OlmoeAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "clamp_", "class", "clip_qkv", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "eps", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "lead", "logger", "make", "math", "matmul", "max", "max_position_embeddings", "min", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_norm", "q_proj", "query_states", "raise", "recommended", "repeat_kv", "reshape", "return", "rms_norm_eps", "rope_theta", "self", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "olmoe/modeling_olmoe.py:OlmoeFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "The", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "apply_rotary_pos_emb", "args", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache_kwargs", "cache_position", "cast", "casted", "clamp_", "class", "clip_qkv", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "max", "might", "min", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "o_proj", "or", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_norm", "q_proj", "query_states", "related", "reshape", "return", "seems", "self", "silently", "sin", "size", "states", "super", "target_dtype", "the", "this", "to", "torch", "training", "transpose", "tuple", "type", "upcasted", "update", "use_cache", "use_top_left_mask", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "you"], "olmoe/modeling_olmoe.py:OlmoeSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "clamp_", "class", "clip_qkv", "config", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_size", "hidden_states", "if", "implementation", "is", "is_causal", "k_norm", "k_proj", "key_states", "layer_idx", "loading", "logger", "manual", "max", "min", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_norm", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "olmoe/modeling_olmoe.py:OlmoeSparseMoeBlock": ["F", "False", "Linear", "ModelMLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "keepdim", "nn", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "olmoe/modeling_olmoe.py:OlmoeDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelRMSNorm", "ModelSparseMoeBlock", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "olmoe/modeling_olmoe.py:OlmoePreTrainedModel": ["Embedding", "False", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "olmoe/modeling_olmoe.py:OlmoeModel": ["AttentionMaskConverter", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "list", "logger", "mask_length", "masked_fill", "min", "min_dtype", "must", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "return_dict", "rms_norm_eps", "rotary_emb", "router_logits", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_static_cache", "v", "vocab_size", "warning_once", "with", "xpu"], "olmoe/modeling_olmoe.py:OlmoeForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "class", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "mimi/modeling_mimi.py:MimiOutput": ["Cache", "FloatTensor", "LongTensor", "ModelOutput", "None", "Optional", "Union", "audio_codes", "audio_values", "class", "decoder_past_key_values", "encoder_past_key_values", "list", "r", "torch"], "mimi/modeling_mimi.py:MimiConv1dPaddingCache": ["Expected", "ModelConv1dPaddingCache", "None", "NotImplementedError", "Tensor", "True", "ValueError", "__init__", "all", "and", "batch_size", "class", "constant", "convolutions", "current_cache", "def", "device", "dtype", "elif", "else", "empty", "f", "for", "from_args_num_layers", "hidden_states", "if", "in", "in_channels", "int", "is", "layer_idx", "len", "list", "mode", "not", "num_layers", "ones", "or", "other", "padding", "padding_cache", "padding_mode", "padding_states", "per_layer_in_channels", "per_layer_is_init", "per_layer_padding", "per_layer_padding_mode", "pop", "raise", "replicate", "return", "self", "shape", "str", "supported", "than", "torch", "update", "using", "values", "zeros"], "mimi/modeling_mimi.py:MimiEncoderOutput": ["Cache", "FloatTensor", "LongTensor", "ModelConv1dPaddingCache", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "Union", "audio_codes", "class", "encoder_past_key_values", "list", "padding_cache", "r", "torch"], "mimi/modeling_mimi.py:MimiDecoderOutput": ["Cache", "FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "Union", "audio_values", "class", "decoder_past_key_values", "list", "r", "torch"], "mimi/modeling_mimi.py:MimiConv1d": ["Conv1d", "False", "LongTensor", "ModelConv1d", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_get_extra_padding_for_conv1d", "_get_output_length", "_pad1d", "and", "apply_weight_norm", "been", "bias", "bool", "cat", "causal", "ceil", "class", "config", "conv", "convolutions", "def", "dilation", "dim", "dtype", "elif", "else", "end", "extra_pad", "extra_padding", "f", "float", "for", "forward", "functional", "groups", "has", "hasattr", "hidden_states", "ideal_length", "if", "in_channels", "initialized", "input_length", "int", "int64", "is", "kernel_size", "layer_idx", "layer_padding_cache", "length", "logger", "max", "max_pad", "mode", "n_frames", "nn", "non", "not", "out_channels", "output_length", "pad", "pad_mode", "padded", "padding", "padding_cache", "padding_left", "padding_right", "padding_total", "paddings", "parametrizations", "persistent", "raise", "reflect", "register_buffer", "remove_weight_norm", "return", "self", "shape", "staticmethod", "str", "stride", "super", "supported", "tensor", "to", "torch", "tuple", "update", "use_causal_conv", "utils", "value", "warning", "weight_norm", "with", "zero"], "mimi/modeling_mimi.py:MimiConvTranspose1d": ["ConvTranspose1d", "ModelConvTranspose1d", "Module", "True", "ValueError", "__init__", "apply_weight_norm", "bias", "causal", "ceil", "class", "config", "conv", "convolutions", "def", "else", "end", "for", "forward", "groups", "hasattr", "hidden_states", "if", "in_channels", "int", "kernel_size", "makes", "math", "nn", "not", "only", "or", "out_channels", "padding_left", "padding_right", "padding_total", "parametrizations", "raise", "remove_weight_norm", "return", "self", "sense", "shape", "stride", "super", "trim_right_ratio", "use_causal_conv", "utils", "weight_norm"], "mimi/modeling_mimi.py:MimiResnetBlock": ["ELU", "Identity", "ModelConfig", "ModelConv1d", "ModelResnetBlock", "Module", "ModuleList", "None", "Number", "ValueError", "__init__", "block", "class", "compress", "config", "def", "dilation", "dilations", "dim", "else", "enumerate", "for", "forward", "hidden", "hidden_states", "i", "if", "in", "in_chs", "int", "isinstance", "kernel", "kernel_size", "kernel_sizes", "layer", "len", "list", "match", "nn", "number", "of", "out_chs", "padding_cache", "raise", "residual", "residual_kernel_size", "return", "self", "shortcut", "should", "sizes", "super", "use_conv_shortcut", "zip"], "mimi/modeling_mimi.py:MimiEncoder": ["ELU", "ModelConfig", "ModelConv1d", "ModelEncoder", "ModelResnetBlock", "Modelconv1d_layer_names", "Module", "ModuleList", "None", "_Modelconv1d_layer_names", "__init__", "append", "audio_channels", "block", "class", "config", "conv_layer", "current_scale", "def", "dilation_growth_rate", "else", "enumerate", "extend", "f", "for", "forward", "get_submodule", "hidden_size", "hidden_states", "if", "in", "isinstance", "j", "kernel_size", "last_kernel_size", "layer", "layer_idx", "layername", "layers", "len", "model", "nn", "num_filters", "num_residual_layers", "padding_cache", "range", "ratio", "return", "reversed", "scaling", "self", "setattr", "stride", "super", "upsampling_ratios"], "mimi/modeling_mimi.py:MimiLayerScale": ["ModelLayerScale", "Module", "Parameter", "Tensor", "True", "__init__", "channels", "class", "config", "def", "forward", "full", "hidden_size", "initial_scale", "layer_scale_initial_scale", "nn", "requires_grad", "return", "scale", "self", "super", "torch", "x"], "mimi/modeling_mimi.py:MimiRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "mimi/modeling_mimi.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "mimi/modeling_mimi.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "mimi/modeling_mimi.py:MimiMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "bias", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "mimi/modeling_mimi.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "mimi/modeling_mimi.py:MimiAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "lead", "logger", "make", "math", "matmul", "max_position_embeddings", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "repeat_kv", "return", "rope_theta", "rotary_emb", "scaling", "self", "shape", "should", "sin", "size", "sliding_window", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "mimi/modeling_mimi.py:MimiFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "StaticCache", "Tensor", "The", "ValueError", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "an", "and", "apply_rotary_pos_emb", "args", "at", "attention_dropout", "attention_mask", "attn_implementation", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache", "cache_kwargs", "cache_position", "cast", "casted", "class", "com", "compatible", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attention_2", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "github", "hasattr", "have", "head_dim", "hidden", "hidden_states", "https", "huggingface", "if", "implementation", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "isinstance", "issue", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "make", "mean", "might", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "o_proj", "open", "or", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "raise", "related", "reshape", "return", "rotary_emb", "sdpa", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "static", "super", "sure", "target_dtype", "the", "this", "time", "to", "torch", "training", "transformers", "transpose", "tuple", "type", "upcasted", "update", "use", "use_cache", "use_top_left_mask", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "with", "you"], "mimi/modeling_mimi.py:MimiSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "rotary_emb", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "mimi/modeling_mimi.py:MimiTransformerLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelConfig", "ModelLayerScale", "ModelMLP", "ModelTransformerLayer", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "mlp_layer_scale", "new_name", "nn", "norm_eps", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "self_attn_layer_scale", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "mimi/modeling_mimi.py:MimiTransformerModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelTransformerLayer", "ModelTransformerModel", "Module", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "__init__", "_attn_implementation", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attentions", "bool", "cache_position", "causal_mask", "checkpointing", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "for", "forward", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_states", "if", "in", "incompatible", "input_embeds", "is", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "list", "logger", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "past_key_values", "past_seen_tokens", "position_ids", "range", "return", "return_dict", "self", "shape", "super", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "warning_once", "with"], "mimi/modeling_mimi.py:MimiDecoder": ["ELU", "ModelConfig", "ModelConv1d", "ModelConvTranspose1d", "ModelDecoder", "ModelResnetBlock", "Module", "ModuleList", "__init__", "audio_channels", "class", "config", "current_scale", "def", "dilation_growth_rate", "for", "forward", "hidden_size", "hidden_states", "in", "int", "j", "kernel_size", "last_kernel_size", "layer", "layers", "len", "model", "nn", "num_filters", "num_residual_layers", "range", "ratio", "return", "scaling", "self", "stride", "super", "upsampling_ratios"], "mimi/modeling_mimi.py:MimiEuclideanCodebook": ["ModelConfig", "ModelEuclideanCodebook", "Module", "None", "Tensor", "True", "__init__", "_embed", "argmin", "cdist", "clamp", "class", "cluster_usage", "codebook_dim", "codebook_size", "config", "decode", "def", "dim", "dists", "dtype", "embed", "embed_ind", "embed_sum", "embedding", "encode", "epsilon", "float", "float32", "functional", "hidden_states", "if", "initialized", "is", "min", "nn", "ones", "p", "property", "quantize", "register_buffer", "reshape", "return", "self", "shape", "super", "tensor", "torch", "view", "zeros"], "mimi/modeling_mimi.py:MimiVectorQuantization": ["ModelConfig", "ModelEuclideanCodebook", "ModelVectorQuantization", "Module", "__init__", "class", "codebook", "config", "decode", "def", "embed_in", "embed_ind", "encode", "hidden_states", "nn", "permute", "quantize", "return", "self", "super"], "mimi/modeling_mimi.py:MimiResidualVectorQuantizer": ["Conv1d", "False", "ModelConfig", "ModelResidualVectorQuantizer", "ModelVectorQuantization", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_indices", "append", "bias", "class", "codebook_size", "codes", "config", "decode", "def", "device", "else", "embeddings", "encode", "enumerate", "for", "frame_rate", "hidden_size", "i", "if", "in", "indices", "input_proj", "int", "is", "layer", "layers", "nn", "not", "num_quantizers", "out_indices", "output_proj", "quantized", "quantized_out", "range", "residual", "return", "self", "stack", "super", "tensor", "torch", "transpose", "vector_quantization_hidden_dimension"], "mimi/modeling_mimi.py:MimiSplitResidualVectorQuantizer": ["ModelConfig", "ModelResidualVectorQuantizer", "ModelSplitResidualVectorQuantizer", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "acoustic_codes", "acoustic_residual_vector_quantizer", "asked", "be", "but", "cat", "class", "codebook_size", "codebooks", "codes", "config", "currently", "decode", "def", "dim", "e", "else", "embeddings", "encode", "f", "float", "frame_rate", "higher", "i", "if", "is", "lower", "max_num_quantizers", "nn", "num_acoustic_quantizers", "num_quantizers", "num_semantic_quantizers", "number", "of", "quantized_out", "quantizers", "raise", "return", "self", "semantic", "semantic_residual_vector_quantizer", "shape", "should", "super", "than", "the", "torch", "total"], "mimi/modeling_mimi.py:MimiPreTrainedModel": ["Conv1d", "ConvTranspose1d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoderLayer", "ModelLayerScale", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "a", "b", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "groups", "if", "in_channels", "init", "initializer_range", "input_values", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "layer_scale_initial_scale", "main_input_name", "math", "mean", "module", "nn", "normal_", "not", "past_key_values", "scale", "self", "sqrt", "std", "supports_gradient_checkpointing", "uniform_", "weight", "zero_"], "mimi/modeling_mimi.py:MimiModel": ["Cache", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelConv1d", "ModelConv1dPaddingCache", "ModelConvTranspose1d", "ModelDecoder", "ModelDecoderOutput", "ModelEncoder", "ModelEncoderOutput", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelSplitResidualVectorQuantizer", "ModelTransformerModel", "None", "Number", "Optional", "Tensor", "The", "Union", "ValueError", "_", "_Modelconv1d_layer_names", "__init__", "_decode_frame", "_encode_frame", "_get_output_length", "a", "and", "append", "arange", "asked", "audio", "audio_codes", "audio_codes_mask", "audio_values", "auto_docstring", "be", "bias", "bits_per_codebook", "bool", "but", "channels", "class", "codebook_size", "codebooks", "codes", "config", "currently", "decode", "decoder", "decoder_outputs", "decoder_past_key_values", "decoder_transformer", "def", "device", "dim", "dims", "downsample", "e", "elif", "else", "embeddings", "encode", "encodec_frame_rate", "encoded_frames", "encoded_lengths", "encoder", "encoder_outputs", "encoder_past_key_values", "encoder_transformer", "expand", "f", "flip", "float", "for", "forward", "frame_rate", "get", "get_audio_codes_mask", "get_encoded_length", "get_encoder", "get_submodule", "got", "groups", "hidden_size", "i", "if", "in", "in_channels", "input_length", "input_values", "int", "is", "kernel_size", "layer_idx", "layer_name", "len", "list", "log2", "lower", "math", "max", "must", "not", "num_layers", "num_quantizers", "number", "of", "ones_like", "or", "output_length", "outputs", "pad_mode", "padding_cache", "padding_mask", "padding_side", "padding_total", "past_key_values", "per_layer_in_channels", "per_layer_padding", "per_layer_padding_mode", "post_init", "power", "quantizer", "quantizers", "r", "raise", "replicate", "return", "return_dict", "right", "self", "shape", "should", "str", "stride", "sum", "super", "than", "the", "to", "torch", "total", "transpose", "tuple", "unsqueeze", "upsample", "upsample_groups", "use_streaming"], "altclip/modeling_altclip.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "altclip/modeling_altclip.py:clip_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "altclip/modeling_altclip.py:AltCLIPOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "altclip/modeling_altclip.py:AltRobertaEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelRobertaEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "altclip/modeling_altclip.py:AltRobertaSelfAttention": ["Dropout", "Embedding", "False", "FloatTensor", "Linear", "ModelRobertaSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bhld", "bhlr", "bhrd", "bool", "class", "config", "context_layer", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "f", "forward", "functional", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_layer", "key_length", "long", "lrd", "math", "matmul", "max_position_embeddings", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "query_length", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "shape", "size", "softmax", "sqrt", "super", "the", "to", "torch", "transpose", "tuple", "value", "value_layer", "view"], "altclip/modeling_altclip.py:AltRobertaSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelRobertaSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "altclip/modeling_altclip.py:AltRobertaAttention": ["False", "FloatTensor", "ModelRobertaAttention", "ModelRobertaSelfOutput", "Model_ROBERTA_SELF_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "altclip/modeling_altclip.py:AltRobertaIntermediate": ["ACT2FN", "Linear", "ModelRobertaIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "altclip/modeling_altclip.py:AltRobertaOutput": ["Dropout", "LayerNorm", "Linear", "ModelRobertaOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "altclip/modeling_altclip.py:AltRobertaLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelRobertaAttention", "ModelRobertaIntermediate", "ModelRobertaLayer", "ModelRobertaOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "altclip/modeling_altclip.py:AltRobertaEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelRobertaEncoder", "ModelRobertaLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "altclip/modeling_altclip.py:AltRobertaPooler": ["Linear", "ModelRobertaPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "altclip/modeling_altclip.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "altclip/modeling_altclip.py:AltCLIPAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "altclip/modeling_altclip.py:AltCLIPMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "altclip/modeling_altclip.py:AltCLIPEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "altclip/modeling_altclip.py:AltCLIPEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "altclip/modeling_altclip.py:AltCLIPVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "altclip/modeling_altclip.py:AltCLIPPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "ModelModel", "ModelPreTrainedModel", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "_is_hf_initialized", "_no_split_module", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "padding_idx", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "v_proj", "vision_embed_dim", "visual_projection", "weight", "zero_"], "altclip/modeling_altclip.py:AltCLIPVisionTransformer": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layrnorm", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "altclip/modeling_altclip.py:AltCLIPVisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embeddings", "forward", "get_input_embeddings", "if", "interpolate_pos_encoding", "is", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "vision_model"], "altclip/modeling_altclip.py:AltRobertaModel": ["BaseModelOutputWithPooling", "BaseModelOutputWithPoolingAndCrossAttentions", "ModelPreTrainedModel", "ModelRobertaEmbeddings", "ModelRobertaEncoder", "ModelRobertaModel", "ModelRobertaPooler", "ModelTextConfig", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "altclip/modeling_altclip.py:AltCLIPTextModel": ["BaseModelOutputWithPoolingAndProjection", "Embedding", "False", "LayerNorm", "Linear", "ModelPreTrainedModel", "ModelRobertaModel", "ModelTextConfig", "ModelTextModel", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "else", "embeddings", "eps", "forward", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "last_hidden_state", "layer_norm_eps", "new_num_tokens", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "pooler_output", "position_ids", "post_init", "pre_LN", "project_dim", "projection_state", "r", "resize_token_embeddings", "return", "return_dict", "roberta", "self", "sequence_output", "set_input_embeddings", "super", "token_type_ids", "torch", "transformation", "tuple", "use_return_dict", "value", "word_embeddings"], "altclip/modeling_altclip.py:AltCLIPModel": ["False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelVisionConfig", "ModelVisionTransformer", "None", "Optional", "Parameter", "T", "Tensor", "True", "TypeError", "Union", "__init__", "_attn_implementation", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "class", "clip_loss", "config", "def", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "output", "output_attentions", "output_hidden_states", "p", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "project_dim", "projection_dim", "r", "raise", "return", "return_dict", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "token_type_ids", "torch", "tuple", "type", "use_return_dict", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "True", "__init__", "act_fn", "bias", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "linear_fc1", "linear_fc2", "nn", "return", "self", "super"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionPatchEmbed": ["Conv3d", "ModelVisionPatchEmbed", "Module", "None", "Tensor", "True", "__init__", "bias", "class", "config", "def", "dtype", "embed_dim", "forward", "hidden_size", "hidden_states", "in_channels", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionRotaryEmbedding": ["False", "ModelVisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionPatchMerger": ["False", "GELU", "LayerNorm", "Linear", "ModelVisionConfig", "ModelVisionPatchMerger", "Module", "None", "Tensor", "__init__", "act_fn", "class", "config", "def", "else", "eps", "forward", "hidden_size", "if", "linear_fc1", "linear_fc2", "nn", "norm", "out_hidden_size", "return", "self", "spatial_merge_size", "super", "torch", "use_postshuffle_norm", "view", "x"], "qwen3_vl/modeling_qwen3_vl.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen3_vl/modeling_qwen3_vl.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "qwen3_vl/modeling_qwen3_vl.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen3_vl/modeling_qwen3_vl.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionBlock": ["GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionBlock", "ModelVisionMLP", "None", "Optional", "Tensor", "__init__", "attn", "attn_implementation", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "nn", "norm1", "norm2", "position_embeddings", "return", "rotary_pos_emb", "sdpa", "self", "str", "super", "torch", "tuple"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "apply_interleaved_mrope", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "enumerate", "expand", "float", "for", "forward", "freqs", "freqs_t", "get", "hasattr", "idx", "if", "in", "inv_freq", "inv_freq_expanded", "is", "isinstance", "length", "max_position_embeddings", "max_seq_len_cached", "mps", "mrope_section", "ndim", "nn", "no_grad", "not", "offset", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "slice", "start", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLTextRMSNorm": ["ModelTextRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_vl/modeling_qwen3_vl.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLTextMLP": ["ACT2FN", "False", "Linear", "ModelTextMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLTextDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextMLP", "ModelTextRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLModelOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLPreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextDecoderLayer", "ModelVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLVisionModel": ["Embedding", "F", "False", "ModelPreTrainedModel", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionModel", "ModelVisionPatchEmbed", "ModelVisionPatchMerger", "ModelVisionRotaryEmbedding", "ModuleList", "None", "T", "Tensor", "True", "_", "__init__", "_no_split_modules", "append", "arange", "base_h", "base_h_ceil", "blk", "block_cols", "block_rows", "blocks", "cat", "class", "clip", "col_idx", "config", "coords", "cos", "cu_seqlens", "cumsum", "deepstack_feature", "deepstack_feature_lists", "deepstack_merger_list", "deepstack_visual_indexes", "def", "depth", "device", "dh", "dim", "dtype", "dw", "else", "emb", "embeddings", "empty", "enumerate", "expand", "extend", "fast_pos_embed_interpolate", "flatten", "for", "forward", "freq_table", "gradient_checkpointing", "grid_hs", "grid_thw", "grid_ts", "grid_ws", "h", "h_idxs", "h_idxs_ceil", "h_idxs_floor", "head_dim", "height", "hidden_size", "hidden_states", "i", "idx_list", "idx_tensor", "if", "in", "index", "indices", "inputs", "int", "int32", "intra_col", "intra_row", "is_tracing", "item", "jit", "kwargs", "layer_num", "len", "linspace", "long", "max", "max_hw", "merge_size", "merged_h", "merged_w", "merger", "nn", "num_frames", "num_grid_per_side", "num_heads", "num_position_embeddings", "num_tokens", "offset", "pad", "patch_embed", "patch_pos_embeds", "patch_pos_embeds_permute", "patch_size", "permute", "pos_embed", "pos_embeds", "pos_ids", "position_embeddings", "prod", "range", "repeat", "repeat_interleave", "reshape", "return", "rot_pos_emb", "rotary_pos_emb", "row_idx", "self", "seq_len", "shape", "sin", "size", "spatial_merge_size", "spatial_merge_unit", "split", "stack", "sum", "super", "t", "tensor", "tolist", "torch", "total_tokens", "use_postshuffle_norm", "value", "view", "w", "w_idxs", "w_idxs_ceil", "w_idxs_floor", "weight", "weight_list", "weight_tensor", "weights", "width", "zip"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextRMSNorm", "ModelTextRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "ValueError", "You", "__init__", "_deepstack_process", "_no_split_modules", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "check_model_inputs", "class", "clone", "config", "create_causal_mask", "decoder_layer", "deepstack_visual_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "enumerate", "eps", "exactly", "expand", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "len", "list", "local_this", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "text_position_ids", "to", "torch", "tuple", "use_cache", "view", "visual_embeds", "visual_pos_masks", "vocab_size"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLModel": ["Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelTextDecoderLayer", "ModelTextModel", "ModelVisionBlock", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "_from_config", "_no_split_modules", "accepts_loss_kwargs", "add", "all", "and", "append", "arange", "argwhere", "attention_mask", "attention_mask_tensor", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "cache_position", "can_return_tuple", "cat", "class", "config", "cumsum", "decoder", "deepstack_image_embeds", "deepstack_video_embeds", "deepstack_visual_embeds", "def", "delta", "device", "diagonal", "dict", "dim", "dim1", "dim2", "do", "dtype", "ed", "ed_image", "ed_video", "elif", "else", "embed_joint", "enumerate", "exactly", "expand", "expand_as", "f", "features", "finfo", "flatten", "for", "forward", "full_attention", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_seq_length", "get_video_features", "grid_thw", "h", "h_index", "hidden_states", "i", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_index", "image_mask", "image_mask_joint", "image_nums", "image_token_id", "img_embed", "in", "index", "input_ids", "input_tokens", "inputs_embeds", "int", "is", "is_floating_point", "is_torchdynamo_compiling", "isinstance", "item", "keepdim", "kwargs", "language_model", "last_hidden_state", "len", "list", "llm_grid_h", "llm_grid_t", "llm_grid_w", "llm_pos_ids_list", "llm_positions", "long", "masked_fill_", "masked_scatter", "match", "max", "max_position_ids", "min", "mrope_position_deltas", "must", "n_image_tokens", "n_video_tokens", "ndim", "new_zeros", "not", "numel", "of", "one", "ones", "ones_like", "or", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prefill_compiled_stage", "prefill_noncompiled_stage", "prod", "r", "raise", "range", "remain_images", "remain_videos", "repeat_interleave", "reshape", "return", "rope_deltas", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "special_video_mask", "specify", "split", "split_sizes", "squeeze", "st", "st_idx", "stack", "sum", "super", "t", "t_index", "tensor", "text_config", "text_len", "to", "tokens", "tolist", "torch", "total_input_ids", "tuple", "type", "unsqueeze", "value", "vid_embed", "video", "video_embeds", "video_features", "video_grid_thw", "video_index", "video_mask", "video_mask_joint", "video_nums", "video_token_id", "view", "vision_config", "vision_start_indices", "vision_start_token_id", "vision_tokens", "visual", "visual_pos_masks", "w", "w_index", "zeros", "zip"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen3_vl/modeling_qwen3_vl.py:Qwen3VLForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "If", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_expand_dict_for_generation", "_expand_dict_for_generation_visual", "_expand_inputs_for_generation", "_get_image_nums_and_video_nums", "_repeat_interleave_samples", "_tied_weights_keys", "accepts_loss_kwargs", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "defined", "device", "dict", "dict_to_expand", "dim", "dims", "dtype", "elif", "else", "encoder_outputs", "expand_size", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "get_video_features", "hidden_size", "hidden_states", "if", "image_grid_thw", "image_mask", "image_nums", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "isinstance", "key", "kwargs", "labels", "language_model", "lengths", "list", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "make", "model", "model_inputs", "model_kwargs", "nn", "not", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "prod", "property", "r", "raise", "repeat", "repeat_args", "repeat_interleave", "repeat_times", "result", "return", "roll", "rope_deltas", "sample", "samples", "second_per_grid_ts", "self", "set_decoder", "set_input_embeddings", "shifts", "slice", "slice_indices", "split", "str", "sum", "super", "sure", "tensor", "text_config", "that", "torch", "tuple", "use_cache", "value", "video_grid_thw", "video_mask", "video_nums", "video_token_id", "vision_first_mask", "vision_start_mask", "vision_start_token_id", "visual", "visual_keys", "vocab_size", "weight", "x"], "glpn/modeling_glpn.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "glpn/modeling_glpn.py:GLPNDropPath": ["Model", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "glpn/modeling_glpn.py:GLPNOverlapPatchEmbeddings": ["Conv2d", "LayerNorm", "Model", "Module", "_", "__init__", "class", "def", "embeddings", "flatten", "forward", "height", "hidden_size", "kernel_size", "layer_norm", "nn", "num_channels", "padding", "patch_size", "pixel_values", "proj", "return", "self", "shape", "stride", "super", "transpose", "width"], "glpn/modeling_glpn.py:GLPNEfficientSelfAttention": ["Conv2d", "Dropout", "False", "LayerNorm", "Linear", "Model", "Module", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "f", "forward", "functional", "heads", "height", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "kernel_size", "key", "key_layer", "layer_norm", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_channels", "number", "of", "output_attentions", "outputs", "permute", "query", "query_layer", "raise", "reshape", "return", "self", "seq_len", "seq_length", "sequence_reduction_ratio", "shape", "size", "softmax", "sqrt", "sr", "sr_ratio", "stride", "super", "the", "torch", "transpose", "value", "value_layer", "view", "width"], "glpn/modeling_glpn.py:GLPNSelfOutput": ["Dropout", "Linear", "Model", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super"], "glpn/modeling_glpn.py:GLPNAttention": ["False", "Model", "ModelEfficientSelfAttention", "ModelSelfOutput", "Module", "__init__", "all_head_size", "attention_head_size", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "height", "hidden_size", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "sequence_reduction_ratio", "set", "super", "union", "value", "width"], "glpn/modeling_glpn.py:GLPNDWConv": ["Conv2d", "Model", "Module", "True", "__init__", "batch_size", "bias", "class", "def", "dim", "dwconv", "flatten", "forward", "groups", "height", "hidden_states", "nn", "num_channels", "return", "self", "seq_len", "shape", "super", "transpose", "view", "width"], "glpn/modeling_glpn.py:GLPNMixFFN": ["ACT2FN", "Dropout", "Linear", "Model", "ModelDWConv", "Module", "None", "__init__", "class", "config", "def", "dense1", "dense2", "dropout", "dwconv", "else", "forward", "height", "hidden_act", "hidden_dropout_prob", "hidden_features", "hidden_states", "if", "in_features", "intermediate_act_fn", "isinstance", "nn", "or", "out_features", "return", "self", "str", "super", "width"], "glpn/modeling_glpn.py:GLPNLayer": ["False", "Identity", "LayerNorm", "Model", "ModelAttention", "ModelDropPath", "ModelMixFFN", "Module", "__init__", "attention", "attention_output", "class", "config", "def", "drop_path", "else", "forward", "height", "hidden_features", "hidden_size", "hidden_states", "if", "in_features", "int", "layer_norm_1", "layer_norm_2", "layer_output", "mlp", "mlp_hidden_size", "mlp_output", "mlp_ratio", "nn", "num_attention_heads", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "sequence_reduction_ratio", "super", "width"], "glpn/modeling_glpn.py:GLPNEncoder": ["BaseModelOutput", "False", "LayerNorm", "Model", "ModelLayer", "ModelOverlapPatchEmbeddings", "Module", "ModuleList", "None", "True", "__init__", "all_hidden_states", "all_self_attentions", "append", "attentions", "batch_size", "blk", "block", "block_layer", "blocks", "class", "config", "contiguous", "cpu", "cur", "def", "depths", "device", "dpr", "drop_path", "drop_path_rate", "else", "embedding_layer", "embeddings", "enumerate", "for", "forward", "height", "hidden_size", "hidden_sizes", "hidden_states", "i", "idx", "if", "in", "is", "item", "j", "last_hidden_state", "layer_norm", "layer_outputs", "layers", "linspace", "mlp_ratio", "mlp_ratios", "nn", "norm_layer", "not", "num_attention_heads", "num_channels", "num_encoder_blocks", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_size", "patch_sizes", "permute", "pixel_values", "range", "reshape", "return", "return_dict", "self", "sequence_reduction_ratio", "shape", "sr_ratios", "stride", "strides", "sum", "super", "torch", "tuple", "v", "width", "x", "zip"], "glpn/modeling_glpn.py:GLPNPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "self", "std", "weight", "zero_"], "glpn/modeling_glpn.py:GLPNModel": ["BaseModelOutput", "FloatTensor", "Model", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_prune_heads", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "encoder_outputs", "for", "forward", "heads", "heads_to_prune", "hidden_states", "if", "in", "is", "items", "last_hidden_state", "layer", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "prune_heads", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "glpn/modeling_glpn.py:GLPNSelectiveFeatureFusion": ["BatchNorm2d", "Conv2d", "Model", "Module", "ReLU", "Sequential", "Sigmoid", "__init__", "attn", "cat", "class", "convolutional_layer1", "convolutional_layer2", "convolutional_layer3", "def", "dim", "features", "forward", "global_features", "hybrid_features", "in_channel", "in_channels", "int", "kernel_size", "local_features", "nn", "out_channels", "padding", "return", "self", "sigmoid", "stride", "super", "torch", "unsqueeze"], "glpn/modeling_glpn.py:GLPNDecoderStage": ["Conv2d", "False", "Identity", "Model", "ModelSelectiveFeatureFusion", "Module", "None", "Upsample", "__init__", "align_corners", "bilinear", "class", "convolution", "def", "else", "forward", "fusion", "hidden_state", "if", "in_channels", "is", "kernel_size", "mode", "nn", "not", "out_channels", "residual", "return", "scale_factor", "self", "should_skip", "super", "upsample"], "glpn/modeling_glpn.py:GLPNDecoder": ["False", "Model", "ModelStage", "Module", "ModuleList", "None", "Tensor", "Upsample", "__init__", "align_corners", "append", "bilinear", "class", "config", "decoder_hidden_size", "def", "final_upsample", "for", "forward", "fusion", "hidden_size", "hidden_sizes", "hidden_state", "hidden_states", "in", "list", "mode", "nn", "out_channels", "reserved_hidden_sizes", "return", "scale_factor", "self", "stage", "stage_hidden_state", "stage_hidden_states", "stages", "super", "torch", "zip"], "glpn/modeling_glpn.py:SiLogLoss": ["Module", "SiLogLoss", "__init__", "class", "def", "detach", "diff_log", "forward", "lambd", "log", "loss", "mean", "nn", "pow", "pred", "r", "return", "self", "sqrt", "super", "target", "torch", "valid_mask"], "glpn/modeling_glpn.py:GLPNDepthEstimationHead": ["Conv2d", "False", "Model", "Module", "ReLU", "Sequential", "Tensor", "__init__", "channels", "class", "config", "decoder_hidden_size", "def", "dim", "forward", "head", "head_in_index", "hidden_states", "inplace", "kernel_size", "list", "max_depth", "nn", "padding", "predicted_depth", "return", "self", "sigmoid", "squeeze", "stride", "super", "torch"], "glpn/modeling_glpn.py:GLPNForDepthEstimation": ["DepthEstimatorOutput", "FloatTensor", "Model", "ModelDecoder", "ModelDepthEstimationHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SiLogLoss", "Tensor", "True", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "decoder", "def", "else", "forward", "head", "hidden_states", "if", "is", "labels", "loss", "loss_fct", "not", "out", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "predicted_depth", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "superglue/modeling_superglue.py:concat_pairs": ["Model_pairs", "Tensor", "cat", "def", "for", "in", "return", "tensor0", "tensor1", "tensor_tuple0", "tensor_tuple1", "torch", "tuple", "zip"], "superglue/modeling_superglue.py:normalize_keypoints": ["Model_keypoints", "None", "Tensor", "True", "center", "def", "device", "dtype", "height", "int", "keepdim", "keypoints", "max", "return", "scaling", "size", "tensor", "torch", "values", "width"], "superglue/modeling_superglue.py:log_sinkhorn_iterations": ["Model_cost_matrix", "Model_sinkhorn_iterations", "Model_source_distribution", "Model_target_distribution", "Model_u_scaling", "Model_v_scaling", "Modelsumexp", "Tensor", "_", "def", "dim", "for", "in", "int", "num_iterations", "range", "return", "torch", "unsqueeze", "zeros_like"], "superglue/modeling_superglue.py:log_optimal_transport": ["Model", "Model_normalization", "Model_optimal_transport", "Model_optimal_transport_matrix", "Model_sinkhorn_iterations", "Model_source_distribution", "Model_target_distribution", "None", "Tensor", "batch_size", "cat", "couplings", "def", "expand", "int", "iterations", "new_tensor", "num_columns", "num_columns_tensor", "num_iterations", "num_rows", "num_rows_tensor", "one_tensor", "reg_param", "return", "scores", "shape", "source_reg_param", "target_reg_param", "to", "torch"], "superglue/modeling_superglue.py:arange_like": ["Model_like", "Tensor", "cumsum", "def", "dim", "int", "new_ones", "return", "shape", "torch", "x"], "superglue/modeling_superglue.py:KeypointMatchingOutput": ["FloatTensor", "IntTensor", "ModelMatchingOutput", "ModelOutput", "Models", "None", "Optional", "attentions", "class", "hidden_states", "loss", "mask", "matches", "matching_scores", "r", "torch", "tuple"], "superglue/modeling_superglue.py:SuperGlueMultiLayerPerceptron": ["BatchNorm1d", "Linear", "Model", "ModelConfig", "ModelMultiLayerPerceptron", "Module", "None", "ReLU", "Tensor", "__init__", "activation", "batch_norm", "class", "config", "def", "forward", "hidden_state", "in_channels", "int", "linear", "nn", "out_channels", "return", "self", "torch", "transpose"], "superglue/modeling_superglue.py:SuperGlueKeypointEncoder": ["False", "Linear", "Model", "ModelConfig", "ModelKeypointEncoder", "ModelMultiLayerPerceptron", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_hidden_states", "append", "bool", "cat", "class", "config", "def", "dim", "else", "encoder", "encoder_channels", "for", "forward", "hidden_size", "hidden_state", "i", "if", "in", "keypoint_encoder_sizes", "keypoints", "layer", "layer_sizes", "layers", "len", "nn", "output_hidden_states", "range", "return", "scores", "self", "torch", "tuple", "unsqueeze"], "superglue/modeling_superglue.py:SuperGlueSelfAttention": ["Dropout", "Embedding", "False", "FloatTensor", "Linear", "Model", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bhld", "bhlr", "bhrd", "bool", "class", "config", "context_layer", "contiguous", "current_states", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "functional", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "key_length", "long", "lrd", "math", "matmul", "max_position_embeddings", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "query_length", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "shape", "size", "softmax", "sqrt", "the", "to", "torch", "transpose", "tuple", "value", "value_layer", "view"], "superglue/modeling_superglue.py:SuperGlueSelfOutput": ["Linear", "Model", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "args", "class", "config", "def", "dense", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "torch"], "superglue/modeling_superglue.py:SuperGlueAttention": ["False", "FloatTensor", "Model", "ModelAttention", "ModelSelfOutput", "Model_SELF_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "torch", "tuple", "union", "value"], "superglue/modeling_superglue.py:SuperGlueAttentionalPropagation": ["False", "Linear", "Model", "ModelAttention", "ModelAttentionalPropagation", "ModelConfig", "ModelMultiLayerPerceptron", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_hidden_states", "append", "attention", "attention_mask", "attention_outputs", "bool", "cat", "class", "config", "def", "descriptors", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "hidden_size", "hidden_state", "i", "if", "in", "layer", "layers", "len", "mlp", "mlp_channels", "nn", "output", "output_attentions", "output_hidden_states", "range", "return", "self", "torch", "tuple"], "superglue/modeling_superglue.py:SuperGlueAttentionalGNN": ["False", "Model", "ModelAttentionalGNN", "ModelAttentionalPropagation", "ModelConfig", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_attentions", "all_hidden_states", "attention_mask", "batch_size", "bool", "class", "config", "cross", "def", "delta", "descriptors", "else", "encoder_attention_mask", "encoder_hidden_states", "flip", "for", "forward", "gnn_layer", "gnn_layers_types", "gnn_outputs", "hidden_size", "if", "in", "is", "layer_type", "layers", "layers_types", "len", "mask", "nn", "not", "num_keypoints", "output_attentions", "output_hidden_states", "range", "reshape", "return", "self", "shape", "torch", "tuple", "zip"], "superglue/modeling_superglue.py:SuperGlueFinalProjection": ["Linear", "Model", "ModelConfig", "ModelFinalProjection", "Module", "None", "Tensor", "True", "__init__", "bias", "class", "config", "def", "descriptors", "final_proj", "forward", "hidden_size", "nn", "return", "self", "torch"], "superglue/modeling_superglue.py:SuperGluePreTrainedModel": ["BatchNorm1d", "Conv2d", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "bin_score", "class", "config", "data", "def", "elif", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "superglue/modeling_superglue.py:SuperGlueForKeypointMatching": ["AutoModelForKeypointDetection", "FloatTensor", "Input", "KeypointMatchingOutput", "LongTensor", "Model", "ModelAttentionalGNN", "ModelConfig", "ModelFinalProjection", "ModelForKeypointMatching", "ModelKeypointEncoder", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "Union", "ValueError", "_", "__init__", "_match_image_pair", "a", "absolute_keypoints", "all_attentions", "all_hidden_states", "arange_like", "attentions", "auto_docstring", "batch_size", "be", "bin_score", "bool", "cat", "channels", "class", "clone", "config", "def", "descriptors", "device", "dim", "dtype", "else", "encoded_keypoints", "exp", "extended_attention_mask", "final_descriptors", "final_descriptors0", "final_descriptors1", "final_projection", "finfo", "for", "forward", "from_config", "gather", "get_extended_attention_mask", "gnn", "gnn_outputs", "height", "hidden_size", "hidden_states", "if", "in", "indices", "indices0", "indices1", "input_shape", "int", "is", "iterations", "keypoint_detections", "keypoint_detector", "keypoint_detector_config", "keypoint_encoder", "keypoints", "labels", "last_hidden_state", "log_optimal_transport", "logical_and", "loss", "mask", "mask0", "mask1", "masked_fill", "matches", "matches0", "matches1", "matching_scores", "matching_scores0", "matching_scores1", "matching_threshold", "max", "max0", "max1", "min", "must", "mutual0", "mutual1", "ndim", "new_full", "new_tensor", "new_zeros", "nn", "no", "normalize_keypoints", "not", "num_channels", "num_keypoints", "of", "ones", "or", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "projected_descriptors", "provided", "r", "raise", "register_parameter", "reshape", "return", "return_dict", "scores", "self", "shape", "should", "sinkhorn_iterations", "size", "tensor", "to", "torch", "trainable", "transpose", "tuple", "unsqueeze", "use_return_dict", "v", "valid0", "valid1", "values", "where", "width", "x", "zero"], "fsmt/modeling_fsmt.py:invert_mask": ["Model_mask", "assert", "attention_mask", "def", "dim", "eq", "return"], "fsmt/modeling_fsmt.py:triu_onnx": ["Model_onnx", "arange", "def", "device", "diagonal", "expand", "if", "l", "mask", "masked_fill", "return", "shape", "torch", "unsqueeze", "x"], "fsmt/modeling_fsmt.py:_prepare_fsmt_decoder_inputs": ["None", "_prepare_Model_decoder_inputs", "bsz", "causal_mask", "causal_mask_dtype", "config", "decoder_input_ids", "decoder_padding_mask", "def", "device", "dtype", "else", "fill_with_neg_inf", "float32", "if", "input_ids", "invert_mask", "is", "make_padding_mask", "pad_token_id", "return", "shift_tokens_right", "size", "tgt_len", "to", "torch", "triu_onnx", "zeros"], "fsmt/modeling_fsmt.py:PretrainedFSMTModel": ["Embedding", "False", "Linear", "ModelConfig", "ModelModel", "ModelModelModel", "None", "Parameter", "SinusoidalPositionalEmbedding", "_init_weights", "attention_mask", "base_model_prefix", "bias", "class", "config", "data", "def", "detach_", "device", "dummy_inputs", "elif", "get_embedding", "if", "init_std", "input_ids", "is", "isinstance", "mean", "model", "module", "ne", "nn", "normal_", "not", "pad_token", "pad_token_id", "padding_idx", "property", "requires_grad", "return", "self", "shape", "std", "tensor", "torch", "weight", "zero_"], "fsmt/modeling_fsmt.py:_make_linear_from_emb": ["False", "Linear", "_make_linear_from_emb", "bias", "data", "def", "emb", "emb_size", "lin_layer", "nn", "return", "shape", "vocab_size", "weight"], "fsmt/modeling_fsmt.py:_check_shapes": ["AssertionError", "_check_shapes", "def", "f", "if", "mismatch", "raise", "shape", "shape2", "shape_1"], "fsmt/modeling_fsmt.py:shift_tokens_right": ["Model_tokens_right", "clone", "def", "dim", "gather", "index_of_eos", "input_ids", "masked_fill_", "ne", "pad_token_id", "prev_output_tokens", "return", "squeeze", "sum", "unsqueeze"], "fsmt/modeling_fsmt.py:make_padding_mask": ["Model_padding_mask", "None", "any", "def", "eq", "if", "input_ids", "not", "padding_idx", "padding_mask", "return"], "fsmt/modeling_fsmt.py:EncoderLayer": ["ACT2FN", "Attention", "False", "LayerNorm", "Linear", "ModelConfig", "ModelLayer", "Model_attention_heads", "Model_ffn_dim", "Model_padding_mask", "Module", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attn_weights", "class", "config", "d_model", "def", "dropout", "embed_dim", "fc1", "fc2", "final_layer_norm", "forward", "functional", "key", "key_padding_mask", "layer_head_mask", "nn", "output_attentions", "p", "query", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "training", "x"], "fsmt/modeling_fsmt.py:FSMTEncoder": ["BaseModelOutput", "EncoderLayer", "False", "Model", "ModelConfig", "Module", "ModuleList", "None", "Optional", "SinusoidalPositionalEmbedding", "Tensor", "The", "True", "ValueError", "You", "_", "__init__", "all_attentions", "and", "assert", "at", "attention_mask", "attentions", "attn", "be", "bool", "both", "but", "cannot", "class", "config", "def", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "embedding_dim", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "eq", "f", "for", "forward", "functional", "have", "head_mask", "hidden_states", "idx", "if", "in", "input_ids", "inputs_embeds", "invert_mask", "is", "it", "last_hidden_state", "layer_head_mask", "layerdrop", "layers", "len", "masked_fill", "math", "max_position_embeddings", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "padding_idx", "position_ids", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "transpose", "tuple", "v", "x"], "fsmt/modeling_fsmt.py:DecoderLayer": ["ACT2FN", "Attention", "False", "LayerNorm", "Linear", "ModelConfig", "ModelLayer", "Model_attention_heads", "Model_ffn_dim", "Model_padding_mask", "Module", "None", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "assert", "attention_dropout", "attn_mask", "cache_key", "cache_position", "causal_mask", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "def", "dropout", "embed_dim", "encoder_Model_attention", "encoder_attn", "encoder_attn_layer_norm", "encoder_attn_mask", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "key", "key_padding_mask", "layer_head_mask", "layer_idx", "layer_state", "nn", "num_heads", "output_attentions", "p", "query", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "training", "x"], "fsmt/modeling_fsmt.py:FSMTDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DecoderLayer", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "GatheredParameters", "Linear", "Model", "ModelConfig", "Module", "ModuleList", "None", "Optional", "Passing", "SinusoidalPositionalEmbedding", "Tensor", "The", "Transformers", "True", "ValueError", "You", "__init__", "_tie_weights", "a", "all_cross_attns", "all_hidden_states", "all_self_attns", "an", "and", "assert", "at", "attentions", "attn_mask", "be", "bias", "bool", "both", "but", "cache_position", "cannot", "causal_mask", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_causal_mask", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "decoder_padding_mask", "deepspeed", "def", "deprecated", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_dim", "embed_positions", "embed_scale", "embed_tokens", "embed_tokens_weight_shape", "embedding_dim", "encoder_attn_mask", "encoder_hidden_states", "encoder_padding_mask", "enumerate", "eq", "f", "for", "forward", "from_legacy_cache", "functional", "g", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "input_ids", "inputs_embeds", "instance", "instead", "invert_mask", "is", "is_deepspeed_zero3_enabled", "isinstance", "it", "last_hidden_state", "layer_cross_attn", "layer_head_mask", "layer_idx", "layer_self_attn", "layer_state", "layerdrop", "layers", "len", "logger", "mask_name", "masked_fill", "math", "max_position_embeddings", "modifier_rank", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "output_projection", "p", "padding_idx", "pass", "past_key_values", "position_ids", "positions", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "transpose", "tuple", "use_cache", "v", "v4", "warning_once", "weight", "will", "with", "x", "zero", "zip"], "fsmt/modeling_fsmt.py:_reorder_buffer": ["None", "_reorder_buffer", "attn_cache", "def", "for", "if", "in", "index_select", "input_buffer_k", "is", "items", "k", "new_order", "not", "return"], "fsmt/modeling_fsmt.py:Attention": ["Cache", "EncoderDecoderCache", "False", "Head", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "__init__", "a", "and", "assert", "attn_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_key", "cache_position", "class", "contiguous", "cross_Model_cache", "curr_past_key_value", "current_states", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "encoder_decoder", "encoder_decoder_Model", "f", "finfo", "for", "forward", "functional", "get", "head_dim", "if", "is", "is_updated", "isinstance", "k_proj", "key", "key_padding_mask", "key_states", "keys", "layer", "layer_head_mask", "layer_idx", "layer_state", "layers", "list", "mask", "masked_fill", "min", "must", "nn", "not", "num_heads", "of", "or", "out_proj", "output_Models", "p", "permute", "q_proj", "query", "query_states", "reshape", "reshaped", "return", "scaling", "self", "self_Model_cache", "should", "single", "size", "softmax", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "unsqueeze", "update", "v_proj", "value_states", "values", "view"], "fsmt/modeling_fsmt.py:fill_with_neg_inf": ["Model_", "Model_with_neg_inf", "def", "dtype", "finfo", "float", "min", "return", "t", "torch", "type_as"], "fsmt/modeling_fsmt.py:_get_shape": ["None", "_get_shape", "def", "getattr", "return", "shape", "t"], "fsmt/modeling_fsmt.py:FSMTModel": ["BaseModelOutput", "BoolTensor", "Cache", "Embedding", "False", "FloatTensor", "LongTensor", "Make", "Model", "ModelConfig", "ModelDecoder", "ModelEncoder", "None", "Optional", "PretrainedModel", "Seq2SeqModelOutput", "Tensor", "Union", "ValueError", "__init__", "_prepare_Model_decoder_inputs", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "are", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask", "causal_mask_dtype", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_causal_mask", "decoder_embed_tokens", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_padding_mask", "def", "dtype", "elif", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_embed_tokens", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "nn", "not", "or", "output_attentions", "output_hidden_states", "output_projection", "pad_token_id", "padding_idx", "passed", "past_key_values", "post_init", "r", "raise", "return", "return_dict", "self", "set_input_embeddings", "set_output_embeddings", "src_vocab_size", "super", "sure", "tgt_vocab_size", "that", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "weight"], "fsmt/modeling_fsmt.py:FSMTForConditionalGeneration": ["BoolTensor", "Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "LongTensor", "Model", "ModelConfig", "ModelModel", "None", "Optional", "PretrainedModelModel", "Seq2SeqLMOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "auto_docstring", "base_model", "base_model_prefix", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_encoder", "get_output_embeddings", "head_mask", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_logits", "logits", "loss", "loss_fct", "masked_lm_loss", "model", "not", "output", "output_attentions", "output_hidden_states", "output_projection", "outputs", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "r", "return", "return_dict", "self", "set_output_embeddings", "shift_tokens_right", "super", "tgt_vocab_size", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "weight"], "fsmt/modeling_fsmt.py:SinusoidalPositionalEmbedding": ["Any", "Embedding", "False", "ModelPositionalEmbedding", "None", "Optional", "Parameter", "Tensor", "__init__", "arange", "bsz", "cat", "class", "cos", "cumsum", "def", "detach_", "device", "dim", "dtype", "emb", "embedding_dim", "exp", "float", "forward", "get_embedding", "half_dim", "if", "incremental_state", "input", "int", "int64", "is", "log", "long", "make_positions", "make_weight", "mask", "math", "max_pos", "ne", "nn", "not", "num_embeddings", "num_positions", "padding_idx", "positions", "requires_grad", "return", "self", "seq_len", "shape", "sin", "size", "staticmethod", "super", "tensor", "timestep", "to", "torch", "type_as", "unsqueeze", "view", "weight", "zeros"], "glm4/modeling_glm4.py:Glm4MLP": ["ACT2FN", "False", "FloatTensor", "Linear", "ModelMLP", "Module", "__init__", "activation_fn", "bias", "chunk", "class", "config", "def", "dim", "down_proj", "forward", "gate", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch", "up_states"], "glm4/modeling_glm4.py:Glm4DecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_mlp_layernorm", "post_self_attn_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "glm4/modeling_glm4.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "glm4/modeling_glm4.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "glm4/modeling_glm4.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "glm4/modeling_glm4.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "repeat_interleave", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "glm4/modeling_glm4.py:Glm4Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "glm4/modeling_glm4.py:Glm4RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "glm4/modeling_glm4.py:Glm4RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "glm4/modeling_glm4.py:Glm4PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "glm4/modeling_glm4.py:Glm4Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "glm4/modeling_glm4.py:Glm4ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "tuple", "use_cache", "vocab_size", "weight"], "glm4/modeling_glm4.py:Glm4ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "glm4/modeling_glm4.py:Glm4ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "owlvit/modeling_owlvit.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "owlvit/modeling_owlvit.py:owlvit_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "owlvit/modeling_owlvit.py:OwlViTOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "owlvit/modeling_owlvit.py:_upcast": ["Tensor", "_upcast", "def", "dtype", "else", "float", "float32", "float64", "if", "in", "int", "int32", "int64", "is_floating_point", "return", "t", "torch"], "owlvit/modeling_owlvit.py:box_area": ["Model_area", "Modeles", "Tensor", "_upcast", "def", "return"], "owlvit/modeling_owlvit.py:box_iou": ["Model_area", "Model_iou", "Modeles1", "Modeles2", "None", "area1", "area2", "clamp", "def", "inter", "iou", "left_top", "max", "min", "return", "right_bottom", "torch", "union", "width_height"], "owlvit/modeling_owlvit.py:generalized_box_iou": ["Model_box_iou", "None", "ValueError", "all", "area", "be", "bottom_right", "box_iou", "boxes1", "boxes2", "but", "clamp", "corner", "def", "f", "format", "got", "if", "in", "iou", "max", "min", "must", "not", "raise", "return", "top_left", "torch", "union", "width_height", "x0", "x1", "y0", "y1"], "owlvit/modeling_owlvit.py:OwlViTObjectDetectionOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "class", "class_embeds", "def", "dict", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits", "loss", "loss_dict", "not", "pred_boxes", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "owlvit/modeling_owlvit.py:OwlViTImageGuidedObjectDetectionOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelImageGuidedObjectDetectionOutput", "ModelOutput", "None", "Optional", "class", "class_embeds", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits", "not", "query_image_embeds", "query_pred_boxes", "r", "return", "self", "target_pred_boxes", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "owlvit/modeling_owlvit.py:OwlViTVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "bool", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "owlvit/modeling_owlvit.py:OwlViTTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "arange", "class", "config", "def", "else", "embeddings", "expand", "forward", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "max_position_embeddings", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "shape", "super", "token_embedding", "torch", "vocab_size"], "owlvit/modeling_owlvit.py:OwlViTAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_shape", "and", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bmm", "bool", "bsz", "but", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_states", "mask", "must", "nn", "not", "num_attention_heads", "num_heads", "of", "out_proj", "output_attentions", "p", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scale", "self", "seq_len", "should", "size", "softmax", "src_len", "super", "tensor", "tgt_len", "to", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights"], "owlvit/modeling_owlvit.py:OwlViTMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "owlvit/modeling_owlvit.py:OwlViTEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "owlvit/modeling_owlvit.py:OwlViTPreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "ModelModel", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "logit_scale", "logit_scale_init_value", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_embedding", "v_proj", "vision_embed_dim", "visual_projection", "weight", "zero_"], "owlvit/modeling_owlvit.py:OwlViTEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "for", "forward", "gradient_checkpointing", "hidden_states", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "owlvit/modeling_owlvit.py:OwlViTTextTransformer": ["BaseModelOutputWithPooling", "LayerNorm", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "final_layer_norm", "forward", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "r", "return", "return_dict", "self", "shape", "size", "super", "to", "torch", "tuple", "use_return_dict", "view"], "owlvit/modeling_owlvit.py:OwlViTTextModel": ["BaseModelOutputWithPooling", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "tuple", "value"], "owlvit/modeling_owlvit.py:OwlViTVisionTransformer": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "dtype", "else", "embeddings", "encoder", "encoder_outputs", "eps", "expected_input_dtype", "forward", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layernorm", "return", "return_dict", "self", "super", "to", "torch", "tuple", "use_return_dict", "weight"], "owlvit/modeling_owlvit.py:OwlViTVisionModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "vision_model"], "owlvit/modeling_owlvit.py:OwlViTModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextTransformer", "ModelVisionConfig", "ModelVisionTransformer", "Model_loss", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "Union", "__init__", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "class", "config", "def", "device", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "linalg", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "ord", "output", "output_attentions", "output_hidden_states", "pixel_values", "pooler_output", "post_init", "projection_dim", "r", "raise", "return", "return_base_image_embeds", "return_dict", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_embeds_norm", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "use_return_dict", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "owlvit/modeling_owlvit.py:OwlViTBoxPredictionHead": ["FloatTensor", "GELU", "Linear", "ModelBoxPredictionHead", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense0", "dense1", "dense2", "forward", "gelu", "hidden_size", "image_features", "int", "nn", "out_dim", "output", "return", "self", "super", "torch", "vision_config", "width"], "owlvit/modeling_owlvit.py:OwlViTClassPredictionHead": ["ELU", "FloatTensor", "Linear", "ModelClassPredictionHead", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "__init__", "batch_size", "class", "config", "def", "dense0", "device", "dim", "dtype", "einsum", "elu", "finfo", "float32", "forward", "hidden_size", "if", "image_class_embeds", "image_embeds", "is", "keepdim", "linalg", "logit_scale", "logit_shift", "min", "ndim", "nn", "norm", "not", "num_patches", "out_dim", "pd", "pq", "pred_logits", "qd", "query_dim", "query_embeds", "query_mask", "return", "self", "shape", "super", "text_config", "to", "torch", "tuple", "unsqueeze", "vision_config", "where", "zeros"], "owlvit/modeling_owlvit.py:OwlViTForObjectDetection": ["False", "FloatTensor", "LayerNorm", "Model", "ModelBoxPredictionHead", "ModelClassPredictionHead", "ModelConfig", "ModelForObjectDetection", "ModelImageGuidedObjectDetectionOutput", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "Please", "Sigmoid", "Tensor", "True", "ValueError", "_", "__init__", "all", "an", "append", "arange", "argmin", "as", "attention_mask", "auto_docstring", "axis", "batch_size", "been", "best_box_ind", "best_box_indices", "best_class_embeds", "bool", "box_bias", "box_coord_bias", "box_coordinates", "box_head", "box_indices", "box_iou", "box_predictor", "box_size", "box_size_bias", "broadcast_to", "cat", "center_to_corners_format", "class", "class_embeds", "class_head", "class_predictor", "class_token_out", "clip", "compute_box_bias", "config", "d", "def", "deprecated", "device", "dim", "dtype", "each_query_box", "each_query_pred_boxes", "einsum", "else", "embed_image_query", "eps", "feature_map", "float32", "for", "forward", "full_like", "generalized_box_iou", "has", "height", "hidden_dim", "hidden_size", "i", "id", "if", "image_class_embeds", "image_embedder", "image_embeds", "image_feats", "image_guided_detection", "image_size", "image_text_embedder", "in", "indexing", "input", "input_ids", "instead", "int", "interpolate_pos_encoding", "iou_threshold", "ious", "is", "last_hidden_state", "layer_norm", "layer_norm_eps", "log", "log1p", "logits", "lru_cache", "max", "max_text_queries", "maxsize", "mean", "mean_embeds", "mean_sim", "meshgrid", "new_size", "nn", "nonzero", "normalize_grid_corner_coordinates", "not", "num_patches", "num_patches_height", "num_patches_width", "numel", "output", "output_attentions", "output_hidden_states", "outputs", "pass", "patch_size", "pixel_values", "post_layernorm", "pred_boxes", "pred_boxes_as_corners", "pred_boxes_device", "pred_logits", "query_embeds", "query_feature_map", "query_image_embeds", "query_image_feats", "query_image_features", "query_mask", "query_pixel_values", "query_pred_boxes", "r", "raise", "range", "reshape", "return", "return_dict", "selected_embeddings", "selected_inds", "self", "shape", "sigmoid", "squeeze", "stack", "staticmethod", "super", "target_pred_boxes", "tensor", "text_embeds", "text_model_output", "text_outputs", "to", "to_tuple", "torch", "tuple", "view", "vision_config", "vision_model", "vision_model_output", "vision_outputs", "width", "x", "x_coordinates", "xx", "xy", "y_coordinates", "yy"], "llama4/modeling_llama4.py:Llama4TextExperts": ["ACT2FN", "ModelTextConfig", "ModelTextExperts", "Module", "Parameter", "Tensor", "__init__", "act_fn", "bmm", "chunk", "class", "config", "def", "dim", "down_proj", "empty", "expert_dim", "forward", "gate", "gate_up", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "next_states", "nn", "num_experts", "num_local_experts", "return", "self", "shape", "super", "torch", "up", "view"], "llama4/modeling_llama4.py:Llama4TextMLP": ["ACT2FN", "False", "Linear", "ModelTextMLP", "Module", "None", "__init__", "activation_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "return", "self", "super", "up_proj", "x"], "llama4/modeling_llama4.py:Llama4TextL2Norm": ["ModelTextL2Norm", "Module", "True", "__init__", "_norm", "class", "def", "eps", "extra_repr", "f", "float", "forward", "keepdim", "mean", "nn", "pow", "return", "rsqrt", "self", "super", "torch", "type_as", "x"], "llama4/modeling_llama4.py:Llama4TextRMSNorm": ["ModelTextRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "eps", "extra_repr", "f", "float", "forward", "hidden_size", "keepdim", "mean", "nn", "ones", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x"], "llama4/modeling_llama4.py:Llama4Router": ["False", "Linear", "ModelRouter", "__init__", "bias", "class", "config", "def", "dim", "dtype", "float", "forward", "full_like", "functional", "hidden_size", "hidden_states", "inf", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "return", "router_indices", "router_logits", "router_scores", "router_top_value", "scatter_", "self", "sigmoid", "super", "to", "top_k", "topk", "torch"], "llama4/modeling_llama4.py:Llama4TextMoe": ["ModelRouter", "ModelTextExperts", "ModelTextMLP", "ModelTextMoe", "Module", "__init__", "add_", "class", "config", "def", "dim", "experts", "forward", "hidden_dim", "hidden_size", "hidden_states", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "out", "repeat", "reshape", "return", "routed_in", "routed_out", "router", "router_logits", "router_scores", "self", "shape", "shared_expert", "sum", "super", "top_k", "transpose"], "llama4/modeling_llama4.py:Llama4TextRotaryEmbedding": ["False", "Model3", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "class", "config", "cpu", "def", "default", "device", "device_type", "dynamic_rope_update", "else", "enabled", "expand", "float", "forward", "freqs", "freqs_cis", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "ones_like", "original_inv_freq", "original_max_seq_len", "persistent", "polar", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "llama4/modeling_llama4.py:apply_rotary_emb": ["Model_rotary_emb", "None", "Tensor", "def", "flatten", "float", "freqs_cis", "reshape", "return", "shape", "torch", "tuple", "type_as", "view_as_complex", "view_as_real", "xk", "xk_", "xk_out", "xq", "xq_", "xq_out"], "llama4/modeling_llama4.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "llama4/modeling_llama4.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "float", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "torch", "training", "transpose", "value", "value_states"], "llama4/modeling_llama4.py:vision_eager_attention_forward": ["Model_eager_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "float", "functional", "head_dim", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "torch", "training", "transpose", "value", "value_states"], "llama4/modeling_llama4.py:Llama4TextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextL2Norm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "apply_rotary_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_scale", "attn_scales", "attn_temperature_tuning", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "def", "deprecate_kwarg", "device", "dropout", "dtype", "eager", "eager_attention_forward", "else", "expand", "float", "floor", "floor_scale", "forward", "getattr", "hasattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "log1p", "new_name", "nn", "no_rope_layers", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "qk_norm", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "super", "to", "torch", "training", "transpose", "tuple", "update", "use_qk_norm", "use_rope", "v_proj", "value_states", "version", "view"], "llama4/modeling_llama4.py:Llama4TextDecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelTextAttention", "ModelTextDecoderLayer", "ModelTextMLP", "ModelTextMoe", "ModelTextRMSNorm", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "attention_mask", "attention_states", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "feed_forward", "forward", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "intermediate_size", "intermediate_size_mlp", "is_moe_layer", "kwargs", "layer_idx", "layer_types", "moe_layers", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "shape", "super", "torch", "tuple", "use_cache", "version", "view"], "llama4/modeling_llama4.py:Llama4PreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelTextExperts", "ModelTextRMSNorm", "ModelVisionModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "bias", "class", "class_embedding", "config", "data", "def", "down_proj", "elif", "else", "fill_", "gate_up_proj", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "positional_embedding_vlm", "scale", "self", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "llama4/modeling_llama4.py:Llama4TextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextMoe", "ModelTextRMSNorm", "ModelTextRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "_no_split_modules", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "base_model_prefix", "bool", "cache_position", "can_return_tuple", "causal_mask_mapping", "check_model_inputs", "chunked_attention", "class", "config", "create_causal_mask", "create_chunked_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "freq_cis", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_kwargs", "model", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "router_logits", "self", "shape", "specify", "super", "to", "torch", "tuple", "unsqueeze", "use_cache", "vocab_size", "weight"], "llama4/modeling_llama4.py:Llama4ForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_no_split_modules", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "tuple", "use_cache", "vocab_size", "weight"], "llama4/modeling_llama4.py:Llama4CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "llama4/modeling_llama4.py:Llama4VisionMLP2": ["F", "False", "GELU", "Linear", "ModelVisionMLP2", "Module", "__init__", "activation_fn", "bias", "class", "config", "def", "dropout", "fc1", "fc2", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "p", "projector_dropout", "projector_input_dim", "projector_output_dim", "return", "self", "super", "torch", "training"], "llama4/modeling_llama4.py:Llama4MultiModalProjector": ["False", "Linear", "ModelMultiModalProjector", "Module", "__init__", "bias", "class", "config", "def", "forward", "hidden_size", "hidden_states", "image_features", "linear_1", "nn", "return", "self", "super", "text_config", "vision_config", "vision_output_dim"], "llama4/modeling_llama4.py:pixel_shuffle": ["Model_shuffle", "batch_size", "channels", "contiguous", "def", "height", "input_tensor", "int", "math", "num_patches", "output_tensor", "patch_size", "permute", "reshaped_tensor", "return", "shape", "shuffle_ratio", "size", "sqrt", "view", "width"], "llama4/modeling_llama4.py:Llama4VisionPixelShuffleMLP": ["ModelVisionMLP2", "ModelVisionPixelShuffleMLP", "Module", "Tensor", "__init__", "class", "config", "def", "encoded_patches", "forward", "inner_dim", "int", "mlp", "nn", "output_dim", "pixel_shuffle", "pixel_shuffle_ratio", "projector_input_dim", "projector_output_dim", "return", "self", "super", "torch"], "llama4/modeling_llama4.py:reshape_for_broadcast": ["Model_for_broadcast", "Tensor", "d", "def", "else", "enumerate", "for", "freqs_ci", "i", "if", "in", "ndim", "or", "query", "return", "shape", "torch", "view"], "llama4/modeling_llama4.py:vision_apply_rotary_emb": ["Model_apply_rotary_emb", "Tensor", "def", "device", "flatten", "float", "freqs_ci", "key", "key_", "key_out", "query", "query_", "query_out", "reshape", "reshape_for_broadcast", "return", "shape", "to", "torch", "tuple", "type_as", "view_as_complex", "view_as_real"], "llama4/modeling_llama4.py:Llama4VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelVisionAttention", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "class", "config", "contiguous", "def", "dropout", "eager", "else", "embed_dim", "flex_attention", "forward", "freqs_ci", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "in", "input_shape", "is_causal", "k_proj", "key_states", "kwargs", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "o_proj", "past_key_values", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "vision_apply_rotary_emb", "vision_eager_attention_forward"], "llama4/modeling_llama4.py:Llama4VisionMLP": ["GELU", "Linear", "ModelVisionMLP", "Module", "Tensor", "True", "__init__", "activation_fn", "bias", "class", "config", "def", "fc1", "fc2", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "llama4/modeling_llama4.py:Llama4VisionEncoderLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoderLayer", "ModelVisionMLP", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "def", "forward", "freqs_ci", "hidden_size", "hidden_state", "if", "input_layernorm", "mlp", "nn", "output_attentions", "outputs", "post_attention_layernorm", "residual", "return", "self", "self_attn", "super", "torch"], "llama4/modeling_llama4.py:Llama4VisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "encoder_layer", "encoder_states", "for", "forward", "freqs_ci", "gradient_checkpointing", "hidden_state", "hidden_states", "if", "in", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "llama4/modeling_llama4.py:Llama4UnfoldConvolution": ["False", "Linear", "ModelUnfoldConvolution", "Module", "Tensor", "Unfold", "__init__", "bias", "class", "config", "def", "forward", "hidden_size", "hidden_states", "if", "int", "isinstance", "kernel_size", "linear", "nn", "num_channels", "patch_size", "permute", "return", "self", "stride", "super", "torch", "unfold"], "llama4/modeling_llama4.py:Llama4VisionRotaryEmbedding": ["ModelVisionRotaryEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "config", "contiguous", "cos", "def", "device", "dim", "dtype", "float", "forward", "freq_cis", "freq_dim", "freqs", "freqs_ci", "freqs_x", "freqs_y", "frequencies_x", "frequencies_y", "hidden_size", "hidden_states", "idx", "image_size", "img_idx", "int32", "masked_fill", "nn", "num_attention_heads", "patch_size", "repeat_interleave", "reshape", "return", "rope_freq", "rope_theta", "self", "sin", "stack", "super", "to", "torch", "view_as_complex"], "llama4/modeling_llama4.py:Llama4VisionModel": ["BaseModelOutput", "LayerNorm", "ModelPreTrainedModel", "ModelUnfoldConvolution", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "ModelVisionModel", "ModelVisionPixelShuffleMLP", "ModelVisionRotaryEmbedding", "None", "Optional", "Parameter", "Tensor", "Union", "_", "__init__", "_no_split_modules", "attention_mask", "attentions", "base_model_prefix", "batch_size_times_num_tiles", "bool", "cat", "class", "class_embedding", "config", "def", "device", "dim", "dtype", "else", "expand", "for", "forward", "freqs_ci", "get_input_embeddings", "height", "hidden_dim", "hidden_size", "hidden_state", "hidden_states", "if", "image_size", "in", "is", "last_hidden_state", "layernorm_post", "layernorm_pre", "model", "nn", "not", "num_channels", "num_chunks", "num_concurrent_media", "num_patches", "output", "output_attentions", "output_hidden_states", "patch_embedding", "patch_size", "pixel_values", "positional_embedding", "positional_embedding_vlm", "post_init", "r", "randn", "reshape", "return", "return_dict", "rotary_embedding", "scale", "self", "shape", "super", "to", "torch", "tuple", "use_return_dict", "v", "view", "vision_adapter", "vision_model", "width"], "llama4/modeling_llama4.py:Llama4ForConditionalGeneration": ["Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Image", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForCausalLM", "ModelForConditionalGeneration", "ModelMultiModalProjector", "ModelPreTrainedModel", "ModelTextDecoderLayer", "ModelVisionEncoderLayer", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unexpected", "Union", "Unpack", "ValueError", "You", "__init__", "_no_split_modules", "_tp_plan", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bool", "both", "cache_position", "cannot", "class", "config", "contiguous", "decoder", "def", "default", "deprecate_kwarg", "device", "do", "dtype", "either", "else", "exactly", "expand_as", "f", "feature", "features", "for", "forward", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_placeholder_mask", "hidden_state", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "items", "k", "kwargs", "labels", "language_model", "last_hidden_state", "list", "logits", "logits_to_keep", "long", "loss", "loss_fct", "masked_scatter", "match", "model_inputs", "multi_modal_projector", "must", "n_image_tokens", "new_embeddings", "nn", "not", "numel", "of", "one", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "projected_vision_flat", "r", "raise", "return", "return_dict", "same", "select", "self", "set_decoder", "set_input_embeddings", "set_output_embeddings", "shape", "shift_attention_mask", "shift_labels", "shift_logits", "size", "special_image_mask", "specify", "str", "strategy", "sum", "super", "tensor", "text_config", "the", "time", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "value", "version", "view", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_flat", "vision_model", "vocab_size"], "mamba/modeling_mamba.py:_lazy_load_causal_conv1d": ["None", "_causal_conv1d_cache", "_causal_conv1d_kernel", "_lazy_load_causal_conv1d", "causal", "causal_conv1d_fn", "causal_conv1d_update", "community", "conv1d", "def", "elif", "else", "get_kernel", "global", "if", "is", "is_causal_conv1d_available", "is_kernels_available", "kernels", "not", "return"], "mamba/modeling_mamba.py:MambaCache": ["LongTensor", "ModelCache", "None", "PretrainedConfig", "Tensor", "True", "Union", "_", "__init__", "_dtype", "_dynamo", "append", "cache_position", "clamp", "class", "config", "conv_kernel", "conv_kernel_size", "conv_state", "conv_states", "def", "device", "dims", "dtype", "else", "float16", "for", "if", "in", "int", "intermediate_size", "is", "is_compileable", "layer_idx", "len", "list", "mark_static_address", "max_batch_size", "new_conv_state", "new_ssm_state", "not", "num_hidden_layers", "range", "reset", "return", "roll", "self", "shifts", "ssm_state", "ssm_state_size", "ssm_states", "state_size", "str", "to", "torch", "update_conv_state", "update_ssm_state", "zero_", "zeros"], "mamba/modeling_mamba.py:MambaMixer": ["A", "ACT2FN", "AILab", "A_log", "B", "C", "Conv1d", "D", "Dao", "Falling", "False", "For", "ImportError", "Linear", "LongTensor", "Model", "ModelCache", "ModelConfig", "ModelMixer", "Model_inner_fn", "Modelpy", "Module", "None", "Optional", "Parameter", "Tensor", "The", "To", "True", "_", "__init__", "_dynamo", "_lazy_load_causal_conv1d", "act", "activation", "all", "alxndrTL", "and", "append", "arange", "as", "attention_mask", "available", "back", "backend", "batch_size", "because", "bias", "but", "cache_params", "cache_position", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk", "class", "clone", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_kernel", "conv_kernel_size", "conv_state", "conv_states", "conv_weights", "copy_", "cuda", "cuda_kernels_forward", "def", "deltaB_u", "delta_bias", "delta_softplus", "device", "dim", "discrete_A", "discrete_B", "discrete_time_step", "dt_proj", "dt_softplus", "dtype", "else", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "hasattr", "hidden_act", "hidden_size", "hidden_states", "hs", "https", "i", "if", "implementation", "in", "in_channels", "in_proj", "input_states", "install", "installed", "int", "intermediate_size", "is", "is_Modelpy_available", "is_compiling", "is_fast_path_available", "it", "kernel_size", "kernels", "layer_idx", "library", "log", "logger", "matmul", "nn", "not", "of", "one", "ones", "or", "out_channels", "out_proj", "package", "pad", "padding", "path", "pip", "projected_states", "pscan", "py", "raise", "range", "return", "return_last_state", "scan_output", "scan_outputs", "selective_scan_fn", "selective_state_update", "self", "seq_len", "sequential", "set", "shape", "size", "slow_forward", "softplus", "spaces", "split", "squeeze", "ssm_parameters", "ssm_state", "ssm_state_size", "ssm_states", "stack", "state", "state_size", "sum", "super", "the", "time_proj_bias", "time_step", "time_step_rank", "to", "torch", "training", "transpose", "type", "unsqueeze", "update_conv_state", "update_ssm_state", "use_Modelpy", "use_bias", "use_conv_bias", "using", "view", "warn_slow_implementation", "warning_once", "weight", "x_proj", "zeros"], "mamba/modeling_mamba.py:MambaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "mamba/modeling_mamba.py:MambaBlock": ["GradientCheckpointingLayer", "LongTensor", "ModelBlock", "ModelCache", "ModelMixer", "ModelRMSNorm", "None", "Optional", "__init__", "attention_mask", "cache_params", "cache_position", "class", "config", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "if", "layer_idx", "layer_norm_epsilon", "mixer", "norm", "residual", "residual_in_fp32", "return", "self", "super", "to", "torch", "weight"], "mamba/modeling_mamba.py:MambaPreTrainedModel": ["A", "A_log", "D", "Embedding", "False", "Linear", "ModelBlock", "ModelConfig", "ModelMixer", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_reinit", "_no_split_modules", "a", "arange", "backbone", "base_model_prefix", "bias", "clamp", "class", "config", "constant", "constant_", "contiguous", "conv1d", "copy_", "data", "def", "dt", "dt_init_std", "dt_proj", "dtype", "elif", "exp", "expand", "expm1", "fill_", "float32", "getattr", "if", "init", "initializer_range", "intermediate_size", "inv_dt", "is", "isinstance", "kaiming_uniform_", "log", "math", "min", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "p", "rand", "random", "rescale_prenorm_residual", "self", "sqrt", "ssm_state_size", "std", "supports_gradient_checkpointing", "time_step_floor", "time_step_init_scheme", "time_step_max", "time_step_min", "time_step_rank", "time_step_scale", "torch", "uniform_", "weight", "zeros_"], "mamba/modeling_mamba.py:MambaOutput": ["FloatTensor", "ModelCache", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "mamba/modeling_mamba.py:MambaCausalLMOutput": ["FloatTensor", "ModelCache", "ModelCausalLMOutput", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "mamba/modeling_mamba.py:MambaModel": ["Embedding", "False", "LongTensor", "ModelBlock", "ModelCache", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_register_load_state_dict_pre_hook", "a", "all_hidden_states", "and", "arange", "are", "args", "attention_mask", "auto_docstring", "automatically", "be", "because", "bool", "break", "cache_params", "cache_position", "case", "class", "config", "conv_kernel", "def", "device", "don", "dtype", "elif", "else", "embedding", "embeddings", "eps", "exactly", "for", "forward", "get_input_embeddings", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "initialized", "input_ids", "inputs_embeds", "is", "it", "k", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layers", "load_hook", "manually", "mixer_block", "must", "new_embeddings", "nn", "norm_f", "not", "num_hidden_layers", "of", "one", "or", "output_hidden_states", "pass", "passed", "pop", "post_init", "prefilling", "prefix", "r", "raise", "range", "replace", "return", "return_dict", "self", "set_input_embeddings", "size", "specify", "stage", "state_dict", "super", "t", "that", "the", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "vocab_size", "when", "will", "you"], "mamba/modeling_mamba.py:MambaForCausalLM": ["Any", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCache", "ModelCausalLMOutput", "ModelForCausalLM", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "_update_model_kwargs_for_generation", "and", "arange", "attention_mask", "auto_docstring", "backbone", "bias", "bool", "cache_params", "cache_position", "cat", "class", "config", "contiguous", "conv_kernel", "def", "device", "dict", "dim", "dtype", "else", "float", "for", "forward", "get", "get_input_embeddings", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "items", "key", "kwargs", "labels", "lm_head", "logits", "loss", "loss_fct", "max_batch_size", "model_inputs", "model_kwargs", "new_embeddings", "new_ones", "nn", "not", "num_new_tokens", "output", "output_hidden_states", "outputs", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "set_input_embeddings", "shape", "shift_labels", "shift_logits", "size", "str", "super", "to", "torch", "tuple", "unsqueeze", "update", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "vision_encoder_decoder/modeling_vision_encoder_decoder.py:shift_tokens_right": ["Make", "Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "attribute", "clone", "configuration", "decoder_start_token_id", "def", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "of", "pad_token_id", "raise", "return", "s", "set", "shape", "sure", "the", "to", "torch"], "vision_encoder_decoder/modeling_vision_encoder_decoder.py:VisionEncoderDecoderModel": ["AutoConfig", "AutoModel", "AutoModelForCausalLM", "BaseModelOutput", "BoolTensor", "Cache", "Config", "Cross", "Decoder", "Either", "False", "FloatTensor", "GenerationMixin", "Got", "Head", "If", "In", "Initializing", "LM", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "None", "Optional", "Please", "PreTrainedModel", "PretrainedConfig", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "ValueError", "You", "__class__", "__init__", "_attn_implementation", "_can_compile_fullgraph", "_supports_flash_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "a", "add_cross_attention", "added", "allows", "an", "and", "architecture", "are", "argument", "as", "attention", "attention_mask", "attentions", "attributes", "auto_docstring", "base_model_prefix", "be", "bool", "by", "cache_position", "causal", "class", "classmethod", "cls", "config", "config_class", "configuration", "cross", "cross_attention_hidden_size", "cross_attentions", "decoder", "decoder_", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_model", "decoder_outputs", "decoder_pretrained_model_name_or_path", "decoder_start_token_id", "def", "defined", "del", "disabled", "do", "elif", "else", "enc_to_dec_proj", "encoder", "encoder_", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_model", "encoder_outputs", "encoder_pretrained_model_name_or_path", "equal", "f", "for", "forward", "from", "from_config", "from_encoder_decoder_configs", "from_encoder_decoder_pretrained", "from_pretrained", "get_encoder", "get_input_embeddings", "get_output_embeddings", "has", "have", "hidden_size", "hidden_states", "if", "in", "info", "initialize", "initialized", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "it", "items", "key", "kwargs", "kwargs_decoder", "kwargs_encoder", "labels", "last_hidden_state", "layers", "len", "logger", "logits", "loss", "loss_function", "main_input_name", "make", "mask", "model", "model_args", "new_embeddings", "nn", "not", "num_items_in_batch", "of", "or", "order", "output_attentions", "output_hidden_states", "overwritten", "pad_token_id", "pass", "passed", "past_key_values", "pixel_values", "pop", "prepare_decoder_input_ids_from_labels", "provided", "r", "raise", "randomly", "return", "return_dict", "return_unused_kwargs", "s", "self", "set", "set_output_embeddings", "shared", "shift_tokens_right", "should", "specified", "specify", "startswith", "str", "super", "supports_gradient_checkpointing", "sure", "that", "the", "tie_word_embeddings", "to", "to_dict", "torch", "tuple", "type", "use", "use_cache", "use_return_dict", "value", "vocab_size", "warning", "without"], "t5gemma/modeling_t5gemma.py:T5GemmaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "t5gemma/modeling_t5gemma.py:T5GemmaMLP": ["ACT2FN", "Dropout", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "dropout", "dropout_rate", "forward", "gate_proj", "hidden_activation", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "t5gemma/modeling_t5gemma.py:T5GemmaRotaryEmbedding": ["False", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "t5gemma/modeling_t5gemma.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "t5gemma/modeling_t5gemma.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "t5gemma/modeling_t5gemma.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "t5gemma/modeling_t5gemma.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_dim", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softcap", "softmax", "tanh", "to", "torch", "training", "transpose", "tuple", "value", "value_states"], "t5gemma/modeling_t5gemma.py:T5GemmaSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelModuleConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_logit_softcapping", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_pre_attn_scalar", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "softcap", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "t5gemma/modeling_t5gemma.py:T5GemmaCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Cross", "Encoder", "False", "FlashAttentionKwargs", "Linear", "ModelCrossAttention", "ModelModuleConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "attention", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_logit_softcapping", "attn_output", "attn_weights", "be", "bias", "class", "config", "contiguous", "cross", "cross_attention_cache", "cross_attention_hidden_size", "curr_past_key_value", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "encoder_hidden_shape", "encoder_hidden_states", "encoder_input_shape", "for", "forward", "get", "getattr", "head_dim", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_updated", "k_proj", "key_states", "keys", "kwargs", "layer_idx", "layers", "needs", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "past_key_value", "past_key_values", "q_proj", "query_pre_attn_scalar", "query_states", "raise", "required", "reshape", "return", "scaling", "self", "shape", "sliding_window", "softcap", "specified", "state", "super", "to", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "t5gemma/modeling_t5gemma.py:T5GemmaEncoderLayer": ["Dropout", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelEncoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSelfAttention", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "attention_type", "class", "config", "def", "dropout", "dropout_rate", "eps", "forward", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "layer_types", "mlp", "nn", "past_key_values", "position_embeddings", "position_ids", "post_feedforward_layernorm", "post_self_attn_layernorm", "pre_feedforward_layernorm", "pre_self_attn_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple"], "t5gemma/modeling_t5gemma.py:T5GemmaDecoderLayer": ["EncoderDecoderCache", "False", "FloatTensor", "LongTensor", "ModelCrossAttention", "ModelDecoderLayer", "ModelEncoderLayer", "ModelRMSNorm", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn", "def", "deprecate_kwarg", "dropout", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "forward", "hidden_size", "hidden_states", "if", "int", "is", "kwargs", "layer_idx", "mlp", "new_name", "not", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_cross_attn_layernorm", "post_feedforward_layernorm", "post_self_attn_layernorm", "pre_cross_attn_layernorm", "pre_feedforward_layernorm", "pre_self_attn_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attention_cache", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "t5gemma/modeling_t5gemma.py:T5GemmaClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "Tensor", "__init__", "class", "classifier_dropout_rate", "def", "dropout", "float", "forward", "hidden_size", "hidden_states", "int", "nn", "num_labels", "out_proj", "p", "return", "self", "super", "torch"], "t5gemma/modeling_t5gemma.py:T5GemmaLMHead": ["False", "Linear", "ModelLMHead", "Module", "Tensor", "__init__", "bias", "bool", "class", "def", "forward", "hidden_size", "hidden_states", "int", "logits", "nn", "out_proj", "return", "self", "super", "torch", "vocab_size"], "t5gemma/modeling_t5gemma.py:T5GemmaPreTrainedModel": ["ModelClassificationHead", "ModelConfig", "ModelCrossAttention", "ModelDecoderLayer", "ModelEncoderLayer", "ModelLMHead", "ModelPreTrainedModel", "ModelSelfAttention", "None", "OutputRecorder", "PreTrainedModel", "RMSNorm", "True", "ValueError", "__class__", "__name__", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_shift_right", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "attentions", "base_model_prefix", "be", "bias", "bos_token_id", "class", "clone", "config", "cross_attn", "data", "decoder", "decoder_start_token_id", "def", "defined", "elif", "has", "hasattr", "hidden_states", "if", "in", "index", "initializer_range", "input_ids", "is", "isinstance", "layer_name", "masked_fill_", "mean", "model", "module", "new_zeros", "normal_", "not", "out_proj", "pad_token_id", "past_key_values", "raise", "return", "scale", "self", "self_attn", "shape", "shifted_input_ids", "std", "super", "supports_gradient_checkpointing", "tie_word_embeddings", "to", "weight", "zero_"], "t5gemma/modeling_t5gemma.py:bidirectional_mask_function": ["Callable", "Model_mask_function", "None", "Optional", "Tensor", "attention_mask", "batch_idx", "bool", "def", "dtype", "head_idx", "if", "inner_mask", "int", "is", "kv_idx", "ones", "q_idx", "return", "to", "torch"], "t5gemma/modeling_t5gemma.py:sliding_window_bidirectional_mask_function": ["Callable", "Model_window", "Model_window_bidirectional_mask_function", "batch_idx", "bool", "def", "head_idx", "inner_mask", "int", "kv_idx", "q_idx", "return"], "t5gemma/modeling_t5gemma.py:make_default_2d_attention_mask": ["LongTensor", "Model_default_2d_attention_mask", "None", "Optional", "Tensor", "ValueError", "attention_mask", "def", "device", "dtype", "else", "for", "hidden_states", "if", "information", "int", "is", "long", "not", "ones", "pad_token_id", "padding", "raise", "required", "return", "shape", "to", "token_ids", "torch"], "t5gemma/modeling_t5gemma.py:T5GemmaEncoder": ["BaseModelOutput", "Dropout", "Embedding", "False", "FloatTensor", "LongTensor", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelSelfAttention", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "and_mask_function", "arange", "attention_mask", "attention_type", "attentions", "bidirectional_mask_function", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "def", "device", "dict", "dropout", "dropout_rate", "dtype", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_module", "layers", "make_default_2d_attention_mask", "mask_kwargs", "must", "nn", "norm", "normalizer", "not", "num_hidden_layers", "of", "one", "or", "or_mask_function", "pad_token_id", "padding_idx", "past_key_values", "pop", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "self_attn_mask_mapping", "shape", "sliding_attention", "sliding_window", "sliding_window_bidirectional_mask_function", "specify", "super", "tensor", "torch", "unsqueeze", "vocab_size"], "t5gemma/modeling_t5gemma.py:T5GemmaDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "EncoderDecoderCache", "FloatTensor", "LongTensor", "ModelCrossAttention", "ModelDecoder", "ModelDecoderLayer", "ModelEncoder", "ModelSelfAttention", "ModuleList", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "and", "arange", "attention_mask", "attention_type", "attentions", "be", "bidirectional_mask_function", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "cross_attentions", "cross_attn_mask_mapping", "decoder", "def", "device", "dict", "dropout", "dtype", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "exactly", "for", "forward", "full_attention", "get_seq_length", "given", "hidden_size", "hidden_states", "if", "in", "index", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_module", "layers", "make_default_2d_attention_mask", "mask_kwargs", "must", "nn", "norm", "normalizer", "not", "num_hidden_layers", "of", "one", "or", "or_mask_function", "pad_token_id", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "self_attention_cache", "self_attn_mask_mapping", "shape", "sliding_attention", "specify", "super", "tensor", "torch", "training", "unsqueeze", "use_cache"], "t5gemma/modeling_t5gemma.py:T5GemmaModel": ["BaseModelOutput", "BoolTensor", "EncoderDecoderCache", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelEncoderModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "TransformersKwargs", "Unpack", "Use", "ValueError", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_position_ids", "def", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get", "get_encoder", "get_input_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "instead", "is", "is_encoder_decoder", "kwargs", "last_hidden_state", "modeling", "new_embeddings", "not", "only", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "raise", "return", "self", "set_input_embeddings", "super", "support", "torch", "use_cache"], "t5gemma/modeling_t5gemma.py:T5GemmaEncoderModel": ["BaseModelOutput", "FloatTensor", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "Use", "ValueError", "__init__", "attention_mask", "auto_docstring", "can_return_tuple", "class", "config", "def", "encoder", "encoder_outputs", "forward", "get_input_embeddings", "if", "input_ids", "inputs_embeds", "instead", "is_encoder_decoder", "kwargs", "model", "new_embeddings", "only", "position_ids", "post_init", "raise", "return", "self", "set_input_embeddings", "super", "supports", "torch"], "t5gemma/modeling_t5gemma.py:T5GemmaForConditionalGeneration": ["BaseModelOutput", "BoolTensor", "EncoderDecoderCache", "FloatTensor", "ForMaskedLM", "GenerationMixin", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Seq2SeqModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_pp_plan", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "_tp_plan", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "cross_attentions", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_position_ids", "def", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "final_logit_softcapping", "forward", "get_decoder", "get_encoder", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "loss_type", "model", "new_embeddings", "not", "out_proj", "past_key_values", "position_ids", "post_init", "prepare_decoder_input_ids_from_labels", "r", "return", "self", "set_output_embeddings", "slice", "slice_indices", "super", "tanh", "tie_word_embeddings", "torch", "tuple", "use_cache", "vocab_size", "weight"], "t5gemma/modeling_t5gemma.py:T5GemmaForSequenceClassification": ["BaseModelOutput", "Cannot", "False", "FloatTensor", "If", "LongTensor", "ModelClassificationHead", "ModelConfig", "ModelEncoderModel", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Please", "Results", "Seq2SeqModelOutput", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_shift_right", "and", "arange", "are", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bool", "can_return_tuple", "cannot", "clamp", "class", "classifier_dropout", "classifier_dropout_rate", "config", "conjunction", "currently", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_position_ids", "def", "defined", "detect", "device", "dtype", "either", "elif", "else", "embeddings", "encoder", "encoder_outputs", "f", "for", "forward", "get_input_embeddings", "getattr", "handle", "hidden_size", "hidden_states", "if", "in", "input", "input_ids", "inputs_embeds", "int32", "is", "is_encoder_decoder", "kwargs", "labels", "last_hidden_state", "last_non_pad_token", "logger", "logits", "loss", "loss_function", "max", "may", "mode", "model", "no", "non_pad_mask", "not", "num_labels", "or", "outputs", "pad_token_id", "padding", "pass", "passed", "pooled_logits", "position_ids", "post_init", "r", "raise", "return", "score", "self", "set_input_embeddings", "shape", "sizes", "super", "supported", "to", "token", "token_indices", "tokens", "torch", "unexpected", "use_cache", "using", "value", "warning_once", "will", "with"], "t5gemma/modeling_t5gemma.py:T5GemmaForTokenClassification": ["BaseModelOutput", "False", "FloatTensor", "If", "LongTensor", "ModelClassificationHead", "ModelConfig", "ModelEncoderModel", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Please", "Seq2SeqModelOutput", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_shift_right", "and", "are", "attention_mask", "attentions", "auto_docstring", "be", "bool", "can_return_tuple", "cannot", "class", "classifier_dropout", "classifier_dropout_rate", "config", "currently", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_position_ids", "def", "either", "else", "embeddings", "encoder", "encoder_outputs", "f", "for", "forward", "get_input_embeddings", "getattr", "hidden_size", "hidden_states", "if", "in", "input", "input_ids", "inputs_embeds", "is", "is_encoder_decoder", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "mode", "model", "no", "not", "num_labels", "or", "outputs", "pass", "passed", "position_ids", "post_init", "r", "raise", "return", "score", "self", "set_input_embeddings", "super", "supported", "torch", "use_cache", "value"], "speech_encoder_decoder/modeling_speech_encoder_decoder.py:shift_tokens_right": ["Make", "Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "attribute", "clone", "configuration", "decoder_start_token_id", "def", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "of", "pad_token_id", "raise", "return", "s", "set", "shape", "sure", "the", "to", "torch"], "speech_encoder_decoder/modeling_speech_encoder_decoder.py:SpeechEncoderDecoderModel": ["AutoConfig", "AutoModel", "AutoModelForCausalLM", "BaseModelOutput", "BoolTensor", "Cache", "Config", "Cross", "CrossEntropyLoss", "Decoder", "Either", "False", "FloatTensor", "GenerationMixin", "Got", "Head", "If", "In", "Initializing", "LM", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "None", "NotImplementedError", "Optional", "Please", "PreTrainedModel", "PretrainedConfig", "Resizing", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "ValueError", "You", "__class__", "__init__", "_attn_implementation", "_get_feature_vector_attention_mask", "_supports_flash_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "a", "add_cross_attention", "added", "allows", "an", "and", "architecture", "are", "args", "argument", "as", "at", "attention", "attention_mask", "attentions", "attributes", "auto_docstring", "base_model_prefix", "be", "bool", "both", "by", "cannot", "causal", "class", "classmethod", "cls", "config", "config_class", "configuration", "cross", "cross_attention_hidden_size", "cross_attentions", "decoder", "decoder_", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_model", "decoder_outputs", "decoder_pretrained_model_name_or_path", "decoder_start_token_id", "def", "defined", "del", "directly", "disabled", "do", "either", "elif", "else", "embedding", "enc_to_dec_proj", "encoder", "encoder_", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_model", "encoder_output_dim", "encoder_outputs", "encoder_pretrained_model_name_or_path", "equal", "f", "for", "forward", "freeze_feature_encoder", "from", "from_config", "from_encoder_decoder_configs", "from_encoder_decoder_pretrained", "from_pretrained", "get_encoder", "get_input_embeddings", "get_output_embeddings", "getattr", "has", "have", "hidden_size", "hidden_states", "if", "in", "info", "initialize", "initialized", "input_features", "input_ids", "input_values", "inputs", "inputs_embeds", "is", "is_decoder", "isinstance", "it", "items", "key", "kwargs", "kwargs_decoder", "kwargs_encoder", "labels", "layers", "len", "logger", "logits", "loss", "loss_fct", "main_input_name", "make", "mask", "methods", "model", "model_args", "new_embeddings", "nn", "not", "num_items_in_batch", "object", "of", "or", "order", "output_attentions", "output_hidden_size", "output_hidden_states", "overwritten", "pad_token_id", "pass", "passed", "past_key_values", "pop", "prepare_decoder_input_ids_from_labels", "provided", "r", "raise", "randomly", "reshape", "resize_token_embeddings", "respective", "return", "return_dict", "return_unused_kwargs", "s", "same", "self", "set", "set_output_embeddings", "shape", "shared", "shift_tokens_right", "should", "specified", "specify", "startswith", "str", "super", "supported", "supports_gradient_checkpointing", "sure", "that", "the", "tie_word_embeddings", "time", "to", "to_dict", "torch", "tuple", "type", "use", "use_cache", "use_return_dict", "value", "via", "vocab_size", "warning", "without", "wrapped"], "lightglue/modeling_lightglue.py:LightGlueKeypointMatchingOutput": ["FloatTensor", "IntTensor", "ModelKeypointMatchingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "keypoints", "loss", "mask", "matches", "matching_scores", "prune", "r", "torch", "tuple"], "lightglue/modeling_lightglue.py:LightGluePositionalEncoder": ["False", "Linear", "ModelConfig", "ModelPositionalEncoder", "Module", "Optional", "Tensor", "Union", "__init__", "bias", "bool", "class", "config", "cos", "cosines", "def", "descriptor_dim", "dim", "else", "embeddings", "forward", "if", "keypoints", "nn", "num_attention_heads", "output", "output_hidden_states", "projected_keypoints", "projector", "repeat_interleave", "return", "self", "sin", "sines", "super", "torch", "tuple"], "lightglue/modeling_lightglue.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "rot_x", "stack", "torch", "x", "x1", "x2"], "lightglue/modeling_lightglue.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "float", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "lightglue/modeling_lightglue.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "lightglue/modeling_lightglue.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "lightglue/modeling_lightglue.py:LightGlueAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "class", "config", "contiguous", "cos", "current_attention_mask", "current_states", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_cross_attention", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "version", "view"], "lightglue/modeling_lightglue.py:LightGlueMLP": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "True", "__init__", "activation_fn", "class", "config", "def", "elementwise_affine", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "layer_norm", "nn", "return", "self", "super", "torch"], "lightglue/modeling_lightglue.py:LightGlueTransformerLayer": ["False", "ModelAttention", "ModelConfig", "ModelMLP", "ModelTransformerLayer", "Module", "None", "Optional", "Tensor", "__init__", "all_attentions", "all_hidden_states", "attention_mask", "attention_output", "batch_size", "bool", "cat", "class", "config", "cross_attention", "cross_attention_hidden_states", "cross_attention_output", "cross_attentions", "cross_intermediate_states", "cross_mlp", "cross_output_states", "def", "descriptor_dim", "descriptors", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "flip", "forward", "if", "int", "intermediate_states", "is", "keypoints", "layer_idx", "nn", "not", "num_keypoints", "output_attentions", "output_hidden_states", "output_states", "position_embeddings", "reshape", "return", "self", "self_attention", "self_attention_descriptors", "self_attention_hidden_states", "self_attentions", "self_mlp", "shape", "super", "torch", "tuple"], "lightglue/modeling_lightglue.py:sigmoid_log_double_softmax": ["Model_log_double_softmax", "Tensor", "batch_size", "certainties", "contiguous", "def", "functional", "logModel", "log_softmax", "matchability0", "matchability1", "new_full", "nn", "num_keypoints_0", "num_keypoints_1", "return", "scores", "scores0", "scores1", "shape", "similarity", "squeeze", "torch", "transpose"], "lightglue/modeling_lightglue.py:LightGlueMatchAssignmentLayer": ["Linear", "ModelConfig", "ModelMatchAssignmentLayer", "Module", "None", "Tensor", "True", "__init__", "batch_size", "bias", "class", "config", "def", "descriptor_dim", "descriptors", "device", "dtype", "final_projection", "finfo", "forward", "functional", "get_matchability", "if", "is", "m_descriptors", "m_descriptors0", "m_descriptors1", "mask", "mask0", "mask1", "masked_fill", "matchability", "matchability_0", "matchability_1", "min", "nn", "not", "num_keypoints", "reshape", "return", "scores", "self", "shape", "sigmoid", "sigmoid_log_double_softmax", "similarity", "squeeze", "super", "tensor", "torch", "transpose", "unsqueeze"], "lightglue/modeling_lightglue.py:LightGlueTokenConfidenceLayer": ["Linear", "ModelConfig", "ModelTokenConfidenceLayer", "Module", "Tensor", "__init__", "class", "config", "def", "descriptor_dim", "descriptors", "detach", "forward", "functional", "nn", "return", "self", "sigmoid", "squeeze", "super", "token", "torch"], "lightglue/modeling_lightglue.py:LightGluePreTrainedModel": ["False", "Model", "ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "main_input_name", "pixel_values", "supports_gradient_checkpointing"], "lightglue/modeling_lightglue.py:get_matches_from_scores": ["Model_matches_from_scores", "None", "Tensor", "_", "arange", "batch_size", "def", "device", "exp", "float", "gather", "indices", "indices0", "indices1", "matches", "matches0", "matches1", "matching_scores", "matching_scores0", "matching_scores1", "max", "max0", "max1", "mutual0", "mutual1", "new_tensor", "reshape", "return", "scores", "shape", "stack", "threshold", "torch", "transpose", "tuple", "valid0", "valid1", "values", "where", "zero"], "lightglue/modeling_lightglue.py:normalize_keypoints": ["Model_keypoints", "None", "Tensor", "def", "device", "dtype", "height", "int", "keypoints", "max", "return", "scale", "shift", "size", "tensor", "torch", "values", "width"], "lightglue/modeling_lightglue.py:LightGlueForKeypointMatching": ["AutoModelForKeypointDetection", "False", "FloatTensor", "Identity", "Input", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForKeypointMatching", "ModelKeypointMatchingOutput", "ModelMatchAssignmentLayer", "ModelPositionalEncoder", "ModelPreTrainedModel", "ModelTokenConfidenceLayer", "ModelTransformerLayer", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_concat_early_stopped_outputs", "_do_final_keypoint_pruning", "_do_layer_keypoint_pruning", "_get_confidence_threshold", "_get_early_stopped_image_pairs", "_get_keypoint_matching", "_get_pruning_mask", "_keypoint_processing", "_match_image_pair", "_matches", "_matching_scores", "a", "absolute_keypoints", "all", "all_attentions", "all_hidden_states", "and", "any", "arange", "attention", "attention_mask", "attentions", "auto_docstring", "batch_first", "batch_size", "be", "bias", "bool", "break", "can_return_tuple", "channels", "clamp", "class", "clip", "clone", "confidences", "config", "contiguous", "def", "depth_confidence", "descriptor_decoder_dim", "descriptor_dim", "descriptors", "descriptors_matchability", "detach", "device", "dim", "do_early_stop", "do_keypoint_pruning", "dtype", "early_stopped_image_indices", "early_stopped_matches", "early_stopped_matching_scores", "early_stopped_pairs", "early_stops", "early_stops_indices", "else", "exp", "expand", "extend", "extended_attention_mask", "filter_threshold", "final_pruned_keypoints_indices", "final_pruned_keypoints_iterations", "float", "for", "forward", "from_config", "full", "gather", "get_extended_attention_mask", "get_matchability", "get_matches_from_scores", "height", "hidden_states", "i", "ids", "if", "image_indices", "in", "indices", "indices0", "indices1", "initial_num_keypoints", "input_projection", "input_shape", "int", "is", "keep", "keypoint_1", "keypoint_confidences", "keypoint_detections", "keypoint_detector", "keypoint_detector_config", "keypoint_detector_descriptor_dim", "keypoint_encoding_output", "keypoints", "keypoints_0", "labels", "layer_idx", "layer_index", "layer_output", "list", "loss", "mask", "masked_fill", "match_assignment_layers", "matches", "matches0", "matches1", "matching_scores", "matching_scores0", "matching_scores1", "min", "must", "ndim", "new_full", "new_zeros", "nn", "no", "normalize_keypoints", "not", "np", "num_channels", "num_hidden_layers", "num_keypoints", "num_layers", "num_points", "num_points_per_pair", "of", "ones", "ones_like", "or", "order_indices", "output_attentions", "output_hidden_states", "pad_sequence", "padding_value", "pixel_values", "positional_encoder", "post_init", "projected_descriptors", "provided", "prune", "prune_output", "pruned_descriptors", "pruned_indices", "pruned_keypoints", "pruned_keypoints_0", "pruned_keypoints_1", "pruned_keypoints_indices", "pruned_keypoints_iterations", "pruned_keypoints_mask", "pruned_mask", "pruned_tensor", "raise", "range", "ratio_confident", "repeat_interleave", "reshape", "return", "scores", "self", "shape", "should", "size", "stack", "sum", "super", "t", "tensor", "threshold", "to", "token_confidence", "torch", "trainable", "transformer_layers", "trust_remote_code", "tuple", "where", "width", "width_confidence", "zeros", "zip"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch", "video_hidden_states"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple", "video_hidden_states"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoPooler": ["AvgPool2d", "Conv2d", "Has", "MaxPool2d", "ModelPooler", "Module", "Unknown", "ValueError", "_", "__init__", "average", "batch_size", "be", "class", "config", "contiguous", "conv", "def", "dim", "elif", "else", "f", "flatten", "forward", "getattr", "hidden_size", "if", "image_features", "image_features_spatial", "image_features_spatial_pool", "image_size", "in_channels", "int", "kernel_size", "math", "max", "mode", "nn", "of", "one", "ori_height", "ori_width", "out_channels", "patch_size", "permute", "pool", "pooling", "raise", "return", "self", "shape", "spatial_pool_mode", "spatial_pool_out_channels", "spatial_pool_stride", "sqrt", "stride", "super", "to", "transpose", "view", "vision_config"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoMultiModalProjector": ["ACT2FN", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "bias", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "image_features", "int", "isinstance", "len", "linear_1", "linear_2", "multimodal_projector_bias", "nn", "num_feature_layers", "projector_hidden_act", "return", "self", "super", "text_config", "vision_config", "vision_feature_layer"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoPreTrainedModel": ["Linear", "LlamaDecoderLayer", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "embed_std", "get_text_config", "getattr", "hidden_size", "if", "image_newline", "initializer_range", "is", "isinstance", "math", "mean", "module", "nn", "normal_", "not", "past_key_values", "self", "sqrt", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "llava_next_video/modeling_llava_next_video.py:get_anyres_image_grid_shape": ["Model_anyres_image_grid_shape", "Tensor", "TypeError", "a", "be", "def", "either", "f", "grid_pinpoints", "height", "if", "image_size", "invalid", "isinstance", "list", "lists", "ndarray", "not", "np", "of", "or", "patch_size", "raise", "return", "select_best_resolution", "should", "tensor", "tolist", "torch", "tuple", "tuples", "type", "valid", "width"], "llava_next_video/modeling_llava_next_video.py:image_size_to_num_patches": ["Model_size", "Model_size_to_num_patches", "Tensor", "TypeError", "a", "be", "best_resolution", "def", "f", "for", "grid_pinpoints", "height", "i", "if", "in", "int", "invalid", "isinstance", "j", "list", "lists", "ndarray", "not", "np", "num_patches", "of", "or", "patch_size", "raise", "range", "return", "select_best_resolution", "should", "tolist", "torch", "tuple", "tuples", "type", "value", "width", "with"], "llava_next_video/modeling_llava_next_video.py:unpad_image": ["Model_image", "Modelded_tensor", "Tensor", "TypeError", "be", "current_aspect_ratio", "current_height", "current_width", "def", "either", "else", "f", "if", "image_size", "int", "invalid", "isinstance", "list", "ndarray", "new_height", "new_width", "not", "np", "or", "original_aspect_ratio", "original_height", "original_size", "original_width", "padding", "raise", "return", "round", "scale_factor", "shape", "should", "tensor", "tolist", "torch", "tuple", "type", "valid"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoModel": ["AutoModel", "CLS", "Cache", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "Unpack", "ValueError", "Videos", "You", "_", "__init__", "_checkpoint_conversion_mapping", "_pixel_values_list", "a", "all", "and", "append", "attention_mask", "attentions", "auto_docstring", "base_image_feature", "batch_size", "be", "bool", "cache_position", "can_return_tuple", "cat", "channels", "class", "config", "contiguous", "decoder", "def", "default", "device", "dim", "dimensions", "do", "does", "dtype", "elif", "else", "embed_std", "encoder", "enumerate", "exactly", "expand", "expand_as", "expect", "f", "feature", "feature_lens", "features", "flatten", "for", "forward", "frames", "from_config", "get_anyres_image_grid_shape", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_video_features", "grid_pinpoints", "have", "height", "hidden_size", "hidden_states", "hs_pool", "if", "image", "image_feature", "image_features", "image_grid_pinpoints", "image_hidden_states", "image_idx", "image_newline", "image_num_patches", "image_size", "image_size_to_num_patches", "image_sizes", "image_token_id", "imsize", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "language_model", "last_hidden_state", "layer_idx", "line", "list", "logger", "long", "masked_scatter", "match", "math", "may", "model", "multi_modal_projector", "must", "n_image_tokens", "n_video_tokens", "new_image_features", "nn", "not", "np", "num_patch", "num_patch_height", "num_patch_width", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pack_image_features", "pad_token_id", "past_key_values", "patch", "patch_size", "permute", "pix_val", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prod", "provided", "r", "raise", "randn", "reshape", "return", "return_dict", "selected_image_feature", "selected_video_features", "self", "set_decoder", "set_input_embeddings", "shape", "size", "special_image_mask", "special_video_mask", "specify", "split", "sqrt", "str", "sum", "super", "tensor", "text_config", "that", "the", "to", "tokens", "torch", "transpose", "tuple", "unpad_image", "unsqueeze", "up", "use_cache", "use_return_dict", "using", "value", "video_feature_lens", "video_features", "video_hidden_states", "video_token_id", "view", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_resampler", "vision_tower", "visual", "vocab_size", "warning_once", "width", "with", "zip"], "llava_next_video/modeling_llava_next_video.py:LlavaNextVideoForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_prepare_4d_causal_attention_mask_with_cache_position", "_tied_weights_keys", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "causal_mask", "class", "clone", "config", "decoder", "def", "device", "diagonal", "dim", "dtype", "else", "expand", "fill_value", "finfo", "forward", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_video_features", "hidden_size", "hidden_states", "if", "image_features", "image_hidden_states", "image_newline", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "mask_length", "masked_fill", "min", "min_dtype", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "pack_image_features", "padding_mask", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "reshape", "return", "return_dict", "self", "sequence_length", "set_decoder", "set_input_embeddings", "shape", "slice", "slice_indices", "staticmethod", "str", "super", "target_length", "text_config", "to", "torch", "triu", "tuple", "use_cache", "use_return_dict", "value", "video_hidden_states", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2GenerationOutput": ["FloatTensor", "IntTensor", "ModelGenerationOutput", "ModelOutput", "None", "Optional", "class", "r", "sequences", "torch", "tuple", "unit_sequences", "waveform", "waveform_lengths"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2TextToUnitDecoderOutput": ["FloatTensor", "ModelOutput", "ModelTextToUnitDecoderOutput", "None", "Optional", "Tensor", "attentions", "class", "hidden_states", "last_hidden_state", "padding_mask", "r", "torch", "tuple"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2TextToUnitOutput": ["FloatTensor", "ModelOutput", "ModelTextToUnitOutput", "None", "Optional", "Tensor", "class", "decoder_attentions", "decoder_hidden_states", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "loss", "padding_mask", "r", "torch", "tuple"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:_compute_new_attention_mask": ["Tensor", "_compute_new_attention_mask", "arange", "batch_size", "bool_mask", "def", "device", "expand", "hidden_states", "indices", "mask", "mask_seq_len", "masked_fill", "new_ones", "return", "seq_lens", "shape", "torch", "unsqueeze"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:format_speech_generation_kwargs": ["Model_speech_generation_kwargs", "def", "elif", "else", "for", "generation_config", "if", "in", "items", "key", "kwargs", "kwargs_speech", "kwargs_text", "len", "not", "return", "speech_", "startswith", "text_", "value"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelConformerFeatureProjection", "Module", "__init__", "class", "config", "def", "dropout", "dtype", "eps", "feature_projection_input_dim", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "speech_encoder_dropout", "super", "to", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelConformerFeedForward", "Module", "None", "__init__", "act_fn", "class", "config", "def", "dropout", "else", "forward", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "is", "isinstance", "nn", "not", "output_dense", "output_dropout", "return", "self", "speech_encoder_dropout", "speech_encoder_hidden_act", "speech_encoder_intermediate_size", "str", "super"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerConvolutionModule": ["ACT2FN", "Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelConformerConvolutionModule", "Module", "None", "SAME", "ValueError", "__init__", "a", "activation", "attention_mask", "be", "bias", "bool", "class", "config", "conv_depthwise_kernel_size", "def", "depthwise_conv", "depthwise_layer_norm", "dim", "dropout", "for", "forward", "functional", "glu", "groups", "hidden_size", "hidden_states", "if", "is", "kernel_size", "layer_norm", "masked_fill", "nn", "not", "number", "odd", "pad", "padding", "pointwise_conv1", "pointwise_conv2", "raise", "return", "self", "should", "speech_encoder_dropout", "speech_encoder_hidden_act", "stride", "super", "torch", "transpose", "unsqueeze"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerSelfAttention": ["Dropout", "Embedding", "False", "Linear", "ModelConformerSelfAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "arange", "attention_mask", "attn_output", "attn_weights", "batch_size", "bhld", "bhlr", "bool", "clamp", "class", "config", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "else", "forward", "head_size", "hidden_size", "hidden_states", "if", "is", "key", "key_length", "left_max_position_embeddings", "linear_k", "linear_out", "linear_q", "linear_v", "long", "lrd", "math", "matmul", "nn", "not", "num_heads", "num_positions", "output_attentions", "p", "position_embeddings_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_key_states", "query_length", "relative_key", "relative_position_attn_weights", "reshape", "return", "right_max_position_embeddings", "self", "sequence_length", "shape", "size", "softmax", "speech_encoder_attention_heads", "speech_encoder_dropout", "sqrt", "super", "to", "torch", "transpose", "tuple", "use_position_embeddings", "value", "value_states", "view"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelConformerConvolutionModule", "ModelConformerEncoderLayer", "ModelConformerFeedForward", "ModelConformerSelfAttention", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "conv_attention_mask", "conv_module", "def", "dropout", "embed_dim", "ffn1", "ffn1_layer_norm", "ffn2", "ffn2_layer_norm", "final_layer_norm", "forward", "hidden_size", "hidden_states", "nn", "output_attentions", "residual", "return", "self", "self_attn", "self_attn_dropout", "self_attn_layer_norm", "speech_encoder_dropout", "super", "torch"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelConformerEncoder", "ModelConformerEncoderLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "_apply_chunk_attention", "all_hidden_states", "all_self_attentions", "and", "arange", "attention_mask", "attentions", "bool", "chunk_indices", "chunk_mask", "clamp_", "class", "config", "conv_attention_mask", "def", "device", "div", "dropout", "dropout_probability", "dtype", "else", "end_indices", "enumerate", "eps", "expand", "finfo", "for", "forward", "full_like", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "indices", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layers", "long", "masked_fill", "max", "min", "nn", "not", "or", "output_attentions", "output_hidden_states", "rand", "range", "return", "return_dict", "self", "sequence_len", "shape", "skip_the_layer", "speech_encoder_chunk_size", "speech_encoder_dropout", "speech_encoder_layerdrop", "speech_encoder_layers", "speech_encoder_left_chunk_num", "start_indices", "super", "synced_gpus", "to", "torch", "training", "tuple", "unsqueeze", "v"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerAdapterLayer": ["Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelConformerAdapterLayer", "ModelConformerFeedForward", "ModelConformerSelfAttention", "Module", "None", "Optional", "Tensor", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_prepare_4d_attention_mask", "act_fn", "activation", "adaptor_dropout", "adaptor_kernel_size", "adaptor_stride", "attention_mask", "attn_weights", "bool", "class", "config", "def", "device", "dim", "dropout", "dtype", "embed_dim", "ffn", "ffn_layer_norm", "floor", "forward", "hidden_size", "hidden_states", "if", "int", "is", "kernel_size", "nn", "not", "output_attentions", "pad", "padding", "relu", "residual", "residual_conv", "residual_layer_norm", "return", "self", "self_attn", "self_attn_conv", "self_attn_dropout", "self_attn_layer_norm", "seq_lens", "size", "stride", "sub_sampled_lengths", "sum", "super", "to", "torch", "transpose", "use_position_embeddings"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ConformerAdapter": ["ModelConformerAdapter", "ModelConformerAdapterLayer", "Module", "ModuleList", "_", "__init__", "attention_mask", "class", "config", "def", "for", "forward", "hidden_states", "in", "layer", "layers", "nn", "num_adapter_layers", "range", "return", "self", "super"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2SinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "contiguous", "cos", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "detach", "device", "dim", "dtype", "else", "emb", "emb_weights", "embedding_dim", "exp", "expand", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sequence_length", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2Attention": ["Cache", "EncoderDecoderCache", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "__class__", "__init__", "__name__", "a", "and", "attention_mask", "attention_scores", "attn_output", "attn_weights", "batch_size", "be", "bias", "bool", "by", "cache_position", "caching", "call", "class", "config", "context_states", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "during", "else", "embed_dim", "encoder_hidden_states", "errors", "f", "float", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "keys", "layer_idx", "layers", "lead", "logger", "make", "matmul", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "permute", "provide", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "seq_length", "shape", "softmax", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "type_as", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2FeedForwardNetwork": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelFeedForwardNetwork", "Module", "Tensor", "__init__", "act", "activation_dropout", "activation_function", "and", "class", "config", "def", "dropout", "dtype", "fc1", "fc2", "ffn_dim", "forward", "hidden_size", "hidden_states", "if", "int", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "uint8", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2EncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelFeedForwardNetwork", "None", "Tensor", "__init__", "activation_dropout", "attention_dropout", "attention_mask", "attn_dropout", "attn_weights", "bool", "class", "config", "def", "dropout", "else", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "ffn", "ffn_dim", "ffn_dropout", "ffn_layer_norm", "forward", "hidden_size", "hidden_states", "if", "is", "nn", "num_heads", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2DecoderLayer": ["ACT2FN", "Cache", "Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFeedForwardNetwork", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_dropout", "bool", "cache_position", "class", "config", "cross_attention", "cross_attention_layer_norm", "cross_attn_weights", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_hidden_states", "ffn", "ffn_dim", "ffn_dropout", "ffn_layer_norm", "forward", "hidden_size", "hidden_states", "if", "is", "is_decoder", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "use_cache", "version"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2TextToUnitDecoderLayer": ["ACT2FN", "Conv1d", "Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelTextToUnitDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "conv1", "conv2", "conv_dropout", "conv_layer_norm", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "else", "embed_dim", "forward", "hidden_size", "hidden_states", "if", "is", "is_decoder", "kernel_size", "masked_fill", "nn", "not", "num_heads", "output_attentions", "padding", "padding_mask", "residual", "return", "same", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "stride", "super", "torch", "transpose", "unsqueeze"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2PreTrainedModel": ["Conv1d", "ConvTranspose1d", "Embedding", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelConformerEncoderLayer", "ModelConformerFeatureProjection", "ModelConformerSelfAttention", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextToUnitDecoder", "ModelTextToUnitDecoderLayer", "Module", "None", "PreTrainedModel", "True", "ValueError", "_", "_compute_sub_sample_lengths_from_attention_mask", "_count_character_length_in_subword", "_get_char_input_ids", "_hard_upsample", "_indices_to_subwords", "_init_weights", "_no_split_modules", "a", "adaptor_kernel_size", "adaptor_stride", "and", "append", "attention_mask", "b", "base_model_prefix", "batch_first", "batch_id", "batch_size", "bias", "break", "ch", "char_count_per_id", "char_ids", "char_len", "char_seq_len", "char_seqs", "char_to_id", "class", "config", "data", "def", "dim", "duration", "durations", "elif", "else", "enumerate", "fill_", "floor", "for", "generation_config", "get", "groups", "hasattr", "hidden_state", "hidden_states", "i", "id_to_text", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_ids", "int", "is", "is_next_start_with_space", "is_punc", "isalpha", "isinstance", "isnumeric", "item", "k", "kaiming_normal_", "kernel_size", "len", "list", "logger", "math", "max", "max_len", "mean", "merge_space_with_prev_subword", "module", "ne", "new_zeros", "nn", "normal_", "not", "pad", "pad_sequence", "pad_token_id", "padding_idx", "pos_bias_u", "pos_bias_v", "pos_emb_alpha", "pos_emb_alpha_char", "projection", "raise", "range", "repeat_interleave", "return", "rnn", "self", "seq_lens", "sequence_len", "shape", "size", "space", "sqrt", "std", "str", "stride", "subword", "subword_idx", "subword_indices", "subword_lens", "subwords", "subwords_batch", "sum", "supports_gradient_checkpointing", "tensor", "to", "torch", "total", "training", "uniform_", "unk_token_id", "utils", "view", "warning_once", "weight", "xavier_uniform_", "zero_", "zip"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2SpeechEncoder": ["LayerNorm", "ModelConfig", "ModelConformerAdapter", "ModelConformerEncoder", "ModelConformerFeatureProjection", "ModelConformerFeedForward", "ModelPreTrainedModel", "ModelSpeechEncoder", "None", "Optional", "Tensor", "Union", "ValueError", "Wav2Vec2BaseModelOutput", "__init__", "act_fn", "adapter", "add_adapter", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "else", "encoder", "encoder_outputs", "expanded_hidden_states", "feature_projection", "forward", "hidden_size", "hidden_states", "if", "inner_layer_norm", "input_features", "intermediate_ffn", "is", "kwargs", "last_hidden_state", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "raise", "relu", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2Encoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Pass", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "and", "append", "at", "attention_mask", "attentions", "bool", "both", "cannot", "class", "config", "def", "device", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder", "encoder_attention_heads", "encoder_ffn_dim", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "for", "forward", "functional", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "input", "input_ids", "input_shape", "inputs_embeds", "instead", "is", "is_t2u_encoder", "kwargs", "last_hidden_state", "layer_norm", "layer_outputs", "layerdrop", "layers", "math", "max_position_embeddings", "max_source_positions", "model", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "shape", "specify", "sqrt", "super", "text_to_units", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2Decoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "append", "at", "attention_mask", "attentions", "auto_docstring", "be", "bool", "both", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "decoder_attention_heads", "decoder_ffn_dim", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "embedding_dim", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "logger", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "num_embeddings", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2TextToUnitDecoder": ["Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModelTextToUnitDecoder", "ModelTextToUnitDecoderLayer", "ModelTextToUnitDecoderOutput", "ModelVariancePredictor", "ModuleList", "None", "Optional", "Parameter", "Union", "_", "__init__", "_compute_new_attention_mask", "_hard_upsample", "_prepare_4d_attention_mask", "all_hidden_states", "all_self_attns", "append", "attention_mask", "attentions", "bool", "char_count_per_id", "char_hidden_states", "char_input_ids", "char_padding_mask", "char_positions", "char_vocab_size", "clamp", "class", "config", "continue", "decoder_attention_heads", "decoder_ffn_dim", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "dur_out", "duration_predictor", "else", "embed_char", "embed_char_positions", "embed_positions", "embed_scale", "embed_tokens", "embedding_dim", "encoder_hidden_states", "enumerate", "expm1", "for", "forward", "functional", "gradient_checkpointing", "hidden_size", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_norm", "layer_outputs", "layerdrop", "layers", "log_dur_pred", "long", "masked_fill", "math", "max_position_embeddings", "max_target_positions", "min", "nn", "not", "num_embeddings", "ones", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "padding_mask", "pos_emb_alpha", "pos_emb_alpha_char", "positions", "post_init", "r", "rand", "range", "return", "return_dict", "round", "scale_embedding", "self", "sqrt", "sum", "super", "torch", "training", "tuple", "use_return_dict", "v", "variance_pred_dropout", "variance_predictor_embed_dim", "variance_predictor_hidden_dim", "variance_predictor_kernel_size", "vocab_size", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2TextToUnitModel": ["BaseModelOutput", "Embedding", "FloatTensor", "LongTensor", "ModelConfig", "ModelEncoder", "ModelPreTrainedModel", "ModelTextToUnitDecoder", "ModelTextToUnitModel", "ModelTextToUnitOutput", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "True", "Union", "__init__", "and", "attention_mask", "attentions", "bool", "char_count_per_id", "char_input_ids", "class", "config", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_outputs", "def", "elif", "else", "embed_tokens_decoder", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_t2u_encoder", "isinstance", "last_hidden_state", "len", "nn", "not", "output_attentions", "output_hidden_states", "padding_mask", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2TextToUnitForConditionalGeneration": ["CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelPreTrainedModel", "ModelTextToUnitForConditionalGeneration", "ModelTextToUnitModel", "ModelTextToUnitOutput", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "True", "Union", "__init__", "__setattr__", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "attention_mask", "auto_docstring", "bias", "bool", "char_count_per_id", "char_input_ids", "class", "config", "copy", "decoder", "decoder_attentions", "decoder_hidden_states", "deepcopy", "def", "device", "else", "embed_tokens", "embed_tokens_decoder", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "get_decoder", "get_encoder", "get_input_embeddings", "get_output_embeddings", "getattr", "hidden_size", "if", "in", "input_ids", "inputs_embeds", "is", "items", "kwargs", "labels", "last_hidden_state", "lm_head", "lm_logits", "loss", "loss_fct", "masked_lm_loss", "model", "nn", "not", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "padding_mask", "param", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "speech_encoder", "startswith", "super", "t2u_", "t2u_vocab_size", "text_decoder", "text_encoder", "tie_word_embeddings", "to", "to_dict", "torch", "tuple", "use_return_dict", "val", "value", "view", "vocab_size", "vocoder", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:HifiGanResidualBlock": ["Conv1d", "ModelGanResidualBlock", "Module", "ModuleList", "_", "__init__", "apply_weight_norm", "channels", "class", "conv1", "conv2", "convs1", "convs2", "def", "dilation", "for", "forward", "functional", "get_padding", "hasattr", "hidden_states", "i", "if", "in", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "nn", "padding", "parametrizations", "range", "remove_weight_norm", "residual", "return", "self", "stride", "super", "utils", "weight_norm", "zip"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2VariancePredictor": ["Conv1d", "Dropout", "LayerNorm", "Linear", "ModelVariancePredictor", "Module", "None", "Optional", "ReLU", "Tensor", "__init__", "activation_function", "bool", "class", "conv1", "conv2", "def", "dim", "dropout_module", "embed_dim", "forward", "hidden_dim", "hidden_states", "if", "is", "kernel_size", "ln1", "ln2", "masked_fill", "nn", "not", "p", "padding", "padding_mask", "proj", "return", "same", "self", "squeeze", "super", "transpose", "unsqueeze", "var_pred_dropout"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2HifiGan": ["Conv1d", "ConvTranspose1d", "FloatTensor", "HifiGanResidualBlock", "ModelConfig", "ModelHifiGan", "Module", "ModuleList", "__init__", "append", "channels", "class", "config", "conv_post", "conv_pre", "def", "dilation", "enumerate", "for", "forward", "functional", "hidden_states", "i", "in", "input_embeds", "j", "kernel_size", "lang_embed_dim", "leaky_relu", "leaky_relu_slope", "len", "model_in_dim", "nn", "num_kernels", "num_upsamples", "padding", "r", "range", "res_state", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "self", "spkr_embed_dim", "squeeze", "stride", "super", "tanh", "torch", "unit_embed_dim", "upsample_initial_channel", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "waveform", "zip"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2CodeHifiGan": ["Conv1d", "ConvTranspose1d", "Embedding", "LayerNorm", "Linear", "LongTensor", "ModelCodeHifiGan", "ModelConfig", "ModelHifiGan", "ModelVariancePredictor", "Module", "None", "PreTrainedModel", "Tensor", "True", "Union", "__init__", "_conv_out_length", "_get_dur_output_lengths", "_get_output_hifigan_lengths", "_init_weights", "_no_split_modules", "_transpose_conv_out_length", "and", "apply_weight_norm", "batch_first", "bias", "cat", "clamp", "class", "config", "conv_post", "conv_pre", "cumsum", "cumulative_dur_out", "data", "def", "dil", "dilation", "dim", "div", "dur_out", "dur_predictor", "duration", "elif", "else", "embed_dim", "enumerate", "expm1", "fill_", "floor", "for", "forward", "gather", "hasattr", "hidden_state", "hidden_states", "hifi_gan", "i", "if", "in", "index", "initializer_range", "input_embeds", "input_ids", "input_length", "input_lengths", "int", "is", "isinstance", "kernel_size", "lang", "lang_embed_dim", "lang_id", "language_embedding", "layer", "len", "lengths", "log_dur_pred", "logger", "long", "main_input_name", "mean", "min", "module", "nn", "normal_", "not", "pad", "pad_sequence", "pad_token_id", "padding_idx", "parametrizations", "post_init", "range", "remove_weight_norm", "repeat", "repeat_interleave", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "rnn", "round", "rounding_mode", "self", "shape", "size", "speaker_embedding", "speaker_id", "spkr", "spkr_embed_dim", "squeeze", "std", "stride", "sum", "super", "t2u_pad_token_id", "torch", "training", "transpose", "tuple", "unit_embed_dim", "unit_embedding", "unit_hifi_gan_vocab_size", "unit_lengths", "unsqueeze", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "utils", "var_pred_dropout", "variance_predictor_kernel_size", "view", "vocoder_num_langs", "vocoder_num_spkrs", "warning", "weight", "weight_norm", "zero_", "zip"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ForTextToText": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelForTextToText", "ModelPreTrainedModel", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "Union", "ValueError", "__", "__init__", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "changed", "class", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "device", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "forward", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "logits_processor", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pop", "post_init", "prefix_allowed_tokens_fn", "provided", "r", "raise", "replace", "return", "return_dict", "self", "set_input_embeddings", "shared", "shift_tokens_right", "since", "speech_encoder", "stopping_criteria", "super", "synced_gpus", "t2u_model", "tensor", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_tgt_lang_id", "tgt_lang", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "warning", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ForSpeechToText": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelDecoder", "ModelForSpeechToText", "ModelPreTrainedModel", "ModelSpeechEncoder", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "Union", "ValueError", "__", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "changed", "class", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "device", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "forward", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_features", "input_ids", "inputs", "inputs_embeds", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "logits_processor", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pop", "post_init", "prefix_allowed_tokens_fn", "provided", "r", "raise", "replace", "return", "return_dict", "self", "seq_lens", "set_input_embeddings", "shared", "shift_tokens_right", "since", "speech_encoder", "stopping_criteria", "sub_sampled_lengths", "super", "synced_gpus", "t2u_model", "tensor", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_tgt_lang_id", "tgt_lang", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "warning", "weight"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ForTextToSpeech": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "If", "It", "Linear", "LongTensor", "ModelCodeHifiGan", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelForTextToSpeech", "ModelForTextToText", "ModelGenerationOutput", "ModelPreTrainedModel", "ModelTextToUnitForConditionalGeneration", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "__", "__init__", "_compute_new_attention_mask", "_count_character_length_in_subword", "_get_char_input_ids", "_indices_to_subwords", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "and", "argmax", "argument", "as", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "cat", "changed", "char_count_per_id", "char_input_ids", "class", "clone", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "detach", "device", "dim", "do_sample", "doesn", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos_token_id", "f", "for", "format_speech_generation_kwargs", "forward", "functional", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "getattr", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "key", "kwargs", "kwargs_speech", "kwargs_text", "labels", "lang_code_to_id", "lang_id", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "main_input_name", "masked_fill", "masked_lm_loss", "method", "model", "multinomial", "must", "new_zeros", "nn", "no_grad", "not", "num_return_sequences", "num_samples", "or", "output", "output_attentions", "output_hidden_states", "output_scores", "output_unit_ids", "outputs", "pad_token_id", "pad_zero", "padding_mask", "past_key_values", "post_init", "probs", "provided", "r", "raise", "repeat_interleave", "replace", "replace_mask", "reshape", "return", "return_dict", "return_dict_in_generate", "return_intermediate_token_ids", "same", "self", "seq_lens", "sequences", "set_input_embeddings", "shape", "shared", "shift_tokens_right", "since", "softmax", "speaker_id", "specify", "speech", "speech_encoder", "str", "sum", "super", "t", "t2u_char_count_per_id", "t2u_char_input_ids", "t2u_eos_token_id", "t2u_input_embeds", "t2u_input_ids", "t2u_lang_code_to_id", "t2u_logits", "t2u_model", "t2u_model_attention_mask", "t2u_output", "t2u_pad_token_id", "t2u_subwords", "temperature", "tensor", "text", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_generation_output", "text_tgt_lang_id", "tgt_lang", "the", "tie_word_embeddings", "to", "torch", "translated", "tuple", "unit", "unit_ids", "unit_sequences", "use", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "vocoder_lang_code_to_id", "vocoder_offset", "vocoder_tgt_lang_id", "want", "warning", "waveform", "waveform_lengths", "weight", "where", "you"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2ForSpeechToSpeech": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "If", "It", "Linear", "LongTensor", "ModelCodeHifiGan", "ModelDecoder", "ModelForSpeechToSpeech", "ModelForSpeechToText", "ModelGenerationOutput", "ModelPreTrainedModel", "ModelSpeechEncoder", "ModelTextToUnitForConditionalGeneration", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "__", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_count_character_length_in_subword", "_get_char_input_ids", "_indices_to_subwords", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "and", "argmax", "argument", "as", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cat", "changed", "char_count_per_id", "char_input_ids", "class", "clone", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "detach", "device", "dim", "do_sample", "doesn", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos_token_id", "f", "for", "format_speech_generation_kwargs", "forward", "functional", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "getattr", "hidden_size", "hidden_states", "if", "in", "input_features", "input_ids", "inputs_embeds", "int", "is", "isinstance", "key", "kwargs", "kwargs_speech", "kwargs_text", "labels", "lang_code_to_id", "lang_id", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "main_input_name", "masked_fill", "masked_lm_loss", "method", "multinomial", "must", "new_zeros", "nn", "no_grad", "not", "num_return_sequences", "num_samples", "or", "output", "output_attentions", "output_hidden_states", "output_scores", "output_unit_ids", "outputs", "pad_token_id", "pad_zero", "padding_mask", "past_key_values", "post_init", "probs", "provided", "r", "raise", "repeat_interleave", "replace", "replace_mask", "reshape", "return", "return_dict", "return_dict_in_generate", "return_intermediate_token_ids", "same", "self", "seq_lens", "sequences", "set_input_embeddings", "shape", "shared", "shift_tokens_right", "since", "softmax", "speaker_id", "specify", "speech", "speech_encoder", "str", "sub_sampled_lengths", "sum", "super", "t", "t2u_char_count_per_id", "t2u_char_input_ids", "t2u_eos_token_id", "t2u_input_embeds", "t2u_input_ids", "t2u_lang_code_to_id", "t2u_logits", "t2u_model", "t2u_model_attention_mask", "t2u_output", "t2u_pad_token_id", "t2u_subwords", "temperature", "tensor", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_generation_output", "text_tgt_lang_id", "tgt_lang", "the", "tie_word_embeddings", "to", "torch", "translated", "tuple", "unit_ids", "unit_sequences", "use", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "vocoder_lang_code_to_id", "vocoder_offset", "vocoder_tgt_lang_id", "want", "warning", "waveform", "waveform_lengths", "weight", "where", "you"], "seamless_m4t_v2/modeling_seamless_m4t_v2.py:SeamlessM4Tv2Model": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "If", "It", "Linear", "LongTensor", "Make", "ModelCodeHifiGan", "ModelDecoder", "ModelEncoder", "ModelForSpeechToText", "ModelForTextToText", "ModelGenerationOutput", "ModelModel", "ModelPreTrainedModel", "ModelSpeechEncoder", "ModelTextToUnitForConditionalGeneration", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "__", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_count_character_length_in_subword", "_get_char_input_ids", "_indices_to_subwords", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "all", "and", "are", "argmax", "argument", "as", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "been", "bias", "bool", "both", "but", "calls", "cat", "changed", "char_count_per_id", "char_input_ids", "class", "clone", "config", "cross_attentions", "current_modality", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "depending", "detach", "device", "dim", "do_sample", "elif", "else", "embed_tokens", "empty", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos_token_id", "exclusive", "f", "for", "format_speech_generation_kwargs", "forward", "functional", "generate", "generate_speech", "generation_config", "get", "get_encoder", "get_input_embeddings", "getattr", "given", "has", "hidden_size", "hidden_states", "if", "ignored", "in", "input", "input_features", "input_ids", "inputs_embeds", "int", "is", "isinstance", "key", "keys_to_check", "kwargs", "kwargs_speech", "kwargs_text", "labels", "lang_code_to_id", "lang_id", "last_hidden_state", "least", "len", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "main_input_name", "masked_fill", "masked_lm_loss", "method", "modality", "multinomial", "must", "mutually", "new_zeros", "nn", "no_grad", "non", "not", "num_return_sequences", "num_samples", "of", "on", "one", "or", "output", "output_attentions", "output_hidden_states", "output_scores", "output_unit_ids", "outputs", "pad_token_id", "pad_zero", "padding_mask", "past_key_values", "post_init", "priority", "probs", "provided", "r", "raise", "repeat_interleave", "replace", "replace_mask", "reshape", "return", "return_dict", "return_dict_in_generate", "return_intermediate_token_ids", "same", "self", "seq_lens", "sequences", "set_input_embeddings", "set_modality", "shape", "shared", "shift_tokens_right", "since", "softmax", "speaker_id", "specify", "speech", "speech_encoder", "str", "sub_sampled_lengths", "sum", "super", "sure", "t2u_char_count_per_id", "t2u_char_input_ids", "t2u_eos_token_id", "t2u_input_embeds", "t2u_input_ids", "t2u_lang_code_to_id", "t2u_logits", "t2u_model", "t2u_model_attention_mask", "t2u_output", "t2u_pad_token_id", "t2u_subwords", "temperature", "tensor", "text", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_generation_output", "text_tgt_lang_id", "tgt_lang", "that", "the", "them", "through", "tie_word_embeddings", "to", "torch", "translated", "tuple", "unit_ids", "unit_sequences", "use", "use_cache", "use_return_dict", "used", "valid", "value", "view", "vocab_size", "vocoder", "vocoder_lang_code_to_id", "vocoder_offset", "vocoder_tgt_lang_id", "want", "warning", "waveform", "waveform_lengths", "weight", "where", "will", "you"], "convnext/modeling_convnext.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "convnext/modeling_convnext.py:ConvNextDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "convnext/modeling_convnext.py:ConvNextLayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "convnext/modeling_convnext.py:ConvNextEmbeddings": ["FloatTensor", "Make", "Model2d", "ModelEmbeddings", "ModelLayerNorm", "Module", "Tensor", "ValueError", "__init__", "channel", "channels_first", "class", "config", "configuration", "data_format", "def", "dimension", "embeddings", "eps", "forward", "hidden_sizes", "if", "in", "kernel_size", "layernorm", "match", "nn", "num_channels", "of", "one", "patch_embeddings", "patch_size", "pixel", "pixel_values", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "values", "with"], "convnext/modeling_convnext.py:ConvNextLayer": ["ACT2FN", "Identity", "Linear", "Model2d", "ModelDropPath", "ModelLayer", "ModelLayerNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "act", "class", "config", "def", "dim", "drop_path", "dwModel", "else", "eps", "features", "forward", "groups", "hidden_act", "if", "is", "kernel_size", "layer_scale_init_value", "layer_scale_parameter", "layernorm", "nn", "not", "ones", "padding", "permute", "pwModel1", "pwModel2", "requires_grad", "residual", "return", "self", "super", "torch"], "convnext/modeling_convnext.py:ConvNextStage": ["Model2d", "ModelLayer", "ModelLayerNorm", "ModelStage", "Module", "ModuleList", "None", "Tensor", "__init__", "channels_first", "class", "config", "data_format", "def", "depth", "dim", "downsampling_layer", "drop_path", "drop_path_rates", "else", "eps", "features", "for", "forward", "if", "in", "in_channels", "j", "kernel_size", "layer", "layers", "nn", "or", "out_channels", "range", "return", "self", "stride", "super", "torch"], "convnext/modeling_convnext.py:ConvNextEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_hidden_states", "append", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "drop_path_rate", "drop_path_rates", "else", "for", "forward", "hidden_sizes", "hidden_states", "i", "if", "in", "in_channels", "is", "last_hidden_state", "layer_module", "linspace", "nn", "not", "num_stages", "out_channels", "out_chs", "output_hidden_states", "prev_chs", "range", "return", "self", "split", "stage", "stages", "stride", "sum", "super", "tolist", "torch", "x"], "convnext/modeling_convnext.py:ConvNextPreTrainedModel": ["LayerNorm", "Linear", "Model", "Model2d", "ModelConfig", "ModelLayer", "ModelLayerNorm", "ModelPreTrainedModel", "None", "PreTrainedModel", "_can_record_outputs", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "layer_scale_init_value", "layer_scale_parameter", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "convnext/modeling_convnext.py:ConvNextModel": ["BaseModelOutputWithNoAttention", "BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "ValueError", "You", "__init__", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_sizes", "hidden_states", "if", "is", "last_hidden_state", "layer_norm_eps", "layernorm", "mean", "nn", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_init", "raise", "return", "self", "specify", "super", "to", "torch"], "convnext/modeling_convnext.py:ConvNextForImageClassification": ["BaseModelOutputWithPoolingAndNoAttention", "False", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "__init__", "accepts_loss_kwargs", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "kwargs", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "outputs", "pixel_values", "pooled_logits", "pooled_output", "pooler_output", "post_init", "r", "return", "self", "super", "torch"], "convnext/modeling_convnext.py:ConvNextBackbone": ["BackboneMixin", "BackboneOutput", "BaseModelOutputWithPoolingAndNoAttention", "False", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelLayerNorm", "ModelPreTrainedModel", "ModuleDict", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "_out_features", "append", "auto_docstring", "bool", "can_return_tuple", "channels", "channels_first", "class", "config", "data_format", "def", "else", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_state", "hidden_states", "hidden_states_norms", "if", "in", "is", "nn", "num_channels", "num_features", "out_features", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "self", "stage", "stage_names", "super", "torch", "tuple", "zip"], "oneformer/modeling_oneformer.py:_get_clones": ["ModuleList", "N", "_get_clones", "copy", "deepcopy", "def", "for", "i", "in", "module", "nn", "range", "return"], "oneformer/modeling_oneformer.py:multi_scale_deformable_attention": ["False", "Model_scale_deformable_attention", "Tensor", "Union", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "contiguous", "def", "dim", "enumerate", "flatten", "for", "functional", "grid_sample", "height", "hidden_dim", "in", "level_id", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "view", "width", "zeros"], "oneformer/modeling_oneformer.py:dice_loss": ["Model_loss", "Tensor", "def", "denominator", "flatten", "inputs", "int", "labels", "loss", "num_masks", "numerator", "probs", "r", "return", "sigmoid", "sum"], "oneformer/modeling_oneformer.py:sigmoid_cross_entropy_loss": ["BCEWithLogitsLoss", "Model_cross_entropy_loss", "Tensor", "criterion", "cross_entropy_loss", "def", "inputs", "int", "labels", "loss", "mean", "nn", "none", "num_masks", "r", "reduction", "return", "sum", "torch"], "oneformer/modeling_oneformer.py:pair_wise_dice_loss": ["Model_wise_dice_loss", "None", "T", "Tensor", "def", "denominator", "flatten", "inputs", "labels", "loss", "matmul", "numerator", "return", "sigmoid", "sum", "torch"], "oneformer/modeling_oneformer.py:pair_wise_sigmoid_cross_entropy_loss": ["BCEWithLogitsLoss", "Model_wise_sigmoid_cross_entropy_loss", "T", "Tensor", "criterion", "cross_entropy_loss_neg", "cross_entropy_loss_pos", "def", "height_and_width", "inputs", "labels", "loss", "loss_neg", "loss_pos", "matmul", "nn", "none", "ones_like", "r", "reduction", "return", "shape", "torch", "zeros_like"], "oneformer/modeling_oneformer.py:sample_point": ["False", "Model_point", "Tensor", "True", "add_dim", "def", "dim", "functional", "grid_Model", "if", "input_features", "kwargs", "nn", "point_coordinates", "point_features", "return", "squeeze", "torch", "unsqueeze"], "oneformer/modeling_oneformer.py:OneFormerHungarianMatcher": ["All", "False", "ModelHungarianMatcher", "Module", "NModel", "Tensor", "ValueError", "__init__", "align_corners", "and", "append", "array", "as_tensor", "assigned_indices", "autocast", "be", "can", "class", "class_labels", "class_queries_logits", "cost_class", "cost_dice", "cost_mask", "cost_matrix", "costs", "cpu", "cuda", "def", "device", "device_type", "dtype", "enabled", "float", "for", "forward", "i", "if", "in", "indices", "int", "int64", "j", "labels", "linear_sum_assignment", "list", "mask_labels", "masks_queries_logits", "matched_indices", "nn", "no_grad", "np", "num_points", "num_queries", "pair_wise_dice_loss", "pair_wise_sigmoid_cross_entropy_loss", "point_coords", "pred_mask", "pred_probs", "preds_masks", "preds_probs", "raise", "rand", "repeat", "reshape", "return", "sample_point", "self", "shape", "softmax", "squeeze", "super", "t", "target_mask", "to", "torch", "tuple", "with", "zip"], "oneformer/modeling_oneformer.py:OneFormerLoss": ["CrossEntropyLoss", "False", "ModelHungarianMatcher", "ModelLoss", "Models", "Module", "NModel", "Optional", "Parameter", "PartialState", "Tensor", "True", "_", "__init__", "_get_predictions_permutation_indices", "_get_targets_permutation_indices", "_max_by_axis", "_pad_images_to_max_in_batch", "_shared_state", "abs", "align_corners", "arange", "array", "as_tensor", "aux_outputs", "auxiliary_predictions", "b", "batch_indices", "batch_shape", "batch_size", "bool", "calculate_contrastive_loss", "calculate_uncertainty", "cat", "clamp", "class", "class_labels", "class_queries_logits", "classes", "contrastive_queries_logits", "contrastive_temperature", "copy_", "criterion", "cross_entropy", "def", "del", "device", "dice_loss", "dict", "dim", "dtype", "empty_weight", "enumerate", "eos_coef", "exp", "f", "fill_value", "flatten", "float", "for", "forward", "full", "full_like", "functional", "get_num_masks", "h", "i", "idx", "if", "image_queries", "importance_sample_ratio", "in", "index", "indices", "int", "int64", "is", "is_accelerate_available", "item", "items", "j", "k", "key", "len", "list", "log", "logit_scale", "logits", "logits_per_img", "logits_per_text", "long", "loss_ce", "loss_contrastive", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_img", "loss_labels", "loss_mask", "loss_masks", "loss_text", "losses", "mask_labels", "masks_queries_logits", "matcher", "matmul", "max", "max_size", "maxes", "min", "nn", "no_grad", "normalize", "not", "np", "num_boxes", "num_classes", "num_masks", "num_points", "num_points_sampled", "num_processes", "num_queries", "num_random_points", "num_uncertain_points", "oversample_ratio", "padded_tensor", "padded_tensors", "padding_mask", "padding_masks", "point_coordinates", "point_coords", "point_labels", "point_logits", "point_uncertainties", "pred_logits", "pred_logits_transposed", "pred_masks", "predictions_indices", "rand", "reduce", "register_buffer", "requires_backends", "return", "sample_point", "sample_points_using_uncertainty", "scipy", "self", "shape", "shift", "sigmoid_cross_entropy_loss", "squeeze", "src", "src_idx", "str", "sublist", "sum", "super", "t", "target", "target_classes", "target_classes_o", "target_indices", "target_masks", "tensor", "tensors", "text_queries", "tgt", "tgt_idx", "the_list", "topk", "torch", "transpose", "tuple", "uncertainty_function", "uncertainty_scores", "update", "value", "view", "w", "weight", "weight_dict", "with", "world_size", "zeros", "zip"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderOutput": ["BaseModelOutput", "FloatTensor", "ModelTransformerDecoderOutput", "NModel", "Optional", "auxiliary_predictions", "class", "contrastive_logits", "dict", "object_queries", "prediction_class", "prediction_masks", "r", "str", "torch", "tuple"], "oneformer/modeling_oneformer.py:OneFormerPixelDecoderOutput": ["FloatTensor", "ModelOutput", "ModelPixelDecoderOutput", "NModel", "Optional", "attentions", "class", "mask_features", "multi_scale_features", "r", "torch", "tuple"], "oneformer/modeling_oneformer.py:OneFormerPixelLevelModuleOutput": ["FloatTensor", "ModelOutput", "ModelPixelLevelModuleOutput", "NModel", "Optional", "class", "decoder_features", "decoder_last_feature", "encoder_features", "list", "r", "torch"], "oneformer/modeling_oneformer.py:OneFormerModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "NModel", "Optional", "attentions", "class", "dict", "encoder_hidden_states", "pixel_decoder_hidden_states", "r", "str", "task_token", "text_queries", "torch", "transformer_decoder_auxiliary_predictions", "transformer_decoder_class_predictions", "transformer_decoder_contrastive_queries", "transformer_decoder_hidden_states", "transformer_decoder_mask_predictions", "transformer_decoder_object_queries", "tuple"], "oneformer/modeling_oneformer.py:OneFormerForUniversalSegmentationOutput": ["FloatTensor", "ModelForUniversalSegmentationOutput", "ModelOutput", "NModel", "Optional", "attentions", "auxiliary_predictions", "class", "class_queries_logits", "dict", "encoder_hidden_states", "list", "loss", "masks_queries_logits", "pixel_decoder_hidden_states", "r", "str", "task_token", "text_queries", "torch", "transformer_decoder_auxiliary_predictions", "transformer_decoder_class_predictions", "transformer_decoder_contrastive_queries", "transformer_decoder_hidden_states", "transformer_decoder_mask_predictions", "transformer_decoder_object_queries", "tuple"], "oneformer/modeling_oneformer.py:OneFormerPixelDecoderFrozenBatchNorm2d": ["ModelPixelDecoderFrozenBatchNorm2d", "Models", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "oneformer/modeling_oneformer.py:OneFormerPixelDecoderEncoderMultiscaleDeformableAttention": ["CUDA", "DeformableDetrMultiscaleDeformableAttention", "False", "Last", "Linear", "Make", "ModelPixelDecoderEncoderMultiscaleDeformableAttention", "Module", "NModel", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "d", "d_model", "def", "dim", "dim_per_head", "dimension", "divisible", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "forward", "functional", "got", "head", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "make", "masked_fill", "more", "multi_scale_deformable_attention", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "value", "value_proj", "view", "warn", "warnings", "which", "with", "with_pos_embed"], "oneformer/modeling_oneformer.py:OneFormerPixelDecoderEncoderLayer": ["False", "LayerNorm", "Linear", "ModelConfig", "ModelPixelDecoderEncoderLayer", "ModelPixelDecoderEncoderMultiscaleDeformableAttention", "Module", "NModel", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "conv_dim", "def", "dropout", "dtype", "embed_dim", "encoder_attention_mask", "encoder_feedforward_dim", "encoder_hidden_states", "eps", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "is_training", "isinf", "isnan", "layer_norm_eps", "level_start_index", "max", "min", "n_levels", "n_points", "nn", "num_attention_heads", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "relu", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "spatial_shapes", "super", "torch", "training"], "oneformer/modeling_oneformer.py:OneFormerPixelDecoderEncoderOnly": ["BaseModelOutput", "ModelConfig", "ModelPixelDecoderEncoderLayer", "ModelPixelDecoderEncoderOnly", "Module", "ModuleList", "NModel", "_", "__init__", "all_attentions", "append", "attention_mask", "attentions", "cat", "class", "config", "def", "device", "dropout", "dtype", "else", "encoder_layer", "encoder_layers", "encoder_states", "enumerate", "for", "forward", "get_reference_points", "height", "hidden_states", "i", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "level_start_index", "linspace", "lvl", "meshgrid", "nn", "not", "output_attentions", "output_hidden_states", "position_embeddings", "r", "range", "ref", "ref_x", "ref_y", "reference_points", "reference_points_list", "reshape", "return", "return_dict", "self", "spatial_shapes", "stack", "staticmethod", "super", "torch", "use_return_dict", "valid_ratios", "width"], "oneformer/modeling_oneformer.py:OneFormerPixelDecoder": ["Conv2d", "False", "GroupNorm", "ModelConfig", "ModelPixelDecoder", "ModelPixelDecoderEncoderOnly", "ModelPixelDecoderOutput", "ModelSinePositiModelmbedding", "Module", "ModuleList", "NModel", "Parameter", "ReLU", "Sequential", "Tensor", "True", "_", "__init__", "adapter_", "add_module", "align_corners", "append", "as_tensor", "attention_mask", "attentions", "batch_size", "bias", "bilinear", "bool", "bs", "cat", "class", "common_stride", "config", "conv_dim", "cumsum", "cur_fpn", "def", "device", "dim", "dtype", "else", "encoder", "encoder_outputs", "enumerate", "f", "feats", "feature_channels", "features", "flatten", "float32", "for", "forward", "functional", "get_valid_ratio", "height", "i", "idx", "if", "in", "in_channels", "input_projections", "input_projections_list", "inputs_embeds", "int", "interpolate", "is", "kernel_size", "last_hidden_state", "lateral_conv", "lateral_convs", "layer_", "level", "level_embed", "level_start_index", "log2", "long", "lvl_pos_embed", "lvl_pos_embed_flatten", "m", "mask", "mask_dim", "mask_features", "mask_flatten", "mask_projection", "masks", "min", "mode", "multi_scale_features", "new_zeros", "nn", "normalize", "not", "np", "num_channels", "num_cur_levels", "num_feature_levels", "num_fpn_levels", "num_pos_feats", "o", "out", "output_attentions", "output_conv", "output_convs", "output_hidden_states", "padding", "pos_embed", "position_embedding", "position_embeddings", "position_embeddings_list", "prod", "range", "return", "return_dict", "self", "shape", "size", "source", "source_flatten", "sources", "spatial_shape", "spatial_shapes", "split", "split_size_or_sections", "stack", "stride", "strides", "sum", "super", "to", "torch", "transformer_feature_strides", "transformer_in_channels", "transpose", "valid_height", "valid_ratio", "valid_ratio_height", "valid_ratio_width", "valid_ratios", "valid_width", "view", "width", "x", "y", "z", "zeros", "zip"], "oneformer/modeling_oneformer.py:OneFormerPixelLevelModule": ["False", "ModelConfig", "ModelPixelDecoder", "ModelPixelDecoderOutput", "ModelPixelLevelModule", "ModelPixelLevelModuleOutput", "Module", "Tensor", "__init__", "bool", "channels", "class", "config", "decoder", "decoder_features", "decoder_last_feature", "decoder_output", "def", "encoder", "encoder_features", "feature_channels", "feature_maps", "features", "forward", "list", "load_backbModel", "mask_features", "multi_scale_features", "nn", "output_hidden_states", "pixel_values", "return", "self", "super", "tuple"], "oneformer/modeling_oneformer.py:OneFormerAttention": ["Attention", "False", "Linear", "ModelAttention", "Module", "NModel", "Optional", "Tensor", "True", "ValueError", "__init__", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "int", "is", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_position_embeddings", "key_value_states", "key_value_states_original", "mask", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "permute", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderSelfAttentionLayer": ["ACT2FN", "Dropout", "False", "LayerNorm", "ModelAttention", "ModelTransformerDecoderSelfAttentionLayer", "Module", "NModel", "Optional", "Tensor", "True", "__init__", "activation", "attention_mask", "attention_weights", "class", "def", "dropout", "else", "embed_dim", "eps", "forward", "forward_post", "forward_pre", "hidden_states", "if", "is", "is_decoder", "layer_norm_eps", "nn", "norm", "normalize_before", "num_heads", "output", "output2", "output_attentions", "output_key_padding_mask", "output_mask", "pos", "position_embeddings", "query_pos", "relu", "return", "self", "self_attn", "super", "tensor", "with_pos_embed"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderCrossAttentionLayer": ["ACT2FN", "Dropout", "False", "LayerNorm", "ModelTransformerDecoderCrossAttentionLayer", "Module", "MultiheadAttention", "NModel", "Optional", "Tensor", "__init__", "activation", "attention_weights", "attn_mask", "class", "def", "dropout", "else", "embed_dim", "eps", "forward", "forward_post", "forward_pre", "if", "is", "key", "key_padding_mask", "layer_norm_eps", "memory", "memory_key_padding_mask", "memory_mask", "multihead_attn", "nn", "norm", "normalize_before", "num_heads", "output", "output2", "pos", "query", "query_pos", "relu", "return", "self", "super", "tensor", "value", "with_pos_embed"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderFFNLayer": ["ACT2FN", "Dropout", "False", "LayerNorm", "Linear", "ModelTransformerDecoderFFNLayer", "Module", "NModel", "Optional", "Tensor", "__init__", "activation", "class", "d_model", "def", "dim_feedforward", "dropout", "else", "eps", "forward", "forward_post", "forward_pre", "if", "is", "layer_norm_eps", "linear1", "linear2", "nn", "norm", "normalize_before", "output", "output2", "pos", "relu", "return", "self", "super", "tensor", "with_pos_embed"], "oneformer/modeling_oneformer.py:OneFormerMLPPredictionHead": ["Identity", "ModelMLPPredictionHead", "Module", "PredictionBlock", "ReLU", "Sequential", "Tensor", "__init__", "activation", "append", "class", "def", "else", "enumerate", "for", "forward", "hidden_dim", "i", "if", "in", "in_dim", "in_dims", "input", "input_dim", "int", "layers", "nn", "num_layers", "out_dim", "out_dims", "output_dim", "return", "self", "super", "zip"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderLayer": ["False", "ModelConfig", "ModelTransformerDecoderCrossAttentionLayer", "ModelTransformerDecoderFFNLayer", "ModelTransformerDecoderLayer", "ModelTransformerDecoderSelfAttentionLayer", "Module", "NModel", "Optional", "Tensor", "__init__", "attention_mask", "bool", "class", "config", "cross_attn", "cross_attn_weights", "d_model", "def", "dim_feedforward", "dropout", "embed_dim", "ffn", "forward", "hidden_dim", "if", "index", "int", "layer_norm_eps", "level_index", "list", "memory_key_padding_mask", "memory_mask", "multi_stage_features", "multi_stage_positional_embeddings", "nn", "normalize_before", "num_attention_heads", "num_feature_levels", "num_heads", "output", "output_attentions", "output_key_padding_mask", "output_mask", "outputs", "pos", "pre_norm", "query_embeddings", "query_pos", "return", "self", "self_attn", "self_attn_weights", "shape", "sum", "super", "torch", "where"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderQueryTransformerDecoder": ["False", "ModelTransformerDecoderQueryTransformerDecoder", "Module", "NModel", "Optional", "Tensor", "__init__", "_get_clModels", "append", "class", "decoder_layer", "def", "for", "forward", "if", "in", "intermediate", "is", "layer", "layers", "memory", "memory_key_padding_mask", "memory_mask", "nn", "norm", "not", "num_layers", "output", "output_key_padding_mask", "output_mask", "pop", "pos", "query_pos", "return", "return_intermediate", "self", "stack", "super", "torch", "unsqueeze"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderQueryTransformerDecoderLayer": ["ACT2FN", "Dropout", "False", "LayerNorm", "Linear", "ModelTransformerDecoderQueryTransformerDecoderLayer", "Module", "MultiheadAttention", "NModel", "Optional", "Tensor", "__init__", "activation", "attn_mask", "class", "d_model", "def", "dim_feedforward", "dropout", "dropout1", "dropout2", "dropout3", "else", "eps", "forward", "forward_post", "forward_pre", "if", "is", "k", "key", "key_padding_mask", "layer_norm_eps", "linear1", "linear2", "memory", "memory_key_padding_mask", "memory_mask", "multihead_attn", "nhead", "nn", "norm1", "norm2", "norm3", "normalize_before", "output", "output2", "output_key_padding_mask", "output_mask", "pos", "q", "query", "query_pos", "relu", "return", "self", "self_attn", "super", "tensor", "value", "with_pos_embed"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoderQueryTransformer": ["False", "LayerNorm", "ModelTransformerDecoderQueryTransformer", "ModelTransformerDecoderQueryTransformerDecoder", "ModelTransformerDecoderQueryTransformerDecoderLayer", "Module", "NModel", "__init__", "activation", "batch_size", "class", "d_model", "decoder", "decoder_layer", "decoder_norm", "def", "dim_feedforward", "dropout", "else", "eps", "flatten", "forward", "if", "is", "layer_norm_eps", "mask", "memory_key_padding_mask", "nhead", "nn", "normalize_before", "not", "num_decoder_layers", "permute", "pos", "pos_embed", "queries", "query_embed", "query_pos", "relu", "repeat", "return", "return_intermediate", "return_intermediate_dec", "self", "shape", "src", "super", "task_token", "torch", "transpose", "unsqueeze", "zeros_like"], "oneformer/modeling_oneformer.py:OneFormerTransformerDecoder": ["Conv2d", "False", "Intermediate", "LayerNorm", "Linear", "ModelConfig", "ModelMLPPredictionHead", "ModelTransformerDecoder", "ModelTransformerDecoderLayer", "ModelTransformerDecoderOutput", "ModelTransformerDecoderQueryTransformer", "Module", "ModuleList", "NModel", "ValueError", "_", "__init__", "_get_aux_predictions", "a", "align_corners", "append", "as", "attention_mask", "attention_mask_target_size", "attentions", "aux_list", "auxiliary_predictions", "b", "bchw", "bilinear", "bool", "bqc", "bqhw", "cat", "clModel", "class", "class_embed", "class_queries_logits", "config", "contrastive_logits", "d_model", "decoder", "decoder_layers", "decoder_norm", "decoder_output", "def", "detach", "dim", "dim_feedforward", "dropout", "einsum", "elements", "else", "enumerate", "eps", "flatten", "for", "forward", "forward_prediction_heads", "functional", "have", "hidden_dim", "if", "in", "in_channels", "index", "int", "intermediate_class_predictions", "intermediate_mask_predictions", "interpolate", "is_training", "jit", "kernel_size", "layer", "layer_norm_eps", "layer_outputs", "layers", "len", "mask_dim", "mask_embed", "mask_features", "masks_queries_logits", "mode", "multi_stage_features", "multi_stage_positional_embeddings", "must", "nhead", "nn", "normalize_before", "not", "num_attention_heads", "num_decoder_layers", "num_feature_levels", "num_heads", "num_labels", "number", "object_queries", "of", "output", "output_attentions", "outputs_class", "outputs_mask", "outputs_seg_masks", "permute", "pre_norm", "prediction_class", "prediction_masks", "predictions", "queries", "query_dec_layers", "query_embedder", "query_embeddings", "query_features", "query_input_projection", "query_transformer", "raise", "range", "repeat", "return", "return_intermediate_dec", "same", "self", "sigmoid", "size", "size_list", "super", "task_token", "the", "torch", "transformer", "transpose", "tuple", "unsqueeze", "unused", "use_auxiliary_loss", "use_task_norm", "weight", "zip"], "oneformer/modeling_oneformer.py:OneFormerTransformerModule": ["Conv2d", "Embedding", "False", "ModelConfig", "ModelSinePositiModelmbedding", "ModelTransformerDecoder", "ModelTransformerDecoderOutput", "ModelTransformerModule", "Module", "NModel", "Number", "Sequential", "Tensor", "True", "ValueError", "_", "__init__", "and", "append", "batch_size", "bool", "class", "config", "decoder", "def", "device", "do", "dtype", "elements", "else", "enforce_input_proj", "f", "flatten", "for", "forward", "hidden_dim", "i", "if", "in", "in_channels", "in_features", "input_projections", "int", "kernel_size", "len", "level_embed", "list", "mask_features", "match", "multi_scale_features", "multi_stage_features", "multi_stage_positional_embeddings", "nn", "normalize", "not", "num_feature_levels", "num_pos_feats", "num_queries", "of", "or", "output_attentions", "permute", "position_embedder", "queries_embedder", "query_embedder", "query_embeddings", "query_features", "raise", "range", "repeat", "return", "self", "shape", "size_list", "super", "task_token", "unsqueeze", "weight"], "oneformer/modeling_oneformer.py:OneFormerSinePositionEmbedding": ["False", "ModelSinePositiModelmbedding", "Module", "NModel", "Optional", "Size", "Tensor", "True", "Union", "ValueError", "__init__", "and", "arange", "be", "bool", "cat", "class", "compile_compatible_method_lru_cache", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "else", "eps", "flatten", "float", "floor", "forward", "if", "int", "int64", "is", "mask", "math", "maxsize", "nn", "normalize", "not", "not_mask", "num_pos_feats", "passed", "permute", "pi", "pos", "pos_x", "pos_y", "raise", "return", "rounding_mode", "scale", "self", "shape", "should", "sin", "stack", "str", "super", "temperature", "to", "torch", "x_embed", "y_embed", "zeros"], "oneformer/modeling_oneformer.py:PredictionBlock": ["Linear", "ModelBlock", "Module", "None", "Tensor", "__init__", "activation", "add_module", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_dim", "input", "int", "layer", "layers", "nn", "out_dim", "return", "self", "str", "super"], "oneformer/modeling_oneformer.py:OneFormerTextMapperAttention": ["Dropout", "False", "Linear", "ModelTextMapperAttention", "Module", "NModel", "ValueError", "__init__", "and", "attn", "attn_drop", "batch_size", "bias", "bknm", "bmkc", "bnkc", "class", "def", "different", "dim", "einsum", "f", "forward", "have", "head_dim", "if", "k", "k_proj", "k_sequence_length", "keys", "list", "nn", "not", "num_channels", "num_heads", "or", "output", "proj", "proj_drop", "q", "q_proj", "q_sequence_length", "qk_scale", "qkv_bias", "raise", "reshape", "return", "scale", "self", "shape", "shapes", "softmax", "super", "torch", "v", "v_proj", "values"], "oneformer/modeling_oneformer.py:OneFormerTextTransformerDecoderLayer": ["Dropout", "GELU", "LayerNorm", "Linear", "ModelTextMapperAttention", "ModelTextTransformerDecoderLayer", "Module", "Sequential", "__init__", "class", "cross_attn", "d_model", "def", "dropout", "eps", "forward", "hidden_state", "k", "layer_norm_eps", "mem", "mlp", "nhead", "nn", "norm1", "norm2", "norm3", "proj_drop", "q", "return", "self", "self_attn", "super", "v"], "oneformer/modeling_oneformer.py:OneFormerTextContextDecoder": ["LayerNorm", "Linear", "ModelTextContextDecoder", "ModelTextTransformerDecoderLayer", "Module", "ModuleList", "Sequential", "_", "__init__", "class", "decoder", "def", "dropout", "eps", "for", "forward", "hidden_state", "in", "kwargs", "layer", "layer_norm_eps", "memory_proj", "nn", "out_proj", "range", "return", "self", "super", "text", "text_proj", "transformer_heads", "transformer_layers", "transformer_width", "visual", "visual_dim"], "oneformer/modeling_oneformer.py:OneFormerTextMLP": ["ACT2FN", "Linear", "ModelTextMLP", "Module", "NModel", "Optional", "Tensor", "__init__", "activation_fn", "class", "def", "fc1", "fc2", "forward", "hidden_size", "hidden_states", "int", "intermediate_size", "nn", "output_size", "quick_gelu", "return", "self", "super", "torch"], "oneformer/modeling_oneformer.py:OneFormerTextTransformerLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelTextMLP", "ModelTextTransformerLayer", "MultiheadAttention", "NModel", "Optional", "Tensor", "__init__", "attn_mask", "class", "def", "eps", "forward", "heads", "hidden_states", "int", "key_padding_mask", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "need_weights", "nn", "residual", "return", "self", "self_attn", "super", "torch", "width"], "oneformer/modeling_oneformer.py:OneFormerTextTransformer": ["False", "ModelTextTransformer", "ModelTextTransformerLayer", "Module", "NModel", "Optional", "Sequential", "Tensor", "_", "__init__", "attn_mask", "class", "def", "for", "forward", "heads", "hidden_states", "in", "int", "layer", "layer_norm_eps", "layers", "nn", "num_layers", "range", "return", "self", "super", "torch", "use_checkpoint", "width"], "oneformer/modeling_oneformer.py:OneFormerTextEncoder": ["Embedding", "False", "LayerNorm", "ModelTextEncoder", "ModelTextTransformer", "Module", "Parameter", "__init__", "arange", "argmax", "attn_mask", "build_attention_mask", "class", "context_length", "def", "dim", "empty", "eps", "fill_", "float", "forward", "heads", "hidden_state", "inf", "int", "layer_norm_eps", "layers", "ln_final", "mask", "nn", "permute", "positional_embedding", "return", "self", "shape", "super", "text", "token_embedding", "torch", "transformer", "triu_", "use_checkpoint", "vocab_size", "width"], "oneformer/modeling_oneformer.py:OneFormerTextMapper": ["Embedding", "False", "ModelConfig", "ModelMLPPredictionHead", "ModelTextEncoder", "ModelTextMapper", "Module", "NModel", "NModelType", "Number", "Tensor", "True", "ValueError", "_", "__init__", "batch_size", "be", "cat", "class", "config", "context_length", "def", "dim", "dimensions", "else", "encode_text", "encoded_text", "forward", "hidden_dim", "if", "in", "inputs", "is", "layer_norm_eps", "layers", "must", "ndim", "nn", "not", "num_text", "of", "or", "prompt_ctx", "raise", "repeat", "reshape", "return", "self", "shape", "squeeze_dim", "super", "text", "text_encoder", "text_encoder_context_length", "text_encoder_n_ctx", "text_encoder_num_layers", "text_encoder_proj_layers", "text_encoder_vocab_size", "text_encoder_width", "text_projector", "text_queries", "text_queries_ctx", "torch", "unsqueeze", "vocab_size", "weight", "width"], "oneformer/modeling_oneformer.py:OneFormerTaskModel": ["ModelConfig", "ModelMLPPredictionHead", "ModelTaskModel", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_dim", "inputs", "nn", "return", "self", "super", "task_mlp", "task_seq_len", "task_tokens"], "oneformer/modeling_oneformer.py:OneFormerPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "GroupNorm", "LayerNorm", "Linear", "ModelConfig", "ModelLoss", "ModelMLPPredictionHead", "ModelPixelDecoder", "ModelPixelDecoderEncoderMultiscaleDeformableAttention", "ModelPreTrainedModel", "ModelTextEncoder", "ModelTextTransformer", "ModelTransformerDecoder", "ModelTransformerDecoderLayer", "ModelTransformerDecoderQueryTransformer", "ModelTransformerModule", "Module", "MultiheadAttention", "NModel", "Parameter", "PreTrainedModel", "Sequential", "True", "_init_weights", "abs", "arange", "attention_weights", "attn_std", "base_model_prefix", "bias", "class", "config", "constant_", "contrastive_temperature", "cos", "data", "def", "dim", "dtype", "elif", "fc1", "fc2", "fc_std", "fill_", "float", "for", "gain", "grid_init", "hasattr", "i", "if", "in", "in_proj_bias", "in_proj_weight", "init", "init_std", "init_xavier_std", "input_projection", "input_projections", "int64", "is", "isinstance", "keepdim", "layer", "layers", "level_embed", "log", "logit_scale", "main_input_name", "math", "max", "mean", "mlp", "model", "module", "modules", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "np", "num_layers", "out_proj", "output_proj", "p", "padding_idx", "parameters", "pi", "pixel_values", "positional_embedding", "proj_std", "query_input_projection", "range", "reference_points", "repeat", "sampling_offsets", "self", "self_attn", "sin", "stack", "std", "submodule", "thetas", "token_embedding", "torch", "value_proj", "view", "weight", "width", "with", "xavier_std", "xavier_uniform_", "zero_"], "oneformer/modeling_oneformer.py:OneFormerModel": ["ModelConfig", "ModelModel", "ModelModelOutput", "ModelPixelLevelModule", "ModelPreTrainedModel", "ModelTaskModel", "ModelTextMapper", "ModelTransformerModule", "Models", "NModel", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "attentions", "auto_docstring", "auxiliary_predictions", "batch_size", "bool", "class", "config", "contrastive_logits", "conv_dim", "decoder_features", "decoder_last_feature", "def", "device", "dtype", "else", "encoder_features", "encoder_hidden_states", "f", "for", "forward", "have", "height", "if", "in", "in_features", "is", "is_training", "main_input_name", "mask_features", "multi_scale_features", "not", "object_queries", "output", "output_attentions", "output_hidden_states", "pixel_decoder_hidden_states", "pixel_level_module", "pixel_level_module_output", "pixel_mask", "pixel_values", "post_init", "prediction_class", "prediction_masks", "queries", "r", "raise", "return", "return_dict", "self", "shape", "specify", "super", "task_encoder", "task_inputs", "task_token", "text_inputs", "text_mapper", "text_queries", "to", "torch", "transformer_decoder_auxiliary_predictions", "transformer_decoder_class_predictions", "transformer_decoder_contrastive_queries", "transformer_decoder_hidden_states", "transformer_decoder_mask_predictions", "transformer_decoder_object_queries", "transformer_module", "transformer_module_output", "tuple", "use_return_dict", "v", "values", "width"], "oneformer/modeling_oneformer.py:OneFormerForUniversalSegmentation": ["ModelConfig", "ModelForUniversalSegmentation", "ModelForUniversalSegmentationOutput", "ModelHungarianMatcher", "ModelLoss", "ModelModel", "ModelPreTrainedModel", "NModel", "Optional", "Tensor", "True", "__init__", "and", "auto_docstring", "auxiliary_predictions", "bool", "calculate_contrastive_loss", "class", "class_labels", "class_queries_logits", "class_weight", "config", "contrastive_queries_logits", "contrastive_temperature", "contrastive_weight", "cost_class", "cost_dice", "cost_mask", "criterion", "def", "dice_weight", "dict", "else", "eos_coef", "float", "for", "forward", "get_loss", "get_loss_dict", "if", "importance_sample_ratio", "in", "is", "items", "key", "list", "loss", "loss_contrastive", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_key", "loss_mask", "main_input_name", "mask_labels", "mask_weight", "masks_queries_logits", "matcher", "model", "no_object_weight", "not", "num_classes", "num_labels", "num_points", "or", "output", "output_attentions", "output_auxiliary_logits", "output_hidden_states", "outputs", "oversample_ratio", "pixel_mask", "pixel_values", "post_init", "r", "return", "return_dict", "self", "str", "sum", "super", "task_inputs", "text_inputs", "text_queries", "train_num_points", "transformer_decoder_auxiliary_predictions", "transformer_decoder_class_predictions", "transformer_decoder_contrastive_queries", "transformer_decoder_mask_predictions", "tuple", "use_auxiliary_loss", "use_return_dict", "v", "values", "weight", "weight_dict"], "efficientnet/modeling_efficientnet.py:round_filters": ["ModelConfig", "Model_filters", "config", "def", "depth_divisor", "divisor", "if", "int", "max", "new_dim", "num_channels", "r", "return", "width_coefficient"], "efficientnet/modeling_efficientnet.py:correct_pad": ["Model", "Model_pad", "True", "Union", "adjust", "bool", "def", "else", "if", "int", "isinstance", "kernel_size", "r", "return", "tuple"], "efficientnet/modeling_efficientnet.py:EfficientNetEmbeddings": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "ModelConfig", "ModelEmbeddings", "Module", "Tensor", "ZeroPad2d", "__init__", "activation", "batch_norm_eps", "batch_norm_momentum", "batchnorm", "bias", "class", "config", "convolution", "def", "eps", "features", "forward", "hidden_act", "kernel_size", "momentum", "nn", "num_channels", "out_dim", "padding", "pixel_values", "r", "return", "round_filters", "self", "stride", "super", "torch", "valid"], "efficientnet/modeling_efficientnet.py:EfficientNetDepthwiseConv2d": ["Conv2d", "ModelDepthwiseConv2d", "True", "__init__", "bias", "class", "def", "depth_multiplier", "dilation", "groups", "in_channels", "kernel_size", "nn", "out_channels", "padding", "padding_mode", "self", "stride", "super", "zeros"], "efficientnet/modeling_efficientnet.py:EfficientNetExpansionLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "FloatTensor", "ModelConfig", "ModelExpansionLayer", "Module", "Tensor", "__init__", "batch_norm_eps", "bias", "class", "config", "def", "eps", "expand_act", "expand_bn", "expand_conv", "forward", "hidden_act", "hidden_states", "in_channels", "in_dim", "int", "kernel_size", "nn", "num_features", "out_channels", "out_dim", "padding", "r", "return", "same", "self", "stride", "super", "torch"], "efficientnet/modeling_efficientnet.py:EfficientNetDepthwiseLayer": ["ACT2FN", "BatchNorm2d", "False", "FloatTensor", "ModelConfig", "ModelDepthwiseConv2d", "ModelDepthwiseLayer", "Module", "Tensor", "ZeroPad2d", "__init__", "adjust", "adjust_padding", "batch_norm_eps", "batch_norm_momentum", "bias", "bool", "class", "config", "conv_pad", "correct_pad", "def", "depthwise_act", "depthwise_conv", "depthwise_conv_pad", "depthwise_norm", "else", "eps", "forward", "hidden_act", "hidden_states", "if", "in_dim", "int", "kernel_size", "momentum", "nn", "num_features", "padding", "r", "return", "same", "self", "stride", "super", "torch", "valid"], "efficientnet/modeling_efficientnet.py:EfficientNetSqueezeExciteLayer": ["ACT2FN", "AdaptiveAvgPool2d", "Conv2d", "False", "FloatTensor", "ModelConfig", "ModelSqueezeExciteLayer", "Module", "Sigmoid", "Tensor", "__init__", "act_expand", "act_reduce", "bool", "class", "config", "def", "dim", "dim_se", "else", "expand", "expand_dim", "forward", "hidden_act", "hidden_states", "if", "in_channels", "in_dim", "inputs", "int", "kernel_size", "max", "mul", "nn", "out_channels", "output_size", "padding", "r", "reduce", "return", "same", "self", "squeeze", "squeeze_expansion_ratio", "super", "torch"], "efficientnet/modeling_efficientnet.py:EfficientNetFinalBlockLayer": ["BatchNorm2d", "Conv2d", "Dropout", "False", "FloatTensor", "ModelConfig", "ModelFinalBlockLayer", "Module", "Tensor", "__init__", "and", "apply_dropout", "batch_norm_eps", "batch_norm_momentum", "bias", "bool", "class", "config", "def", "drop_rate", "dropout", "embeddings", "eps", "float", "forward", "hidden_states", "id_skip", "if", "in_channels", "in_dim", "int", "kernel_size", "momentum", "nn", "not", "num_features", "out_channels", "out_dim", "p", "padding", "project_bn", "project_conv", "r", "return", "same", "self", "stride", "super", "torch"], "efficientnet/modeling_efficientnet.py:EfficientNetBlock": ["FloatTensor", "ModelBlock", "ModelConfig", "ModelDepthwiseLayer", "ModelExpansionLayer", "ModelFinalBlockLayer", "ModelSqueezeExciteLayer", "Module", "Tensor", "__init__", "adjust_padding", "bool", "class", "config", "def", "depthwise_conv", "drop_rate", "else", "embeddings", "expand", "expand_dim", "expand_in_dim", "expand_ratio", "expansion", "float", "forward", "hidden_states", "id_skip", "if", "in_dim", "int", "kernel_size", "nn", "out_dim", "projection", "r", "return", "self", "squeeze_excite", "stride", "super", "torch"], "efficientnet/modeling_efficientnet.py:EfficientNetEncoder": ["ACT2FN", "BaseModelOutputWithNoAttention", "BatchNorm2d", "Conv2d", "False", "FloatTensor", "ModelBlock", "ModelConfig", "ModelEncoder", "Module", "ModuleList", "None", "Optional", "True", "__init__", "adjust_padding", "all_hidden_states", "append", "batch_norm_eps", "batch_norm_momentum", "bias", "block", "blocks", "bool", "ceil", "class", "config", "curr_block_num", "def", "depth_coModel", "depthwise_padding", "drop_connect_rate", "drop_rate", "else", "eps", "expand_ratio", "expand_ratios", "for", "forward", "hidden_act", "hidden_dim", "hidden_states", "i", "id_skip", "if", "in", "in_channels", "in_dim", "int", "is", "j", "kernel_size", "kernel_sizes", "last_hidden_state", "len", "math", "momentum", "n", "nn", "not", "num_base_blocks", "num_block_repeats", "num_blocks", "num_features", "out_channels", "out_dim", "output_hidden_states", "padding", "r", "range", "repeats", "return", "return_dict", "round_filters", "round_repeats", "same", "self", "stride", "strides", "sum", "super", "top_activation", "top_bn", "top_conv", "torch", "tuple", "v"], "efficientnet/modeling_efficientnet.py:EfficientNetPreTrainedModel": ["BatchNorm2d", "Conv2d", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "efficientnet/modeling_efficientnet.py:EfficientNetModel": ["AvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "MaxPool2d", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "auto_docstring", "be", "bool", "ceil_mode", "class", "config", "def", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "f", "forward", "got", "have", "hidden_dim", "hidden_states", "if", "is", "last_hidden_state", "max", "mean", "must", "nn", "not", "of", "one", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "pooling", "pooling_type", "post_init", "raise", "reshape", "return", "return_dict", "self", "shape", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "efficientnet/modeling_efficientnet.py:EfficientNetForImageClassification": ["Dropout", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "dropout_rate", "else", "forward", "hidden_dim", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "p", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "mobilebert/modeling_mobilebert.py:NoNorm": ["Module", "NoNorm", "None", "Parameter", "Tensor", "__init__", "bias", "class", "def", "eps", "feat_size", "forward", "input_tensor", "nn", "ones", "return", "self", "super", "torch", "weight", "zeros"], "mobilebert/modeling_mobilebert.py:MobileBertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "ModelEmbeddings", "Module", "NORM2FN", "None", "Optional", "Tensor", "__init__", "arange", "cat", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embed_dim_multiplier", "embedded_input_size", "embedding_size", "embedding_transformation", "embeddings", "expand", "forward", "functional", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "long", "max_position_embeddings", "nn", "normalization_type", "not", "or", "pad", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "trigram_input", "type_vocab_size", "value", "vocab_size", "word_embeddings", "zeros"], "mobilebert/modeling_mobilebert.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "mobilebert/modeling_mobilebert.py:MobileBertSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "all_head_size", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_mask", "hidden_shape", "hidden_size", "if", "input_shape", "int", "is_causal", "key", "key_layer", "key_tensor", "kwargs", "nn", "not", "num_attention_heads", "p", "query", "query_layer", "query_tensor", "reshape", "return", "scaling", "self", "shape", "super", "torch", "training", "transpose", "true_hidden_size", "tuple", "use_bottleneck_attention", "value", "value_layer", "value_tensor", "view"], "mobilebert/modeling_mobilebert.py:MobileBertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "NORM2FN", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_states", "if", "layer_norm_eps", "layer_outputs", "nn", "normalization_type", "not", "residual_tensor", "return", "self", "super", "torch", "true_hidden_size", "use_bottleneck"], "mobilebert/modeling_mobilebert.py:MobileBertAttention": ["FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "if", "index", "key", "key_tensor", "kwargs", "layer_input", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "query_tensor", "return", "self", "set", "super", "torch", "tuple", "union", "value", "value_tensor"], "mobilebert/modeling_mobilebert.py:MobileBertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch", "true_hidden_size"], "mobilebert/modeling_mobilebert.py:OutputBottleneck": ["Dropout", "LayerNorm", "Linear", "ModelBottleneck", "Module", "NORM2FN", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "layer_Models", "layer_norm_eps", "nn", "normalization_type", "residual_tensor", "return", "self", "super", "torch", "true_hidden_size"], "mobilebert/modeling_mobilebert.py:MobileBertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "NORM2FN", "OutputBottleneck", "Tensor", "__init__", "bottleneck", "class", "config", "def", "dense", "dropout", "else", "forward", "hidden_dropout_prob", "if", "intermediate_size", "intermediate_states", "layer_output", "nn", "normalization_type", "not", "residual_tensor_1", "residual_tensor_2", "return", "self", "super", "torch", "true_hidden_size", "use_bottleneck"], "mobilebert/modeling_mobilebert.py:BottleneckLayer": ["LayerNorm", "Linear", "ModelLayer", "Module", "NORM2FN", "Tensor", "__init__", "class", "config", "def", "dense", "eps", "forward", "hidden_size", "hidden_states", "intra_Model_size", "layer_input", "layer_norm_eps", "nn", "normalization_type", "return", "self", "super", "torch"], "mobilebert/modeling_mobilebert.py:Bottleneck": ["Model", "ModelLayer", "Modeled_hidden_states", "Module", "Tensor", "__init__", "attention", "class", "config", "def", "elif", "else", "forward", "hidden_states", "if", "input", "key_query_shared_Model", "nn", "return", "self", "shared_attention_input", "super", "torch", "tuple", "use_Model_attention"], "mobilebert/modeling_mobilebert.py:FFNOutput": ["LayerNorm", "Linear", "Model", "Module", "NORM2FN", "Tensor", "__init__", "class", "config", "def", "dense", "eps", "forward", "hidden_states", "intermediate_size", "layer_norm_eps", "layer_outputs", "nn", "normalization_type", "residual_tensor", "return", "self", "super", "torch", "true_hidden_size"], "mobilebert/modeling_mobilebert.py:FFNLayer": ["FFNOutput", "Model", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_states", "intermediate", "intermediate_output", "layer_outputs", "nn", "output", "return", "self", "super", "torch"], "mobilebert/modeling_mobilebert.py:MobileBertLayer": ["Bottleneck", "FFNLayer", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention", "attention_mask", "attention_output", "bottleneck", "class", "config", "def", "else", "ffn", "ffn_module", "for", "forward", "head_mask", "hidden_states", "if", "in", "intermediate", "intermediate_output", "key_tensor", "kwargs", "layer_input", "layer_output", "nn", "num_feedforward_networks", "output", "query_tensor", "range", "return", "self", "self_attention_output", "super", "torch", "tuple", "use_bottleneck", "value_tensor"], "mobilebert/modeling_mobilebert.py:MobileBertEncoder": ["BaseModelOutput", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "class", "config", "def", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "in", "kwargs", "last_hidden_state", "layer", "layer_module", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "mobilebert/modeling_mobilebert.py:MobileBertPooler": ["Linear", "ModelPooler", "Module", "Tensor", "__init__", "class", "classifier_activation", "config", "def", "dense", "do_activate", "else", "first_token_tensor", "forward", "hidden_size", "hidden_states", "if", "nn", "not", "pooled_output", "return", "self", "super", "tanh", "torch"], "mobilebert/modeling_mobilebert.py:MobileBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "NORM2FN", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "mobilebert/modeling_mobilebert.py:MobileBertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "Tensor", "__init__", "_tie_weights", "bias", "cat", "class", "config", "decoder", "def", "dense", "dim", "embedding_size", "forward", "hidden_size", "hidden_states", "matmul", "nn", "return", "self", "super", "t", "torch", "transform", "vocab_size", "weight", "zeros"], "mobilebert/modeling_mobilebert.py:MobileBertOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "mobilebert/modeling_mobilebert.py:MobileBertPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super", "torch", "tuple"], "mobilebert/modeling_mobilebert.py:MobileBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "NoNorm", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "mobilebert/modeling_mobilebert.py:MobileBertForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "mobilebert/modeling_mobilebert.py:MobileBertModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_full_mask", "add_pooling_layer", "attention", "attention_mask", "auto_docstring", "check_model_inputs", "class", "config", "def", "dtype", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "exactly", "flash", "flex_attention", "for", "forward", "get_head_mask", "get_input_embeddings", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "is", "is_causal", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "make_flex_block_causal_mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "sdpa", "self", "sequence_output", "set_input_embeddings", "specify", "super", "token_type_ids", "torch", "tuple", "value", "word_embeddings"], "mobilebert/modeling_mobilebert.py:MobileBertForPreTraining": ["CrossEntropyLoss", "Embedding", "FloatTensor", "LongTensor", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_get_resized_lm_head", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "can_return_tuple", "class", "cls", "config", "decoder", "def", "dense", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "new_num_tokens", "next_sentence_label", "next_sentence_loss", "nn", "not", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "resize_token_embeddings", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "total_loss", "transposed", "tuple", "view", "vocab_size", "weight"], "mobilebert/modeling_mobilebert.py:MobileBertForMaskedLM": ["CrossEntropyLoss", "Embedding", "False", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_get_resized_lm_head", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bias", "can_return_tuple", "class", "cls", "config", "decoder", "def", "dense", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "new_num_tokens", "nn", "not", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "resize_token_embeddings", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "transposed", "tuple", "view", "vocab_size", "weight"], "mobilebert/modeling_mobilebert.py:MobileBertOnlyNSPHead": ["Linear", "ModelOnlyNSPHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super", "torch"], "mobilebert/modeling_mobilebert.py:MobileBertForNextSentencePrediction": ["CrossEntropyLoss", "FloatTensor", "FutureWarning", "LongTensor", "Model", "ModelForNextSentencePrediction", "ModelModel", "ModelOnlyNSPHead", "ModelPreTrainedModel", "NextSentencePredictorOutput", "None", "Optional", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "a", "and", "argument", "attention_mask", "attentions", "auto_docstring", "be", "can_return_tuple", "class", "cls", "config", "def", "deprecated", "forward", "future", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "outputs", "pooled_output", "pop", "position_ids", "post_init", "r", "removed", "return", "return_dict", "self", "seq_relationship_score", "super", "token_type_ids", "torch", "tuple", "use", "version", "view", "warn", "warnings", "will"], "mobilebert/modeling_mobilebert.py:MobileBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "mobilebert/modeling_mobilebert.py:MobileBertForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "mobilebert/modeling_mobilebert.py:MobileBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "mobilebert/modeling_mobilebert.py:MobileBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "view"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2PreTrainedModel": ["Model", "ModelBasicLayer", "ModelConfig", "PreTrainedModel", "_no_split_modules", "base_model_prefix", "class", "config", "main_input_name", "pixel_values"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2LearnableAffineBlock": ["Model", "Module", "Parameter", "Tensor", "True", "__init__", "bias", "bias_value", "class", "def", "float", "forward", "hidden_state", "nn", "requires_grad", "return", "scale", "scale_value", "self", "super", "tensor", "torch"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2ConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Identity", "Model", "ModelLearnableAffineBlock", "Module", "None", "Tensor", "__init__", "activation", "and", "bias", "bool", "class", "convolution", "def", "else", "forward", "groups", "hidden_state", "if", "in_channels", "input", "int", "is", "kernel_size", "lab", "nn", "normalization", "not", "out_channels", "padding", "relu", "return", "self", "str", "stride", "super", "use_learnable_affine_block"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2ConvLayerLight": ["False", "Model", "ModelConvLayer", "Module", "None", "Tensor", "__init__", "activation", "bool", "class", "conv1", "conv2", "def", "forward", "groups", "hidden_state", "in_channels", "int", "kernel_size", "nn", "out_channels", "return", "self", "super", "use_learnable_affine_block"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2Embeddings": ["F", "Make", "MaxPool2d", "Model", "ModelConfig", "ModelConvLayer", "Module", "Tensor", "True", "ValueError", "__init__", "activation", "cat", "ceil_mode", "channel", "class", "config", "configuration", "def", "dim", "dimension", "emb_stem_2a", "embedding", "forward", "hidden_act", "if", "in", "kernel_size", "match", "nn", "num_channels", "of", "one", "pad", "pixel", "pixel_values", "pool", "pooled_emb", "raise", "return", "self", "set", "shape", "stem1", "stem2a", "stem2b", "stem3", "stem4", "stem_channels", "stride", "super", "sure", "that", "the", "torch", "use_learnable_affine_block", "values", "with"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2BasicLayer": ["Dropout", "False", "Identity", "Model", "ModelConvLayer", "ModelConvLayerLight", "Module", "ModuleList", "Sequential", "Tensor", "__init__", "aggregation", "aggregation_excitation_conv", "aggregation_squeeze_conv", "append", "block", "bool", "cat", "class", "def", "dim", "drop_path", "else", "float", "for", "forward", "hidden_state", "i", "identity", "if", "in", "in_channels", "int", "kernel_size", "layer", "layer_num", "layers", "light_block", "middle_channels", "nn", "out_channels", "output", "range", "residual", "return", "self", "stride", "super", "temp_in_channels", "torch", "total_channels", "use_learnable_affine_block"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2Stage": ["Identity", "Model", "ModelBasicLayer", "ModelConfig", "ModelConvLayer", "Module", "ModuleList", "None", "Tensor", "__init__", "activation", "append", "block", "blocks", "blocks_list", "class", "config", "def", "downsample", "drop_path", "else", "float", "for", "forward", "groups", "hidden_state", "i", "if", "in", "in_channels", "int", "kernel_size", "light_block", "mid_channels", "nn", "num_blocks", "num_layers", "out_channels", "range", "residual", "return", "self", "stage_downsample", "stage_in_channels", "stage_index", "stage_kernel_size", "stage_light_block", "stage_mid_channels", "stage_num_blocks", "stage_numb_of_layers", "stage_out_channels", "stride", "super", "use_learnable_affine_block"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2Encoder": ["BaseModelOutputWithNoAttention", "False", "Model", "ModelConfig", "ModelStage", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "append", "bool", "class", "config", "def", "else", "for", "forward", "hidden_state", "hidden_states", "if", "in", "is", "last_hidden_state", "len", "nn", "not", "output_hidden_states", "range", "resnet_stage", "return", "return_dict", "self", "stage", "stage_in_channels", "stage_index", "stages", "super", "tuple", "v"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2Backbone": ["BackboneMixin", "BackboneOutput", "False", "Model", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "depths", "else", "embedder", "embedding_output", "embedding_size", "encoder", "enumerate", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_features", "out_features", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "stage", "stage_names", "super", "use_return_dict"], "hgnet_v2/modeling_hgnet_v2.py:HGNetV2ForImageClassification": ["AdaptiveAvgPool2d", "Flatten", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "__init__", "auto_docstring", "avg_pool", "bool", "class", "classifier", "config", "def", "else", "embedder", "embedding_output", "encoder", "fc", "flatten", "for", "forward", "hidden_sizes", "hidden_states", "if", "in", "is", "labels", "last_hidden_state", "layer", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "use_return_dict"], "sam/modeling_sam.py:SamVisionEncoderOutput": ["FloatTensor", "ModelOutput", "ModelVisionEncoderOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "sam/modeling_sam.py:SamImageSegmentationOutput": ["FloatTensor", "ModelImageSegmentationOutput", "ModelOutput", "None", "Optional", "class", "iou_scores", "mask_decoder_attentions", "pred_masks", "r", "torch", "tuple", "vision_attentions", "vision_hidden_states"], "sam/modeling_sam.py:SamPatchEmbeddings": ["Conv2d", "Input", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "forward", "height", "hidden_size", "if", "image", "image_size", "in", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "of", "one", "or", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "values", "width", "with"], "sam/modeling_sam.py:SamMLPBlock": ["ACT2FN", "Linear", "ModelMLPBlock", "Module", "Tensor", "__init__", "act", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_states", "lin1", "lin2", "mlp_dim", "nn", "return", "self", "super", "torch"], "sam/modeling_sam.py:SamLayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "sam/modeling_sam.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "sam/modeling_sam.py:SamAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "__init__", "_attn_implementation", "_recombine_heads", "_separate_heads", "attention_downModelple_rate", "attention_interface", "attention_mask", "attention_similarity", "attn_output", "attn_weights", "batch", "c_per_head", "channel", "class", "config", "def", "divide", "downModelple_rate", "dropout", "eager", "eager_attention_forward", "else", "forward", "hidden_size", "hidden_states", "if", "int", "internal_dim", "is", "is_causal", "k_proj", "key", "kwargs", "must", "n_heads", "n_tokens", "nn", "num_attention_heads", "out_proj", "point_batch_size", "q_proj", "query", "raise", "reshape", "return", "scaling", "self", "shape", "super", "transpose", "v_proj", "value"], "sam/modeling_sam.py:SamTwoWayAttentionBlock": ["False", "LayerNorm", "ModelAttention", "ModelMLPBlock", "ModelTwoWayAttentionBlock", "Module", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_downModelple_rate", "attention_similarity", "attn_out", "bool", "class", "config", "cross_attn_image_to_token", "cross_attn_token_to_image", "def", "downModelple_rate", "else", "eps", "forward", "hidden_size", "if", "int", "key", "key_point_embedding", "keys", "kwargs", "layer_norm1", "layer_norm2", "layer_norm3", "layer_norm4", "layer_norm_eps", "mlp", "mlp_out", "nn", "queries", "query", "query_point_embedding", "return", "self", "self_attn", "skip_first_layer_pe", "super", "value"], "sam/modeling_sam.py:SamTwoWayTransformer": ["BaseModelOutput", "LayerNorm", "ModelAttention", "ModelMaskDecoderConfig", "ModelTwoWayAttentionBlock", "ModelTwoWayTransformer", "Module", "ModuleList", "None", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "_", "__init__", "an", "append", "attention_similarity", "attn_out", "class", "config", "def", "final_attn_token_to_image", "flatten", "for", "forward", "have", "hidden_size", "i", "if", "image_embedding", "image_embeddings", "image_positional_embeddings", "in", "is", "key", "key_point_embedding", "keys", "kwargs", "layer", "layer_norm_final_attn", "layers", "nn", "not", "num_hidden_layers", "permute", "point_embeddings", "queries", "query", "query_point_embedding", "raise", "range", "return", "self", "skip_first_layer_pe", "specify", "super", "target_embedding", "to", "tuple", "unsqueeze", "value"], "sam/modeling_sam.py:SamFeedForward": ["F", "False", "Linear", "ModelFeedForward", "Module", "ModuleList", "ReLU", "_", "__init__", "activation", "bool", "class", "def", "for", "forward", "hidden_dim", "hidden_states", "if", "in", "input_dim", "int", "layer", "layers", "nn", "num_layers", "output_dim", "proj_in", "proj_out", "range", "return", "self", "sigmoid", "sigmoid_output", "super"], "sam/modeling_sam.py:SamMaskDecoder": ["ConvTranspose2d", "Embedding", "GELU", "ModelFeedForward", "ModelLayerNorm", "ModelMaskDecoder", "ModelMaskDecoderConfig", "ModelTwoWayTransformer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "activation", "attention_similarity", "batch_size", "bool", "cat", "channels_first", "class", "config", "current_mlp", "data_format", "def", "dense_prompt_embeddings", "dim", "dtype", "else", "for", "forward", "height", "hidden_size", "hyper_in", "hyper_in_list", "i", "if", "image_embeddings", "image_positional_embeddings", "in", "iou_head_depth", "iou_head_hidden_dim", "iou_pred", "iou_prediction_head", "iou_token", "iou_token_out", "is", "kernel_size", "mask_slice", "mask_tokens", "mask_tokens_out", "masks", "mlps_list", "multimask_output", "nn", "not", "num_channels", "num_mask_tokens", "num_multimask_outputs", "output_hypernetworks_mlps", "output_tokens", "point_batch_size", "point_embedding", "point_embeddings", "range", "repeat", "repeat_interleave", "reshape", "return", "self", "shape", "slice", "sparse_prompt_embeddings", "stack", "stride", "super", "target_embedding", "to", "tokens", "torch", "transformer", "transpose", "tuple", "upscale_conv1", "upscale_conv2", "upscale_layer_norm", "upscaled_embedding", "weight", "width"], "sam/modeling_sam.py:SamPositionalEmbedding": ["ModelPositionalEmbedding", "Module", "None", "__init__", "cat", "class", "clone", "config", "coordinates", "cos", "def", "dim", "dtype", "forward", "hidden_size", "if", "input_coords", "input_shape", "is", "nn", "not", "np", "num_pos_feats", "pi", "positional_embedding", "randn", "register_buffer", "return", "scale", "self", "sin", "super", "to", "torch"], "sam/modeling_sam.py:SamMaskEmbedding": ["ACT2FN", "Conv2d", "ModelLayerNorm", "ModelMaskEmbedding", "ModelPromptEncoderConfig", "Module", "__init__", "activation", "channels_first", "class", "config", "conv1", "conv2", "conv3", "data_format", "def", "dense_embeddings", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "layer_norm_eps", "mask_input_channels", "masks", "nn", "return", "self", "stride", "super"], "sam/modeling_sam.py:SamPromptEncoder": ["Embedding", "If", "ModelConfig", "ModelMaskEmbedding", "ModelPositionalEmbedding", "ModelPromptEncoder", "Module", "ModuleList", "None", "Optional", "Tensor", "ValueError", "__init__", "_embed_boxes", "_embed_points", "also", "are", "batch_size", "be", "bool", "box_embeddings", "boxes", "cat", "class", "config", "coords", "corner_embedding", "def", "dense_embeddings", "device", "dim", "else", "expand", "for", "forward", "hidden_size", "i", "if", "image_embedding_size", "image_size", "in", "input_boxes", "input_image_size", "input_labels", "input_masks", "input_points", "input_shape", "is", "labels", "mask_embed", "must", "nb_boxes", "nn", "no_mask_embed", "not", "not_a_point_embed", "num_point_embeddings", "ones", "pad", "padding_label", "padding_point", "point_embed", "point_embedding", "point_embeddings", "points", "prompt_encoder_config", "provided", "raise", "range", "reshape", "return", "self", "shape", "shared_embedding", "sparse_embeddings", "super", "target_labels_shape", "target_point_shape", "torch", "tuple", "vision_config", "weight", "where", "zeros", "zeros_like"], "sam/modeling_sam.py:SamVisionAttention": ["F", "Input", "Linear", "ModelVisionAttention", "Module", "None", "Parameter", "Tensor", "ValueError", "_", "__init__", "arange", "attention_dropout", "attn_output", "attn_probs", "attn_weights", "batch_size", "be", "bhwc", "bhwk", "bias", "class", "config", "decomposed_rel_pos", "def", "dim", "dropout", "dtype", "einsum", "else", "encoding", "float32", "forward", "functional", "get_decomposed_rel_pos", "get_rel_pos", "head_dim", "height", "hidden_size", "hidden_states", "hkc", "if", "image_size", "input_size", "int", "interpolate", "is", "k_coords", "k_size", "key", "key_height", "key_width", "linear", "long", "max", "max_rel_dist", "mode", "must", "nn", "num_attention_heads", "output_attentions", "p", "patch_size", "permute", "positional", "proj", "provided", "q_coords", "q_size", "qkv", "qkv_bias", "query", "query_height", "query_width", "raise", "rel_h", "rel_pos", "rel_pos_h", "rel_pos_resized", "rel_pos_w", "rel_w", "relative", "relative_coords", "relative_position_height", "relative_position_width", "reshape", "reshape_as", "reshaped_query", "return", "scale", "self", "shape", "size", "softmax", "super", "to", "torch", "training", "transpose", "tuple", "unbind", "use_rel_pos", "using", "value", "width", "window_size", "wkc", "zeros"], "sam/modeling_sam.py:SamVisionSdpaAttention": ["Falling", "False", "ModelVisionAttention", "ModelVisionSdpaAttention", "None", "Tensor", "This", "Transformers", "True", "_", "__init__", "argument", "attention", "attn_bias", "attn_implementation", "attn_mask", "attn_output", "back", "batch_size", "be", "but", "can", "class", "config", "decomposed_rel_pos", "def", "does", "eager", "forward", "from", "functional", "get_decomposed_rel_pos", "height", "hidden_states", "if", "implementation", "is", "key", "loading", "logger", "manual", "model", "nn", "not", "num_attention_heads", "onwards", "output_attentions", "permute", "proj", "qkv", "query", "rel_pos_h", "rel_pos_w", "removed", "required", "reshape", "return", "scaled_dot_product_attention", "self", "shape", "specifying", "super", "support", "the", "to", "torch", "unbind", "use_rel_pos", "used", "using", "v5", "value", "version", "view", "warning", "warning_once", "when", "width", "will", "window_size"], "sam/modeling_sam.py:SamVisionLayer": ["F", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelMLPBlock", "ModelVisionLayer", "Model_VISION_ATTENTION_CLASSES", "Tensor", "__init__", "_attn_implementation", "attn", "attn_weights", "batch_size", "channel", "class", "config", "contiguous", "def", "eps", "forward", "height", "hidden_size", "hidden_states", "if", "int", "layer_norm1", "layer_norm2", "layer_norm_eps", "layernorm_output", "mlp", "nn", "original_shape", "pad", "pad_h", "pad_height", "pad_w", "pad_width", "padding_shape", "permute", "reshape", "residual", "return", "self", "shape", "super", "torch", "tuple", "width", "window_partition", "window_size", "window_unpartition", "windows"], "sam/modeling_sam.py:SamVisionNeck": ["Conv2d", "False", "ModelLayerNorm", "ModelVisionConfig", "ModelVisionNeck", "Module", "__init__", "bias", "channels_first", "class", "config", "conv1", "conv2", "data_format", "def", "forward", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "nn", "output_channels", "padding", "permute", "return", "self", "super"], "sam/modeling_sam.py:SamPreTrainedModel": ["Model", "ModelConfig", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionEncoder", "Module", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "class", "config", "data", "def", "elif", "if", "isinstance", "main_input_name", "module", "nn", "pixel_values", "pos_embed", "rel_pos_h", "rel_pos_w", "self", "super", "supports_gradient_checkpointing", "use_abs_pos", "use_rel_pos", "zero_"], "sam/modeling_sam.py:SamVisionEncoder": ["False", "FloatTensor", "ModelPatchEmbeddings", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderOutput", "ModelVisionLayer", "ModelVisionNeck", "ModuleList", "None", "Optional", "Parameter", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "append", "attentions", "check_model_inputs", "class", "config", "def", "else", "for", "forward", "get_input_embeddings", "global_attn_indexes", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "i", "if", "image_size", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_module", "layers", "neck", "nn", "not", "num_hidden_layers", "patch_embed", "patch_size", "pixel_values", "pos_embed", "raise", "range", "return", "self", "specify", "super", "to", "torch", "use_abs_pos", "window_size", "zeros"], "sam/modeling_sam.py:SamVisionModel": ["FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderOutput", "ModelVisionModel", "Module", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "__init__", "auto_docstring", "class", "config", "def", "forward", "get_input_embeddings", "kwargs", "main_input_name", "nn", "patch_embed", "pixel_values", "post_init", "return", "self", "super", "torch", "tuple", "vision_encoder"], "sam/modeling_sam.py:SamModel": ["Either", "FloatTensor", "Got", "LongTensor", "ModelConfig", "ModelImageSegmentationOutput", "ModelMaskDecoder", "ModelModel", "ModelPositionalEmbedding", "ModelPreTrainedModel", "ModelPromptEncoder", "ModelTwoWayAttentionBlock", "ModelVisionEncoder", "ModelVisionEncoderOutput", "Modele", "None", "Of", "Only", "Optional", "OutputRecorder", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_can_record_outputs", "_keys_to_ignore_on_load_missing", "_tie_weights", "_tied_weights_keys", "a", "and", "as", "attention_similarity", "attentions", "auto_docstring", "batch", "batch_size", "be", "bool", "bounding", "box", "box_batch_size", "boxes", "can", "check_model_inputs", "class", "config", "cumsum", "data", "def", "dense_embeddings", "dense_prompt_embeddings", "device", "dim", "dtype", "else", "embeddings", "f", "for", "forward", "get_image_embeddings", "get_image_wide_positional_embeddings", "get_input_embeddings", "get_prompt_embeddings", "got", "grid", "hidden_states", "if", "image", "image_embedding_size", "image_embeddings", "image_positional_embeddings", "index", "input", "input_boxes", "input_labels", "input_masks", "input_points", "int", "iou_predictions", "iou_scores", "is", "kwargs", "last_hidden_state", "len", "low_res_masks", "make", "many", "mask_decoder", "mask_decoder_attentions", "mask_decoder_config", "multimask_output", "multiple", "must", "nb_boxes", "nb_points_per_image", "no_grad", "not", "num_points_per_image", "of", "one", "ones", "ones_like", "or", "pass", "passed", "per", "permute", "pixel_values", "point_batch_size", "points", "positional_embedding", "post_init", "pred_masks", "prompt_encoder", "prompt_encoder_config", "prompt_output", "provide", "provided", "r", "raise", "repeat", "respectively", "return", "self", "shape", "shared_embedding", "shared_image_embedding", "should", "size", "sparse_embeddings", "sparse_prompt_embeddings", "stack", "super", "sure", "target_device", "target_dtype", "target_embedding", "tensor", "that", "the", "to", "torch", "unsqueeze", "vision_attentions", "vision_config", "vision_encoder", "vision_hidden_states", "vision_output", "vision_outputs", "want", "x_embed", "y_embed", "you"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridBaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridLayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLSamVisionNeck": ["Conv2d", "False", "ModelLayerNorm", "ModelVLSamVisionNeck", "Module", "__init__", "bias", "channels_first", "class", "config", "conv1", "conv2", "data_format", "def", "forward", "hidden_size", "hidden_states", "kernel_size", "layer_norm1", "layer_norm2", "nn", "output_channels", "padding", "permute", "return", "self", "super"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLSamVisionProj": ["Conv2d", "False", "ModelVLSamVisionProj", "Module", "Tensor", "__init__", "align_corners", "bias", "bilinear", "class", "config", "conv1", "conv2", "def", "features", "forward", "functional", "int", "interpolate", "kernel_size", "mode", "nn", "output_channels", "output_size", "padding", "return", "self", "size", "stride", "super", "torch"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridAligner": ["GELU", "Linear", "ModelAligner", "ModelConfig", "Module", "Tensor", "__init__", "act", "class", "concat", "config", "def", "dim", "encodings", "forward", "hidden_size", "high_res_in_channels", "high_res_vision_config", "high_res_vision_encodings", "high_res_vision_proj", "in_channels", "nn", "out_channels", "output_channels", "proj", "return", "self", "super", "text_config", "torch", "vision_config", "vision_encodings", "vision_proj"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridPreTrainedModel": ["Conv2d", "False", "Linear", "LlamaDecoderLayer", "ModelConfig", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_param_buffer_assignment", "_supports_sdpa", "base_model_prefix", "bias", "causal_mask", "class", "config", "data", "def", "elif", "fan_out", "fill_", "high_res_vision_alpha", "if", "init", "initializer_range", "is", "isinstance", "kaiming_normal_", "mean", "mode", "model", "module", "nn", "nonlinearity", "normal_", "not", "past_key_values", "relu", "self", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridModel": ["AutoModel", "Both", "Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelAligner", "ModelBaseModelOutputWithPast", "ModelModel", "ModelPreTrainedModel", "ModelVLSamVisionNeck", "ModelVLSamVisionProj", "Model_VL_COMMON_CUSTOM_ARGS", "None", "Optional", "Parameter", "Tensor", "True", "Union", "ValueError", "You", "__init__", "aligner", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "be", "bool", "both", "cache_position", "can_return_tuple", "cannot", "class", "config", "custom_args", "def", "device", "do", "dtype", "either", "else", "expand_as", "f", "features", "forward", "from_config", "get_high_res_image_features", "get_image_features", "get_input_embeddings", "get_low_res_image_features", "get_placeholder_mask", "global_attn_index", "global_attn_indexes", "global_hidden_state", "gradient_checkpointing", "hidden_states", "high_res_pixel_values", "high_res_vision_alpha", "high_res_vision_config", "high_res_vision_encodings", "high_res_vision_model", "high_res_vision_neck", "high_res_vision_proj", "if", "image", "image_attention_mask", "image_embeds", "image_features", "image_hidden_states", "image_size", "image_token_id", "images_embeds", "input_ids", "inputs_embeds", "int", "is", "kwargs", "language_model", "last_hidden_state", "lm_output", "logits_to_keep", "long", "masked_scatter", "match", "must", "n_image_features", "n_image_tokens", "nn", "not", "numel", "one", "output", "output_hidden_states", "output_size", "past_key_values", "patch_size", "permute", "pixel_values", "position_ids", "post_init", "raise", "reshape", "return", "return_dict", "same", "self", "set_input_embeddings", "shape", "should", "special_image_mask", "specified", "specify", "sum", "super", "tensor", "text_config", "the", "time", "to", "tokens", "torch", "unsqueeze", "use_cache", "value", "vision_config", "vision_encodings", "vision_model", "zeros"], "deepseek_vl_hybrid/modeling_deepseek_vl_hybrid.py:DeepseekVLHybridForConditionalGeneration": ["AttributeError", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Model_VL_COMMON_CUSTOM_ARGS", "None", "Not", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "custom_args", "def", "else", "embed_tokens", "for", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "high_res_pixel_values", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "needed", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_embeddings_for_image_generation", "prepare_inputs_for_generation", "r", "raise", "return", "self", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "use_cache", "value", "vocab_size", "weight"], "markuplm/modeling_markuplm.py:XPathEmbeddings": ["Dropout", "Embedding", "Linear", "Model", "Module", "ModuleList", "None", "ReLU", "_", "__init__", "activation", "append", "cat", "class", "config", "def", "dim", "dropout", "for", "forward", "hidden_dropout_prob", "hidden_size", "i", "in", "inner2emb", "max_depth", "max_xpath_subs_unit_embeddings", "max_xpath_tag_unit_embeddings", "nn", "range", "return", "self", "super", "torch", "xpath_embeddings", "xpath_subs_embeddings", "xpath_subs_seq", "xpath_subs_sub_embeddings", "xpath_tag_sub_embeddings", "xpath_tags_embeddings", "xpath_tags_seq", "xpath_unit_hidden_size", "xpath_unitseq2_embeddings", "xpath_unitseq2_inner"], "markuplm/modeling_markuplm.py:MarkupLMEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "XPathEmbeddings", "__init__", "arange", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "list", "long", "mask", "max_depth", "max_position_embeddings", "ne", "nn", "not", "ones", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "sequence_length", "size", "staticmethod", "subs_pad_id", "super", "tag_pad_id", "token_type_embeddings", "token_type_ids", "torch", "tuple", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "words_embeddings", "xpath_embeddings", "xpath_subs_seq", "xpath_tags_seq", "zeros"], "markuplm/modeling_markuplm.py:MarkupLMSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "markuplm/modeling_markuplm.py:MarkupLMIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "markuplm/modeling_markuplm.py:MarkupLMOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "markuplm/modeling_markuplm.py:MarkupLMPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "markuplm/modeling_markuplm.py:MarkupLMPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "markuplm/modeling_markuplm.py:MarkupLMLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "markuplm/modeling_markuplm.py:MarkupLMOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "markuplm/modeling_markuplm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "markuplm/modeling_markuplm.py:MarkupLMSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "bool", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_states", "kwargs", "multiple", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "query", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_states", "view"], "markuplm/modeling_markuplm.py:MarkupLMAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "markuplm/modeling_markuplm.py:MarkupLMLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "markuplm/modeling_markuplm.py:MarkupLMEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "markuplm/modeling_markuplm.py:MarkupLMPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "Optional", "PathLike", "PreTrainedModel", "Union", "_init_weights", "base_model_prefix", "bias", "class", "classmethod", "cls", "config", "data", "def", "elif", "fill_", "from_pretrained", "if", "initializer_range", "is", "isinstance", "kwargs", "mean", "model_args", "module", "nn", "normal_", "not", "os", "padding_idx", "pretrained_model_name_or_path", "return", "self", "std", "str", "super", "weight", "zero_"], "markuplm/modeling_markuplm.py:MarkupLMModel": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "both", "can_return_tuple", "cannot", "class", "config", "def", "device", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "for", "forward", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "next", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "parameters", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "unsqueeze", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "xpath_subs_seq", "xpath_tags_seq", "zeros"], "markuplm/modeling_markuplm.py:MarkupLMForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "True", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "clamp_", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "xpath_subs_seq", "xpath_tags_seq"], "markuplm/modeling_markuplm.py:MarkupLMForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "MaskedLMOutput", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "True", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "xpath_subs_seq", "xpath_tags_seq"], "markuplm/modeling_markuplm.py:MarkupLMForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "True", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "xpath_subs_seq", "xpath_tags_seq"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionModelOutputWithPooling": ["BaseModelOutputWithPooling", "ModelVisionModelOutputWithPooling", "class", "r"], "data2vec/modeling_data2vec_vision.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionDropPath": ["ModelVisionDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionEmbeddings": ["BEiTEmbeddings", "BoolTensor", "Dropout", "False", "Iterable", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "The", "_", "__init__", "abc", "align_corners", "always", "and", "are", "argument", "batch_size", "be", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "collections", "config", "def", "dim", "dropout", "effect", "else", "embeddings", "expand", "for", "forward", "functional", "has", "height", "hidden_dropout_prob", "hidden_size", "if", "image", "image_size", "in", "input", "int", "interpolate", "interpolate_pos_encoding", "interpolated", "is", "is_tracing", "isinstance", "jit", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "no", "not", "num_patches", "num_positions", "patch_embeddings", "patch_height", "patch_pos_embed", "patch_size", "patch_width", "permute", "pixel_values", "position_embeddings", "removed", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "the", "to", "torch", "torch_int", "transformers", "type_as", "unsqueeze", "use_absolute_position_embeddings", "use_mask_token", "v4", "view", "w", "warn", "warnings", "width", "will", "zeros"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionPatchEmbeddings": ["Conv2d", "Iterable", "Make", "ModelVisionPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "flatten", "forward", "height", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_height", "patch_shape", "patch_size", "patch_width", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "width", "with"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionSelfAttention": ["Dropout", "False", "Linear", "ModelVisionConfig", "ModelVisionRelativePositionBias", "ModelVisionSelfAttention", "Module", "None", "Optional", "Tensor", "The", "Union", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dim_size", "dropout", "else", "embedding_size", "f", "forward", "functional", "has_relative_position_bias", "hasattr", "head_mask", "heads", "height", "hidden", "hidden_size", "hidden_states", "if", "int", "interpolate_pos_encoding", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "patch_size", "permute", "query", "query_layer", "raise", "relative_position_bias", "resolution", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "view", "width", "window_size"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionSdpaSelfAttention": ["Falling", "False", "ModelVisionSdpaSelfAttention", "ModelVisionSelfAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "Union", "_", "all_head_size", "argument", "attention", "attention_head_size", "attention_probs_dropout_prob", "attn_bias", "attn_implementation", "attn_mask", "back", "batch_size", "be", "bool", "but", "can", "class", "config", "context_layer", "contiguous", "def", "dim_size", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "has_relative_position_bias", "head_mask", "height", "hidden_states", "if", "implementation", "int", "interpolate_pos_encoding", "is", "is_causal", "key", "key_layer", "loading", "logger", "manual", "math", "model", "new_context_layer_shape", "nn", "not", "num_attention_heads", "onwards", "or", "output_attentions", "patch_size", "permute", "query", "query_layer", "relative_position_bias", "removed", "required", "resolution", "return", "scale", "scaled_dot_product_attention", "scaling", "self", "seq_length", "shape", "size", "specifying", "sqrt", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "used", "using", "v5", "value", "value_layer", "version", "view", "warning", "warning_once", "when", "width", "will", "window_size"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionSelfOutput": ["Dropout", "Linear", "ModelVisionConfig", "ModelVisionSelfOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "gamma", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionAttention": ["False", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionRelativePositionBias", "ModelVisionSelfOutput", "Model_VISION_SELF_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_attn_implementation", "all_head_size", "attention", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "interpolate_pos_encoding", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "relative_position_bias", "resolution", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionIntermediate": ["ACT2FN", "Linear", "ModelVisionConfig", "ModelVisionIntermediate", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionOutput": ["Dropout", "Linear", "ModelVisionConfig", "ModelVisionOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionLayer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionDropPath", "ModelVisionIntermediate", "ModelVisionLayer", "ModelVisionOutput", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "attention", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "float", "forward", "head_mask", "hidden_size", "hidden_states", "if", "init_values", "int", "intermediate", "interpolate_pos_encoding", "is", "lambda_1", "lambda_2", "layer_norm_eps", "layer_output", "layer_scale_init_value", "layernorm_after", "layernorm_before", "nn", "not", "ones", "output", "output_attentions", "outputs", "relative_position_bias", "requires_grad", "resolution", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple", "window_size"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionRelativePositionBias": ["False", "ModelVisionConfig", "ModelVisionRelativePositionBias", "Module", "None", "Parameter", "Tensor", "__init__", "align_corners", "arange", "bilinear", "bool", "cat", "class", "compile_compatible_method_lru_cache", "config", "contiguous", "coords", "coords_flatten", "def", "dim_size", "dtype", "flatten", "forward", "functional", "generate_relative_position_index", "grid", "if", "ij", "indexing", "int", "interpolate", "interpolate_pos_encoding", "maxsize", "meshgrid", "mode", "new_height", "new_num_relative_distance", "new_relative_position_bias_table", "new_sub_table", "new_width", "nn", "num_attention_heads", "num_relative_distance", "old_height", "old_num_relative_distance", "old_relative_position_bias_table", "old_sub_table", "old_width", "permute", "relative_coords", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "reshape", "return", "self", "size", "squeeze", "stack", "sum", "super", "torch", "torch_int", "tuple", "unsqueeze", "view", "window_area", "window_size", "zeros"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionLayer", "ModelVisionRelativePositionBias", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "bool", "class", "config", "cpu", "def", "device", "dim_size", "dpr", "drop_path_rate", "else", "enumerate", "for", "forward", "gradient_checkpointing", "has_relative_position_bias", "head_mask", "height", "hidden_states", "i", "if", "in", "int", "interpolate_pos_encoding", "is", "item", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "linspace", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_size", "range", "relative_position_bias", "resolution", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "use_relative_position_bias", "use_shared_relative_position_bias", "v", "width", "window_size", "x"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionPreTrainedModel": ["Conv2d", "ConvTranspose2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionLayer", "ModelVisionPreTrainedModel", "ModelVisionRelativePositionBias", "Model_vision", "None", "PreTrainedModel", "True", "_init_weights", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "bias", "class", "cls_token", "config", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "lambda_1", "lambda_2", "layer_scale_init_value", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "position_embeddings", "r", "relative_position_bias_table", "relative_position_index", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionModel": ["BoolTensor", "False", "Identity", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionModel", "ModelVisionModelOutputWithPooling", "ModelVisionPooler", "ModelVisionPreTrainedModel", "None", "Optional", "Tensor", "Union", "_", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "head_mask", "head_outputs", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_shape", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "resolution", "return", "return_dict", "self", "sequence_output", "shape", "super", "torch", "tuple", "use_mean_pooling", "use_return_dict", "window_size"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionPooler": ["LayerNorm", "ModelVisionConfig", "ModelVisionPooler", "Module", "None", "Tensor", "__init__", "class", "config", "def", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "is", "layer_norm_eps", "layernorm", "mean", "nn", "not", "patch_tokens", "pooled_output", "return", "self", "super", "torch", "use_mean_pooling"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionForImageClassification": ["False", "Identity", "ImageClassifierOutput", "Linear", "ModelVisionConfig", "ModelVisionForImageClassification", "ModelVisionModel", "ModelVisionPreTrainedModel", "Model_vision", "None", "Optional", "Tensor", "True", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionConvModule": ["BatchNorm2d", "Conv2d", "False", "ModelVisionConvModule", "Module", "None", "ReLU", "Tensor", "Union", "__init__", "activation", "bias", "bn", "bool", "class", "conv", "def", "dilation", "forward", "in_channels", "input", "int", "kernel_size", "nn", "out_channels", "output", "padding", "return", "self", "str", "super", "torch", "tuple"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionPyramidPoolingBlock": ["AdaptiveAvgPool2d", "ModelVisionConvModule", "ModelVisionPyramidPoolingBlock", "Module", "None", "Tensor", "__init__", "add_module", "channels", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_channels", "input", "int", "kernel_size", "layer", "layers", "nn", "pool_scale", "return", "self", "str", "super", "torch"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionPyramidPoolingModule": ["ModelVisionPyramidPoolingBlock", "ModelVisionPyramidPoolingModule", "Module", "None", "Tensor", "__init__", "add_module", "align_corners", "append", "bilinear", "block", "blocks", "bool", "channels", "class", "def", "enumerate", "for", "forward", "functional", "i", "in", "in_channels", "int", "interpolate", "list", "mode", "nn", "pool_scale", "pool_scales", "ppm", "ppm_out", "ppm_outs", "return", "self", "size", "str", "super", "torch", "tuple", "upsampled_ppm_out", "x"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionUperHead": ["Conv2d", "False", "ModelVisionConfig", "ModelVisionConvModule", "ModelVisionPyramidPoolingModule", "ModelVisionUperHead", "Module", "ModuleList", "None", "Tensor", "__init__", "align_corners", "append", "bilinear", "bottleneck", "cat", "channels", "class", "classifier", "config", "def", "dim", "encoder_hidden_states", "enumerate", "extend", "for", "forward", "fpn_bottleneck", "fpn_conv", "fpn_convs", "fpn_outs", "functional", "hidden_size", "i", "in", "in_channels", "inputs", "interpolate", "kernel_size", "l_conv", "lateral_conv", "lateral_convs", "laterals", "len", "mode", "nn", "num_labels", "output", "padding", "pool_scales", "prev_shape", "psp_forward", "psp_modules", "psp_outs", "range", "return", "self", "shape", "size", "super", "torch", "used_backbone_levels", "x"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionFCNHead": ["Conv2d", "Identity", "ModelVisionConfig", "ModelVisionConvModule", "ModelVisionFCNHead", "Module", "None", "Sequential", "Tensor", "Union", "__init__", "append", "auxiliary_channels", "auxiliary_concat_input", "auxiliary_num_convs", "cat", "channels", "class", "classifier", "concat_input", "config", "conv_cat", "conv_padding", "convs", "def", "dilation", "dim", "else", "encoder_hidden_states", "for", "forward", "hidden_size", "hidden_states", "i", "if", "in", "in_channels", "in_index", "int", "kernel_size", "nn", "num_convs", "num_labels", "output", "padding", "range", "return", "self", "super", "torch", "tuple"], "data2vec/modeling_data2vec_vision.py:Data2VecVisionForSemanticSegmentation": ["BatchNorm2d", "ConvTranspose2d", "CrossEntropyLoss", "False", "GELU", "Identity", "MaxPool2d", "ModelVisionConfig", "ModelVisionFCNHead", "ModelVisionForSemanticSegmentation", "ModelVisionModel", "ModelVisionPreTrainedModel", "ModelVisionUperHead", "Model_vision", "None", "One", "Optional", "SemanticSegmenterOutput", "Sequential", "Tensor", "The", "True", "Union", "ValueError", "__init__", "a", "add_pooling_layer", "align_corners", "and", "architecture", "attentions", "auto_docstring", "auxiliary_head", "auxiliary_logits", "auxiliary_loss", "auxiliary_loss_weight", "backbone", "base", "batch_size", "be", "bilinear", "bool", "can", "case", "class", "compute_loss", "config", "decode_head", "def", "else", "encoder_hidden_states", "enumerate", "feature", "features", "for", "forward", "fpn1", "fpn2", "fpn3", "fpn4", "from", "functional", "greater", "head_mask", "hidden_size", "hidden_states", "i", "idx", "if", "ignore_index", "image_size", "in", "integers", "interpolate", "interpolate_pos_encoding", "is", "kernel_size", "labels", "len", "list", "logits", "loss", "loss_fct", "main_loss", "mode", "nn", "not", "num_labels", "number", "of", "one", "ops", "out_indices", "output", "output_attentions", "output_hidden_states", "outputs", "patch_resolution", "patch_size", "permute", "pixel_values", "post_init", "r", "raise", "range", "requires", "reshape", "return", "return_dict", "self", "semantic_loss_ignore_index", "shape", "should", "size", "sized", "specifying", "stride", "super", "than", "the", "to", "torch", "tuple", "upsampled_auxiliary_logits", "upsampled_logits", "use", "use_auxiliary_head", "use_return_dict", "which", "x"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelAudioConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioPadLayer": ["ModelAudioPadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioPositionalConvLayer": ["ACT2FN", "Conv1d", "False", "LayerNorm", "ModelAudioPadLayer", "ModelAudioPositionalConvLayer", "Module", "__init__", "activation", "class", "config", "conv", "conv_pos_kernel_size", "def", "elementwise_affine", "feat_extract_activation", "forward", "groups", "hidden_size", "hidden_states", "kernel_size", "layer_norm", "nn", "num_conv_pos_embedding_groups", "padding", "return", "self", "super", "transpose"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioPositionalConvEmbedding": ["ModelAudioPositionalConvEmbedding", "ModelAudioPositionalConvLayer", "Module", "ModuleList", "_", "__init__", "class", "config", "def", "for", "forward", "hidden_states", "in", "layer", "layers", "nn", "num_conv_pos_embeddings", "range", "return", "self", "super", "transpose"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioFeatureEncoder": ["False", "ModelAudioConvLayer", "ModelAudioFeatureEncoder", "Module", "ModuleList", "None", "True", "__init__", "_freeze_parameters", "_requires_grad", "and", "class", "config", "conv_layer", "conv_layers", "def", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "input_values", "layer_id", "nn", "num_feat_extract_layers", "param", "parameters", "range", "requires_grad", "return", "self", "super", "training"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelAudioFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "data2vec/modeling_data2vec_audio.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAudioAttention", "ModelAudioConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelAudioFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAudioAttention", "ModelAudioEncoderLayer", "ModelAudioFeedForward", "None", "_", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelAudioEncoder", "ModelAudioEncoderLayer", "ModelAudioPositionalConvEmbedding", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "tensor", "torch", "training", "tuple", "unsqueeze", "v"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioAdapterLayer": ["Conv1d", "ModelAudioAdapterLayer", "Module", "__init__", "adapter_kernel_size", "adapter_stride", "class", "config", "conv", "def", "dim", "forward", "functional", "glu", "hidden_states", "nn", "output_hidden_size", "padding", "return", "self", "stride", "super"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioAdapter": ["LayerNorm", "Linear", "ModelAudioAdapter", "ModelAudioAdapterLayer", "Module", "ModuleList", "None", "_", "__init__", "and", "class", "config", "def", "else", "for", "forward", "hidden_size", "hidden_states", "if", "in", "is", "layer", "layerdrop", "layerdrop_prob", "layers", "nn", "not", "np", "num_adapter_layers", "or", "output_hidden_size", "proj", "proj_layer_norm", "random", "range", "return", "self", "super", "training", "transpose"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioPreTrainedModel": ["Conv1d", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelAudioConfig", "ModelAudioFeatureProjection", "ModelAudioPositionalConvLayer", "ModelAudioPreTrainedModel", "Model_audio", "None", "Optional", "PreTrainedModel", "True", "Union", "_", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "adapter_stride", "add_adapter", "arange", "attention_mask", "b", "base_model_prefix", "batch_size", "bias", "bool", "class", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "def", "device", "dim", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "groups", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "module", "nn", "non_padded_lengths", "normal_", "not", "num_adapter_layers", "output_lengths", "projection", "range", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "supports_gradient_checkpointing", "to", "torch", "uniform_", "weight", "zero_", "zeros", "zip"], "data2vec/modeling_data2vec_audio.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioModel": ["False", "FloatTensor", "LongTensor", "ModelAudioAdapter", "ModelAudioBaseModelOutput", "ModelAudioConfig", "ModelAudioEncoder", "ModelAudioFeatureEncoder", "ModelAudioFeatureProjection", "ModelAudioModel", "ModelAudioPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_freeze_parameters", "_get_feature_vector_attention_mask", "_mask_hidden_states", "adapter", "add_adapter", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "dtype", "elif", "else", "encoder", "encoder_outputs", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "freeze_feature_encoder", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "uniform_", "use_return_dict"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioForCTC": ["CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "ModelAudioForCTC", "ModelAudioModel", "ModelAudioPreTrainedModel", "Model_audio", "None", "Optional", "Please", "Tensor", "The", "Transformers", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "define", "deprecated", "dim", "does", "dropout", "dtype", "else", "enabled", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "forward", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "log_probs", "log_softmax", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "post_init", "r", "raise", "reduction", "removed", "return", "return_dict", "s", "self", "size", "sum", "super", "target_lengths", "that", "the", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "ModelAudio", "ModelAudioForSequenceClassification", "ModelAudioModel", "ModelAudioPreTrainedModel", "Model_audio", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioForAudioFrameClassification": ["Audio", "CrossEntropyLoss", "False", "FutureWarning", "Linear", "ModelAudio", "ModelAudioForAudioFrameClassification", "ModelAudioModel", "ModelAudioPreTrainedModel", "Model_audio", "None", "Optional", "Parameter", "Please", "Tensor", "The", "TokenClassifierOutput", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "adapters", "add_adapter", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "axis", "be", "bool", "class", "classification", "classifier", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "feature_extractor", "for", "forward", "frame", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "init_weights", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "param", "parameters", "r", "raise", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "data2vec/modeling_data2vec_audio.py:AMSoftmaxLoss": ["CrossEntropyLoss", "Model", "Module", "Parameter", "True", "__init__", "bool", "class", "cos_theta", "def", "dim", "flatten", "forward", "functional", "hidden_states", "input_dim", "labels", "logits", "loss", "margin", "mm", "nn", "normalize", "num_labels", "one_hot", "onehot", "psi", "randn", "requires_grad", "return", "scale", "self", "super", "torch", "weight", "where"], "data2vec/modeling_data2vec_audio.py:TDNNLayer": ["Detected", "Linear", "LoRA", "LoraLayer", "Model", "Module", "ReLU", "Tensor", "You", "__init__", "activation", "applied", "be", "bias", "class", "config", "conv1d", "def", "dilation", "due", "else", "exclude", "forward", "from", "functional", "hidden_states", "if", "in_conv_dim", "is_peft_available", "isinstance", "kernel", "kernel_size", "layer_id", "modules", "nn", "on", "optimization", "out_conv_dim", "return", "s", "self", "should", "super", "t", "target", "tdnn_dilation", "tdnn_dim", "tdnn_kernel", "to", "torch", "transpose", "view", "warn", "warnings", "weight", "weights", "won"], "data2vec/modeling_data2vec_audio.py:Data2VecAudioForXVector": ["AMSoftmaxLoss", "False", "FutureWarning", "Linear", "LongTensor", "ModelAudioForXVector", "ModelAudioModel", "ModelAudioPreTrainedModel", "Model_audio", "ModuleList", "None", "Optional", "Parameter", "Please", "TDNNLayer", "Tensor", "The", "Transformers", "True", "Union", "XVectorOutput", "_HIDDEN_STATES_START_POSITION", "__init__", "_conv_out_length", "_freeze_parameters", "_get_feat_extract_output_lengths", "_get_tdnn_output_lengths", "and", "append", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cat", "class", "classifier", "config", "def", "deprecated", "dim", "else", "embeddings", "enumerate", "equivalent", "feat_extract_output_lengths", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hidden_size", "hidden_states", "i", "if", "in", "init_weights", "input_length", "input_lengths", "input_values", "instead", "int", "is", "kernel_size", "labels", "layer_weights", "len", "length", "logits", "loss", "mean", "mean_features", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "objective", "ones", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "parameters", "projector", "r", "range", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "statistic_pooling", "std", "std_features", "stride", "sum", "super", "tdnn", "tdnn_dim", "tdnn_kernel", "tdnn_layer", "tdnn_layers", "tdnn_output_lengths", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will", "xvector_output_dim"], "data2vec/modeling_data2vec_text.py:Data2VecTextEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "data2vec/modeling_data2vec_text.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "data2vec/modeling_data2vec_text.py:Data2VecTextSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelTextSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "data2vec/modeling_data2vec_text.py:Data2VecTextCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelTextCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "data2vec/modeling_data2vec_text.py:Data2VecTextSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "data2vec/modeling_data2vec_text.py:Data2VecTextAttention": ["Cache", "False", "FloatTensor", "ModelTextAttention", "ModelTextCrossAttention", "ModelTextSelfAttention", "ModelTextSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "data2vec/modeling_data2vec_text.py:Data2VecTextIntermediate": ["ACT2FN", "Linear", "ModelTextIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "data2vec/modeling_data2vec_text.py:Data2VecTextOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "data2vec/modeling_data2vec_text.py:Data2VecTextLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelTextAttention", "ModelTextIntermediate", "ModelTextLayer", "ModelTextOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "data2vec/modeling_data2vec_text.py:Data2VecTextPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelTextConfig", "ModelTextCrossAttention", "ModelTextForTextEmbeddings", "ModelTextLayer", "ModelTextPreTrainedModel", "ModelTextSelfAttention", "Model_text", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "def", "elif", "fill_", "hasattr", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "data2vec/modeling_data2vec_text.py:Data2VecTextEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelTextEncoder", "ModelTextLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "data2vec/modeling_data2vec_text.py:Data2VecTextPooler": ["Linear", "ModelTextPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "data2vec/modeling_data2vec_text.py:Data2VecTextModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextLayer", "ModelTextModel", "ModelTextPooler", "ModelTextPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "data2vec/modeling_data2vec_text.py:Data2VecTextLMHead": ["LayerNorm", "Linear", "ModelTextLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "data2vec/modeling_data2vec_text.py:Data2VecTextClassificationHead": ["Dropout", "Linear", "ModelTextClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "data2vec/modeling_data2vec_text.py:Data2VecTextForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "ModelTextForCausalLM", "ModelTextLMHead", "ModelTextLMHeadModel", "ModelTextModel", "ModelTextPreTrainedModel", "Model_text", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "data2vec/modeling_data2vec_text.py:Data2VecTextForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "ModelTextForMaskedLM", "ModelTextLMHead", "ModelTextModel", "ModelTextPreTrainedModel", "Model_text", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "device", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "data2vec/modeling_data2vec_text.py:Data2VecTextForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "ModelTextClassificationHead", "ModelTextForSequenceClassification", "ModelTextModel", "ModelTextPreTrainedModel", "Model_text", "None", "Optional", "SequenceClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "view"], "data2vec/modeling_data2vec_text.py:Data2VecTextForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "ModelTextForMultipleChoice", "ModelTextModel", "ModelTextPreTrainedModel", "Model_text", "MultipleChoiceModelOutput", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "to", "token_type_ids", "torch", "tuple", "view"], "data2vec/modeling_data2vec_text.py:Data2VecTextForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelTextForTokenClassification", "ModelTextModel", "ModelTextPreTrainedModel", "Model_text", "None", "Optional", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "view"], "data2vec/modeling_data2vec_text.py:Data2VecTextForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelTextForQuestionAnswering", "ModelTextModel", "ModelTextPreTrainedModel", "Model_text", "None", "Optional", "QuestionAnsweringModelOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingLayer": ["Conv2d", "ModelConfig", "ModelLayer", "Model_depth", "Module", "ReLU", "Tensor", "True", "__init__", "activation1", "activation2", "bias", "class", "config", "convolution1", "convolution2", "convolution3", "def", "forward", "fusion_hidden_size", "hidden_state", "kernel_size", "nn", "padding", "return", "self", "stride", "super", "torch"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingPreActResidualLayer": ["Conv2d", "ModelPreActResidualLayer", "Module", "ReLU", "Tensor", "True", "__init__", "activation1", "activation2", "bias", "class", "config", "convolution1", "convolution2", "def", "forward", "fusion_hidden_size", "hidden_state", "kernel_size", "nn", "padding", "residual", "return", "self", "stride", "super", "torch"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingFeatureFusionLayer": ["Conv2d", "False", "ModelConfig", "ModelFeatureFusionLayer", "ModelLayer", "ModelPreActResidualLayer", "Model_depth", "Model_depth_layer", "Module", "None", "True", "__init__", "align_corners", "bias", "bilinear", "class", "config", "def", "else", "forward", "functional", "fusion_hidden_size", "hidden_state", "if", "interpolate", "is", "kernel_size", "mode", "modifier", "nn", "not", "projection", "res", "residual", "residual_layer1", "residual_layer2", "return", "scale_factor", "self", "shape", "size", "super"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingFeatureFusionStage": ["ModelConfig", "ModelFeatureFusionLayer", "ModelFeatureFusionStage", "Model_depth", "Module", "ModuleList", "None", "_", "__init__", "append", "class", "config", "def", "else", "enumerate", "for", "forward", "fused_hidden_state", "fused_hidden_states", "hidden_state", "hidden_states", "idx", "if", "in", "is", "layer", "layers", "len", "neck_hidden_sizes", "nn", "range", "return", "self", "shape", "size", "super", "zip"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingDepthEstimationHead": ["Conv2d", "ModelDepthEstimationHead", "Module", "ReLU", "Sigmoid", "Tensor", "True", "Unknown", "ValueError", "__init__", "activation1", "activation2", "align_corners", "bilinear", "class", "config", "conv1", "conv2", "conv3", "def", "depth", "depth_estimation_type", "dim", "elif", "else", "estimation", "f", "features", "forward", "functional", "fusion_hidden_size", "head_hidden_size", "head_in_index", "hidden_states", "if", "int", "interpolate", "kernel_size", "list", "max_depth", "metric", "mode", "nn", "padding", "patch_height", "patch_size", "patch_width", "predicted_depth", "raise", "relative", "return", "self", "squeeze", "stride", "super", "target_height", "target_width", "torch", "torch_int", "type"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingPreTrainedModel": ["Model", "ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "base_model_prefix", "class", "config", "main_input_name", "pixel_values", "supports_gradient_checkpointing"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingReassembleLayer": ["Conv2d", "ConvTranspose2d", "Identity", "ModelConfig", "ModelReassembleLayer", "Module", "__init__", "channels", "class", "config", "def", "elif", "factor", "forward", "hidden_state", "if", "in_channels", "int", "kernel_size", "nn", "out_channels", "padding", "projection", "reassemble_hidden_size", "resize", "return", "self", "stride", "super", "torch_int"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingReassembleStage": ["ModelReassembleLayer", "ModelReassembleStage", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "append", "batch_size", "channels", "class", "config", "contiguous", "def", "enumerate", "factor", "for", "forward", "hidden_state", "hidden_states", "i", "in", "layers", "list", "neck_hidden_sizes", "nn", "num_channels", "out", "patch_height", "patch_width", "permute", "reassemble_factors", "reshape", "return", "self", "shape", "super", "torch", "zip"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingNeck": ["Conv2d", "False", "ModelFeatureFusionStage", "ModelNeck", "ModelReassembleStage", "Model_depth", "Module", "ModuleList", "None", "Optional", "Tensor", "The", "TypeError", "ValueError", "__init__", "a", "append", "be", "bias", "channel", "class", "config", "convs", "def", "enumerate", "equal", "feature", "features", "for", "forward", "fusion_hidden_size", "fusion_stage", "hidden", "hidden_states", "i", "if", "in", "int", "isinstance", "kernel_size", "len", "list", "neck", "neck_hidden_sizes", "nn", "not", "number", "of", "or", "output", "padding", "patch_height", "patch_width", "raise", "reassemble_stage", "return", "self", "should", "sizes", "states", "super", "tensors", "the", "to", "torch", "tuple"], "prompt_depth_anything/modeling_prompt_depth_anything.py:PromptDepthAnythingForDepthEstimation": ["DPTViTEmbeddings", "DepthEstimatorOutput", "FloatTensor", "LongTensor", "ModelDepthEstimationHead", "ModelForDepthEstimation", "ModelNeck", "ModelPreTrainedModel", "Model_depth", "None", "NotImplementedError", "Optional", "Tensor", "Training", "Union", "_", "__init__", "_no_split_modules", "attentions", "auto_docstring", "backbone", "batch_size", "bool", "class", "config", "def", "depth_max", "depth_min", "device", "dim", "else", "feature_maps", "forward", "forward_with_filtered_kwargs", "head", "height", "hidden_states", "if", "implemented", "is", "labels", "load_backbone", "loss", "max", "min", "neck", "not", "output", "output_attentions", "output_hidden_states", "outputs", "patch_height", "patch_size", "patch_width", "pixel_values", "post_init", "predicted_depth", "r", "raise", "reshape", "return", "return_dict", "self", "shape", "squeeze", "super", "to", "torch", "tuple", "use_return_dict", "values", "view", "width", "yet"], "modernbert/modeling_modernbert.py:ApplyRotaryEmbUnpad": ["False", "Function", "ModelRotaryEmbUnpad", "Model_rotary", "None", "Optional", "Tensor", "True", "_nheads", "_three", "autograd", "backward", "class", "conjugate", "contiguous", "cos", "ctx", "cu_seqlens", "def", "do", "dqk", "forward", "headdim", "inplace", "int", "interleaved", "max_seqlen", "qk", "qkv", "return", "save_for_backward", "saved_tensors", "seqlen_offsets", "shape", "sin", "staticmethod", "torch", "total_nnz", "view"], "modernbert/modeling_modernbert.py:apply_rotary_unpadded": ["Model", "ModelRotaryEmbUnpad", "Model_rotary_unpadded", "None", "Optional", "Tensor", "cos", "cu_seqlens", "def", "int", "max_seqlen", "qkv", "return", "sin", "torch"], "modernbert/modeling_modernbert.py:ModernBertUnpaddedRotaryEmbedding": ["False", "ModelUnpaddedRotaryEmbedding", "None", "Optional", "RotaryEmbedding", "Tensor", "Union", "__init__", "_cos_cached", "_sin_cached", "_update_cos_sin_cache", "and", "apply_rotary_unpadded", "base", "class", "cu_seqlens", "def", "device", "dim", "dtype", "extra_repr", "f", "float", "forward", "if", "int", "interleaved", "is", "max_seqlen", "not", "qkv", "return", "scale_base", "self", "str", "super", "torch", "tuple"], "modernbert/modeling_modernbert.py:ModernBertEmbeddings": ["Dropout", "Embedding", "LayerNorm", "LongTensor", "ModelConfig", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "True", "__init__", "bias", "class", "compile", "compiled_embeddings", "config", "def", "drop", "dynamic", "else", "embedding_dropout", "eps", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "nn", "norm", "norm_bias", "norm_eps", "not", "pad_token_id", "padding_idx", "reference_compile", "return", "self", "super", "tok_embeddings", "torch", "vocab_size"], "modernbert/modeling_modernbert.py:ModernBertMLP": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "Wi", "Wo", "__init__", "act", "bias", "chunk", "class", "config", "def", "dim", "drop", "forward", "gate", "hidden_activation", "hidden_size", "hidden_states", "input", "int", "intermediate_size", "mlp_bias", "mlp_dropout", "nn", "return", "self", "super", "torch"], "modernbert/modeling_modernbert.py:ModernBertRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "modernbert/modeling_modernbert.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "modernbert/modeling_modernbert.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "modernbert/modeling_modernbert.py:eager_attention_forward": ["False", "LongTensor", "ModelAttention", "Model_attention_forward", "Optional", "Tensor", "Union", "_kwargs", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "bool", "bs", "contiguous", "cos", "def", "dim", "dropout", "dtype", "float32", "functional", "head_dim", "if", "int", "key", "local_attention", "matmul", "module", "nn", "output_attentions", "p", "position_ids", "qkv", "query", "return", "rotary_emb", "scale", "sin", "sliding_window_mask", "softmax", "to", "torch", "training", "transpose", "tuple", "unbind", "value", "view"], "modernbert/modeling_modernbert.py:flash_attention_forward": ["ModelAttention", "ModelUnpaddedRotaryEmbedding", "Model_attention_forward", "Model_attn_varlen_qkvpacked_func", "Tensor", "_kwargs", "attention_dropout", "attn", "bfloat16", "bs", "convert_dtype", "cu_seqlens", "def", "deterministic", "deterministic_Model_attn", "dim", "dropout_p", "dtype", "else", "float16", "if", "in", "int", "local_attention", "max_seqlen", "module", "not", "orig_dtype", "qkv", "return", "rotary_emb", "target_dtype", "to", "torch", "training", "tuple", "view", "window_size"], "modernbert/modeling_modernbert.py:sdpa_attention_forward": ["F", "LongTensor", "ModelAttention", "Model_attention_forward", "Optional", "Tensor", "_kwargs", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_mask", "attn_output", "bs", "contiguous", "cos", "def", "dim", "dropout_p", "else", "if", "int", "key", "local_attention", "module", "position_ids", "qkv", "query", "return", "rotary_emb", "scaled_dot_product_attention", "sin", "sliding_window_mask", "torch", "training", "transpose", "tuple", "unbind", "value", "view"], "modernbert/modeling_modernbert.py:ModernBertAttention": ["Dropout", "False", "Identity", "Linear", "ModelAttention", "ModelConfig", "ModelRotaryEmbedding", "ModelUnpaddedRotaryEmbedding", "Model_ATTENTION_FUNCTION", "Module", "None", "Optional", "Tensor", "The", "ValueError", "Wo", "Wqkv", "__init__", "_attn_implementation", "a", "all_head_size", "attention", "attention_bias", "attention_dropout", "attn_outputs", "base", "bias", "bool", "bs", "class", "config", "config_copy", "copy", "deepcopy", "def", "deterministic_flash_attn", "dim", "else", "f", "flash_attention_2", "forward", "global_attn_every_n_layers", "global_rope_theta", "head_dim", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "kwargs", "layer_id", "local_attention", "local_rope_theta", "max_position_embeddings", "max_seqlen", "multiple", "nn", "not", "num_attention_heads", "num_heads", "number", "of", "out_drop", "output_attentions", "pruned_heads", "qkv", "raise", "return", "rope_theta", "rotary_emb", "self", "set", "shape", "size", "super", "the", "torch", "view"], "modernbert/modeling_modernbert.py:ModernBertEncoderLayer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "None", "Optional", "Tensor", "True", "__init__", "attention_mask", "attn", "attn_norm", "attn_outputs", "bias", "bool", "class", "compile", "compiled_mlp", "config", "cu_seqlens", "def", "dynamic", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "int", "layer_id", "max_seqlen", "mlp", "mlp_norm", "mlp_output", "nn", "norm_bias", "norm_eps", "output_attentions", "position_ids", "reference_compile", "return", "self", "sliding_window_mask", "super", "torch"], "modernbert/modeling_modernbert.py:ModernBertPreTrainedModel": ["Compiling", "Falling", "False", "If", "ImportError", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelEncoderLayer", "ModelForMaskedLM", "ModelForMultipleChoice", "ModelForQuestionAnswering", "ModelForSequenceClassification", "ModelForTokenClassification", "ModelMLP", "ModelPreTrainedModel", "ModelPredictionHead", "Module", "None", "Optional", "PreTrainedModel", "Resizing", "True", "ValueError", "Wi", "Wo", "Wqkv", "_check_and_adjust_attn_implementation", "_flash_attn_2_can_dispatch", "_init_weights", "_maybe_set_compile", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "accelerate", "across", "and", "args", "attn_implementation", "b", "back", "base_model_prefix", "bias", "bool", "class", "classifier", "compile", "compiled", "config", "cpu", "cutoff_factor", "data", "decoder", "def", "dense", "device", "devices", "elif", "else", "embedding", "embeddings", "except", "fill_", "final_out", "flash_attention_2", "float", "hasattr", "hf_device_map", "hidden_size", "if", "in", "init", "init_weight", "initializer_cutoff_factor", "initializer_range", "is", "is_init_check", "is_triton_available", "isinstance", "kwargs", "len", "logger", "math", "mean", "mode", "model", "model_embeds", "module", "mps", "nn", "non", "not", "num_hidden_layers", "out", "pass", "reference_compile", "resize_token_embeddings", "return", "self", "split", "sqrt", "std", "stds", "str", "super", "supported", "supports_gradient_checkpointing", "the", "to", "tok_embeddings", "token", "torch", "trunc_normal_", "try", "type", "using", "warning_once", "weight", "will", "with", "work", "zero_", "zeros_"], "modernbert/modeling_modernbert.py:_unpad_modernbert_input": ["False", "None", "Optional", "Tensor", "_unpad_Model_input", "as_tuple", "attention_mask", "batch", "cu_seqlens", "cumsum", "def", "dim", "dtype", "else", "flatten", "functional", "if", "indices", "inputs", "int", "int32", "is", "item", "labels", "max", "max_seqlen_in_batch", "nn", "nonzero", "not", "pad", "position_ids", "rest", "return", "seqlen", "seqlens_in_batch", "shape", "sum", "torch", "tuple", "unpadded_inputs", "unpadded_labels", "unpadded_position_ids", "view"], "modernbert/modeling_modernbert.py:_pad_modernbert_output": ["Tensor", "_", "_pad_Model_output", "batch", "def", "device", "dim", "dtype", "else", "if", "indices", "inputs", "int", "output", "padded_inputs", "rest", "return", "seqlen", "shape", "torch", "view", "zeros"], "modernbert/modeling_modernbert.py:ModernBertModel": ["BaseModelOutput", "Consider", "Falling", "False", "LayerNorm", "LongTensor", "ModelConfig", "ModelEmbeddings", "ModelEncoderLayer", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Outputting", "Setting", "T", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_maybe_set_compile", "_pad_Model_output", "_prepare_4d_attention_mask", "_unpad_Model_input", "_update_attention_mask", "abs", "all_hidden_states", "all_self_attentions", "and", "arange", "attention", "attention_mask", "attentions", "attn_implementation", "auto_docstring", "back", "batch", "batch_size", "bias", "bool", "class", "config", "cu_seqlens", "def", "device", "distance", "dtype", "eager", "elif", "else", "embeddings", "encoder_layer", "eps", "exactly", "f", "final_norm", "finfo", "flash_attention_2", "for", "forward", "get_input_embeddings", "global_attention_mask", "gradient_checkpointing", "hidden_size", "hidden_states", "hs", "if", "implementation", "in", "indices", "input_ids", "inputs", "inputs_embeds", "int", "is", "last_hidden_state", "layer_id", "layer_outputs", "layers", "len", "local_attention", "logger", "logical_not", "masked_fill", "max_seqlen", "min", "must", "nn", "no_grad", "norm_bias", "norm_eps", "not", "num_hidden_layers", "of", "one", "ones", "only", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "raise", "range", "repad", "return", "return_dict", "rows", "sdpa", "self", "seq_len", "seqlen", "set_input_embeddings", "setting", "shape", "sliding_window_mask", "specify", "super", "supported", "the", "to", "tok_embeddings", "torch", "tuple", "unsqueeze", "use_return_dict", "v", "value", "warn_if_padding_and_no_attention_mask", "warning_once", "window_mask", "with"], "modernbert/modeling_modernbert.py:ModernBertPredictionHead": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelPredictionHead", "Module", "Tensor", "__init__", "act", "bias", "class", "classifier_activation", "classifier_bias", "config", "def", "dense", "eps", "forward", "hidden_size", "hidden_states", "nn", "norm", "norm_bias", "norm_eps", "return", "self", "super", "torch"], "modernbert/modeling_modernbert.py:ModernBertForMaskedLM": ["Linear", "LongTensor", "MaskedLMOutput", "ModelConfig", "ModelForMaskedLM", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "None", "Optional", "Tensor", "True", "Union", "__init__", "_attn_implementation", "_maybe_set_compile", "_pad_Model_output", "_tied_weights_keys", "_unpad_Model_input", "and", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "bias", "bool", "class", "compile", "compiled_head", "config", "cu_seqlens", "decoder", "decoder_bias", "def", "device", "dtype", "dynamic", "else", "flash_attention_2", "forward", "get_output_embeddings", "head", "hidden_size", "hidden_states", "if", "indices", "input_ids", "inputs", "inputs_embeds", "int", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "mask_tokens", "max_seqlen", "model", "new_embeddings", "nn", "no_grad", "not", "nullcontext", "ones", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "reference_compile", "repad_logits_with_grad", "return", "return_dict", "self", "seq_len", "seqlen", "set_output_embeddings", "shape", "sliding_window_mask", "sparse_pred_ignore_index", "sparse_prediction", "super", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight", "with"], "modernbert/modeling_modernbert.py:ModernBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "LongTensor", "MSELoss", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "None", "Optional", "SequenceClassifierOutput", "Tensor", "True", "Union", "__init__", "_maybe_set_compile", "and", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "classifier_dropout", "classifier_pooling", "cls", "config", "cu_seqlens", "def", "device", "dim", "drop", "dtype", "elif", "else", "forward", "head", "hidden_size", "hidden_states", "if", "indices", "input_ids", "inputs_embeds", "int", "is", "keepdim", "kwargs", "labels", "last_hidden_state", "logits", "long", "loss", "loss_fct", "max_seqlen", "mean", "model", "multi_label_classification", "nn", "not", "num_labels", "ones", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "seq_len", "shape", "single_label_classification", "sliding_window_mask", "squeeze", "sum", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "view", "warn_if_padding_and_no_attention_mask"], "modernbert/modeling_modernbert.py:ModernBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "LongTensor", "ModelConfig", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "_maybe_set_compile", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "classifier_dropout", "config", "cu_seqlens", "def", "drop", "else", "forward", "head", "hidden_size", "hidden_states", "if", "indices", "input_ids", "inputs_embeds", "int", "is", "labels", "last_hidden_state", "logits", "loss", "loss_fct", "max_seqlen", "model", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "seq_len", "sliding_window_mask", "super", "torch", "tuple", "use_return_dict", "view"], "modernbert/modeling_modernbert.py:ModernBertForQuestionAnswering": ["Dropout", "Linear", "ModelConfig", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "_maybe_set_compile", "and", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "classifier_dropout", "config", "contiguous", "cu_seqlens", "def", "dim", "drop", "else", "end_logits", "end_positions", "forward", "head", "hidden_size", "hidden_states", "if", "indices", "input_ids", "int", "is", "kwargs", "last_hidden_state", "logits", "loss", "loss_function", "max_seqlen", "model", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "seq_len", "sliding_window_mask", "split", "squeeze", "start_logits", "start_positions", "super", "torch", "tuple", "use_return_dict"], "modernbert/modeling_modernbert.py:ModernBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "LongTensor", "ModelConfig", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "True", "Union", "__init__", "_maybe_set_compile", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "classifier_dropout", "classifier_pooling", "cls", "config", "cu_seqlens", "def", "dim", "drop", "elif", "else", "forward", "head", "hidden_size", "hidden_states", "if", "indices", "input_ids", "inputs_embeds", "int", "is", "keepdim", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_fct", "max_seqlen", "mean", "model", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "seq_len", "shape", "size", "sliding_window_mask", "sum", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "view"], "ministral/modeling_ministral.py:MinistralMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "ministral/modeling_ministral.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "ministral/modeling_ministral.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "ministral/modeling_ministral.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "ministral/modeling_ministral.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "ministral/modeling_ministral.py:MinistralAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "ministral/modeling_ministral.py:MinistralRMSNorm": ["ModelRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "ministral/modeling_ministral.py:MinistralDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "ministral/modeling_ministral.py:MinistralPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "ministral/modeling_ministral.py:MinistralRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "ministral/modeling_ministral.py:MinistralModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "ministral/modeling_ministral.py:MinistralForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "ministral/modeling_ministral.py:MinistralForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "ministral/modeling_ministral.py:MinistralForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "ministral/modeling_ministral.py:MinistralForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "bark/modeling_bark.py:BarkSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "ValueError", "__init__", "_attn", "_merge_heads", "_split_heads", "and", "att_proj", "attention_mask", "attn_dropout", "attn_head_size", "attn_output", "attn_weights", "be", "bias", "block_size", "bool", "by", "cache_position", "class", "config", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "embed_dim", "f", "finfo", "forward", "functional", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "is", "is_causal", "key", "key_length", "layer_idx", "masked_fill", "math", "matmul", "min", "must", "new_shape", "nn", "not", "num_heads", "ones", "out_proj", "output_attentions", "past_key_values", "permute", "query", "query_length", "raise", "register_buffer", "resid_dropout", "return", "self", "size", "softmax", "split", "sqrt", "super", "tensor", "to", "torch", "transpose", "tril", "update", "use_cache", "value", "view"], "bark/modeling_bark.py:BarkSelfFlashAttention2": ["False", "ModelSelfAttention", "ModelSelfFlashAttention2", "None", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_merge_heads", "_split_heads", "args", "att_proj", "attention_mask", "attn_head_size", "attn_output", "batch_size", "cache_position", "class", "def", "dim", "dropout", "else", "embed_dim", "flash_attn_supports_top_left_mask", "forward", "head_dim", "head_mask", "hidden_states", "if", "is", "is_causal", "key", "kwargs", "layer_idx", "new_shape", "not", "num_heads", "out_proj", "output_attentions", "past_key_values", "query", "query_len", "resid_dropout", "return", "self", "size", "split", "super", "tensor", "training", "update", "use_cache", "use_top_left_mask", "value", "view"], "bark/modeling_bark.py:BarkMLP": ["Dropout", "GELU", "Linear", "ModelMLP", "Module", "__init__", "bias", "class", "config", "def", "dropout", "forward", "gelu", "hidden_size", "hidden_states", "in_proj", "nn", "out_proj", "return", "self", "super"], "bark/modeling_bark.py:BarkBlock": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelBlock", "ModelMLP", "Model_ATTENTION_CLASSES", "None", "__init__", "_attn_implementation", "attention_mask", "attn", "attn_output", "attn_outputs", "bias", "cache_position", "class", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "intermediary_hidden_states", "is_causal", "layer_idx", "layernorm_1", "layernorm_2", "mlp", "nn", "output_attentions", "outputs", "past_key_values", "return", "self", "super", "use_cache"], "bark/modeling_bark.py:BarkPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "__init__", "_hf_hook", "_init_weights", "_supports_flash_attn", "and", "bias", "class", "config", "data", "def", "device", "elif", "execution_device", "fill_", "for", "get_parameter_device", "hasattr", "if", "in", "initializer_range", "inputs", "is", "isinstance", "kwargs", "mean", "module", "modules", "nn", "normal_", "not", "padding_idx", "property", "return", "self", "std", "super", "supports_gradient_checkpointing", "torch", "weight", "zero_"], "bark/modeling_bark.py:BarkCausalModel": ["Cache", "CausalLMOutputWithPast", "Dropout", "DynamicCache", "Embedding", "False", "GenerationMixin", "LayerNorm", "Linear", "LongTensor", "Model", "ModelBlock", "ModelCausalModel", "ModelPreTrainedModel", "ModelSubModelConfig", "ModuleList", "None", "NotImplementedError", "Optional", "Passing", "Setting", "Tensor", "Training", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "a", "all_hidden_states", "all_self_attentions", "an", "and", "arange", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bias", "block", "block_size", "bool", "both", "cache_position", "cannot", "checkpointing", "class", "config", "def", "defined", "deprecated", "device", "do", "drop", "dropout", "dtype", "e", "either", "elif", "else", "ensure", "enumerate", "flash_attention_2", "for", "forward", "from_legacy_cache", "g", "get_head_mask", "get_input_embeddings", "get_output_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "has", "have", "head_mask", "hidden_size", "hidden_states", "i", "if", "implemented", "in", "incompatible", "input_embeds", "input_embeds_layer", "input_ids", "input_shape", "input_vocab_size", "inputs_embeds", "instance", "instead", "is", "is_causal", "isinstance", "kwargs", "labels", "layer_idx", "layernorm_final", "layers", "lm_head", "logger", "logits", "long", "loss", "model", "model_inputs", "new_embeddings", "nn", "not", "num_layers", "of", "or", "output_attentions", "output_hidden_states", "output_shape", "output_vocab_size", "outputs", "pass", "past_key_values", "past_length", "pop", "position_embeds", "position_embeds_layer", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "raise", "range", "removed", "return", "return_dict", "same", "self", "seq_length", "set_input_embeddings", "shape", "should", "size", "specify", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "v4", "view", "warning_once", "will", "with", "yet", "you"], "bark/modeling_bark.py:BarkSemanticModel": ["LongTensor", "ModelCausalModel", "ModelEosPrioritizerLogitsProcessor", "ModelSemanticConfig", "ModelSemanticGenerationConfig", "ModelSemanticModel", "None", "Optional", "SuppressTokensLogitsProcessor", "Tensor", "ValueError", "attention_mask", "base_model_prefix", "batch_size", "be", "bool", "cat", "class", "config", "constant", "def", "device", "dict", "dim", "dtype", "early_stopping_logits_processor", "else", "eos_token_id", "extend", "functional", "generate", "generation_config", "get", "has", "history_prompt", "if", "infer_array", "input_embeds", "input_embeds_layer", "input_ids", "int", "is", "kwargs", "len", "list", "logits_processor", "masked_fill", "max_input_semantic_length", "min_eos_p", "mode", "nn", "not", "ones", "output_vocab_size", "pad", "provided", "raise", "range", "repeat_interleave", "return", "self", "semantic", "semantic_generation_config", "semantic_history", "semantic_infer_token", "semantic_output", "semantic_pad_token", "semantic_prompt", "semantic_vocab_size", "shape", "str", "super", "suppress_tokens_logits_processor", "tensor", "text_encoding_offset", "text_pad_token", "to", "tokens_to_suppress", "torch", "value"], "bark/modeling_bark.py:BarkCoarseModel": ["AlternatingCodebooksLogitsProcessor", "F", "LongTensor", "ModelCausalModel", "ModelCoarseConfig", "ModelCoarseGenerationConfig", "ModelCoarseModel", "ModelSemanticGenerationConfig", "None", "Optional", "Tensor", "Union", "ValueError", "_", "alternatingLogitsProcessor", "base_model_prefix", "base_semantic_idx", "batch_size", "be", "bool", "ceil", "class", "clone", "coarse_acoustics", "coarse_generation_config", "coarse_infer_token", "coarse_output", "coarse_prompt", "coarse_rate_hz", "coarse_semantic_pad_token", "codebook_size", "config", "constant", "def", "del", "device", "dict", "dim", "dtype", "else", "floor", "for", "generate", "generation_config", "has", "history_prompt", "hstack", "if", "in", "input_coarse", "input_coarse_len", "int", "is", "item", "kwargs", "len_coarse_history", "logits_processor", "masked_fill_", "max", "max_coarse_history", "max_coarse_input_length", "max_generated_len", "max_new_tokens", "max_semantic_history", "min", "n", "n_coarse_codebooks", "n_coarse_hist_provided", "n_semantic_hist_provided", "n_window_steps", "not", "np", "output_coarse", "output_lengths", "pad", "preprocess_histories", "provided", "raise", "range", "repeat_interleave", "reshape", "return", "return_output_lengths", "round", "self", "semantic_generation_config", "semantic_idx", "semantic_output", "semantic_pad_token", "semantic_prompt", "semantic_rate_hz", "semantic_to_coarse_ratio", "semantic_vocab_size", "shape", "sliding_window_len", "str", "sum", "super", "tensor", "to", "torch", "total_generated_len", "transpose", "tuple", "x_coarse", "x_coarse_history", "x_semantic_history"], "bark/modeling_bark.py:BarkFineModel": ["Cannot", "Dropout", "Embedding", "F", "False", "LayerNorm", "Linear", "LongTensor", "MaskedLMOutput", "ModelBlock", "ModelCoarseGenerationConfig", "ModelFineConfig", "ModelFineGenerationConfig", "ModelFineModel", "ModelPreTrainedModel", "ModelSemanticGenerationConfig", "ModuleList", "None", "NotImplementedError", "Optional", "T", "Tensor", "Training", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_get_resized_embeddings", "_get_resized_lm_head", "_prepare_4d_attention_mask", "_resize_token_embeddings", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "all_hidden_states", "all_self_attentions", "and", "append", "arange", "argmax", "at", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "be", "bias", "block", "block_size", "bool", "both", "by", "cannot", "cat", "ceil", "class", "coarse", "coarse_generation_config", "coarse_output", "codebook", "codebook_idx", "codebook_preds", "codebook_size", "config", "constant", "def", "defined", "del", "device", "dict", "dim", "drop", "dropout", "dtype", "either", "else", "enumerate", "f", "fine_acoustics", "fine_generation_config", "fine_input", "fine_prompt", "flash_attention_2", "for", "forward", "generate", "get", "get_head_mask", "get_input_embeddings", "get_output_embeddings", "getattr", "gradient_checkpointing", "has", "hasattr", "have", "head_mask", "hidden_size", "hidden_states", "history_prompt", "i", "if", "implemented", "in", "input", "input_buffer", "input_embeddings", "input_embeds", "input_embeds_layer", "input_embeds_layers", "input_ids", "input_shape", "input_vocab_size", "int", "int32", "is", "is_causal", "kwargs", "labels", "layer_idx", "layernorm_final", "layers", "lm_heads", "logits", "long", "loss", "main_input_name", "max", "max_fine_history_length", "max_fine_input_length", "mean_resizing", "min", "mode", "model", "model_embeds", "module", "modules", "multinomial", "n_coarse", "n_coarse_codebooks", "n_codes_given", "n_codes_total", "n_fine_codebooks", "n_history", "n_inner", "n_loops", "n_outer", "n_remove_from_end", "new_embeddings", "new_embeddings_list", "new_lm_head_list", "new_num_tokens", "new_output_embeddings", "nn", "no_grad", "not", "np", "num_layers", "num_samples", "old_embeddings", "old_embeddings_list", "old_lm_head", "old_lm_head_list", "or", "output", "output_attentions", "output_embeddings", "output_hidden_states", "output_shape", "output_vocab_size", "outputs", "pad", "pad_to_multiple_of", "position_embeds", "position_embeds_layer", "position_ids", "post_init", "predict", "predicted", "probs", "provided", "r", "raise", "range", "rel_start_fill_idx", "relevant_logits", "remainder", "repeat_interleave", "reshape", "resize_token_embeddings", "return", "return_dict", "same", "self", "semantic_generation_config", "semantic_vocab_size", "seq_len", "seq_length", "set_input_embeddings", "set_output_embeddings", "shape", "should", "size", "softmax", "specify", "start_fill_idx", "start_idx", "str", "sum", "super", "temperature", "tgt_len", "the", "tie_weights", "tie_word_embeddings", "time", "to", "torch", "transpose", "tuple", "unsqueeze", "use_return_dict", "v", "value", "view", "vocab_size", "weight", "x_fine_history", "yet"], "bark/modeling_bark.py:BarkModel": ["AutoModel", "FutureWarning", "ImportError", "LongTensor", "ModelCoarseGenerationConfig", "ModelCoarseModel", "ModelConfig", "ModelFineGenerationConfig", "ModelFineModel", "ModelModel", "ModelPreTrainedModel", "ModelSemanticGenerationConfig", "ModelSemanticModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "_", "__init__", "_hf_hook", "_tie_weights", "accelerate", "accelerator", "accelerator_id", "and", "argument", "attention_mask", "audio", "audio_arr", "batch_first", "be", "bool", "can_generate", "class", "classmethod", "cls", "coarse_", "coarse_acoustics", "coarse_acoustics_config", "coarse_generation_config", "coarse_output", "codebook_size", "codec_config", "codec_decode", "codec_model", "codec_model_hook", "config", "cpu", "cpu_offload_with_hook", "cpu_offloaded_model", "cuda", "current_accelerator", "decode", "decoder", "def", "deprecated", "device", "device_type", "dict", "elif", "else", "emb", "empty_cache", "enable_cpu_offload", "enable_model_cpu_offload", "execution_device", "f", "fine_", "fine_acoustics", "fine_acoustics_config", "fine_acoustics_hook", "fine_generation_config", "fine_output", "for", "from_config", "generate", "generation_config", "get", "get_parameter_device", "getattr", "gpu_id", "hasattr", "history_prompt", "hook", "if", "in", "input_embeds_layer", "input_ids", "instead", "int", "is", "is_accelerate_available", "is_torch_accelerator_available", "items", "key", "kwargs", "kwargs_coarse", "kwargs_fine", "kwargs_semantic", "l", "len", "min_eos_p", "module", "modules", "n_coarse_codebooks", "nn", "no_grad", "not", "of", "offload", "out", "output", "output_lengths", "pad_sequence", "padding_value", "pop", "prev_module_hook", "property", "quantizer", "r", "raise", "removed", "requires", "return", "return_output_lengths", "rnn", "sample", "self", "semantic", "semantic_", "semantic_config", "semantic_generation_config", "semantic_output", "squeeze", "startswith", "str", "super", "tie_weights", "to", "torch", "torch_accelerator_module", "transpose", "type", "unsqueeze", "use", "utils", "value", "version", "warn", "warnings", "will", "zip"], "falcon/modeling_falcon.py:FalconLinear": ["Linear", "ModelLinear", "None", "T", "Tensor", "bias", "class", "def", "forward", "hidden_states", "if", "input", "is", "nn", "return", "self", "torch", "weight"], "falcon/modeling_falcon.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "falcon/modeling_falcon.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "falcon/modeling_falcon.py:FalconRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "falcon/modeling_falcon.py:build_alibi_tensor": ["Model_alibi_tensor", "None", "Tensor", "alibi", "arange", "arange_tensor", "attention_mask", "base", "batch_size", "bfloat16", "cat", "closest_power_of_2", "cumsum", "def", "device", "dim", "dtype", "extra_base", "extra_powers", "float32", "floor", "if", "int", "int32", "log2", "math", "min", "num_heads", "num_remaining_heads", "pow", "powers", "reshape", "return", "seq_length", "shape", "slopes", "tensor", "to", "torch"], "falcon/modeling_falcon.py:dropout_add": ["F", "Model", "Model_add", "Tensor", "bool", "def", "float", "out", "p", "prob", "residual", "return", "torch", "training", "x"], "falcon/modeling_falcon.py:FalconAttention": ["Cache", "Dropout", "F", "False", "Instantiating", "LongTensor", "ModelAttention", "ModelConfig", "ModelLinear", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "_merge_heads", "_split_heads", "a", "alibi", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_logits", "attention_mask", "attention_probs", "attention_probs_reshaped", "attention_scores", "attn_mask", "attn_output", "batch", "batch_size", "batch_size_and_num_heads", "be", "beta", "bfloat16", "bias", "bool", "broadcast_to", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "contiguous", "cos", "creating", "cuda", "def", "dense", "device", "dim", "divisible", "dropout_p", "dtype", "during", "elif", "else", "errors", "f", "flatten", "float16", "float32", "for", "forward", "functional", "fused_qkv", "got", "head_dim", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "input_dtype", "inv_norm_factor", "is", "is_causal", "key", "key_layer", "kv_length", "layer_idx", "layer_past", "lead", "logger", "make", "math", "matmul_result", "max_position_embeddings", "multi_query", "must", "new_decoder_architecture", "nn", "not", "num_attention_heads", "num_heads", "num_kv_heads", "or", "output_attentions", "p", "passing", "permute", "position_embeddings", "position_ids", "provide", "qkv", "qkv_out_dim", "query", "query_key_value", "query_layer", "query_length", "raise", "recommended", "reshape", "return", "rope_theta", "scaled_dot_product_attention", "sdpa", "self", "seq_len", "seq_length", "shape", "sin", "softmax", "split_size", "sqrt", "super", "sure", "the", "this", "three_times_hidden_size", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "used", "value", "value_layer", "view", "warning_once", "when", "will", "without", "x"], "falcon/modeling_falcon.py:FalconFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "The", "True", "ValueError", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "_split_heads", "alibi", "apply_rotary_pos_emb", "args", "attention_dropout", "attention_mask", "attn_dropout", "attn_output", "attn_weights", "back", "batch_size", "be", "bool", "cache_kwargs", "cache_position", "cast", "casted", "class", "config", "cos", "cpu", "def", "dense", "device", "device_type", "dropout", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "fused_qkv", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "have", "head_dim", "head_mask", "hidden", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "key_layer", "kwargs", "layer", "layer_idx", "layer_past", "layers", "logger", "might", "mps", "new_decoder_architecture", "norm", "not", "num_heads", "num_kv_heads", "or", "output_attentions", "position_embeddings", "position_ids", "query_key_value", "query_layer", "query_length", "raise", "related", "reshape", "return", "seems", "self", "shape", "silently", "sin", "states", "super", "supported", "target_dtype", "the", "this", "to", "torch", "training", "transpose", "tuple", "type", "upcasted", "update", "use_cache", "use_flash_attn", "use_top_left_mask", "value_layer", "warning_once", "weight", "when", "will", "you"], "falcon/modeling_falcon.py:FalconMLP": ["ModelConfig", "ModelLinear", "ModelMLP", "Module", "Tensor", "__init__", "act", "activation", "bias", "class", "config", "def", "dense_4h_to_h", "dense_h_to_4h", "ffn_hidden_size", "forward", "get_activation", "hidden_dropout", "hidden_size", "nn", "return", "self", "super", "torch", "x"], "falcon/modeling_falcon.py:FalconDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "Union", "__init__", "_attn_implementation", "alibi", "and", "attention_dropout", "attention_layernorm_out", "attention_mask", "attention_output", "attn_weights", "bool", "cache_position", "class", "config", "def", "dropout_add", "else", "eps", "forward", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_layernorm", "is", "kwargs", "layer_idx", "layer_norm_epsilon", "layer_past", "ln_attn", "ln_mlp", "mlp", "mlp_layernorm_out", "mlp_output", "new_decoder_architecture", "not", "num_attention_heads", "num_heads", "num_ln_in_parallel_attn", "or", "output", "output_attentions", "parallel_attn", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attention", "super", "torch", "training", "tuple", "use_cache"], "falcon/modeling_falcon.py:FalconPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelLinear", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "__init__", "_attn_implementation", "_can_compile_fullgraph", "_check_and_enable_sdpa", "_init_weights", "_is_bettertransformer", "_no_split_modules", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "bool", "class", "classmethod", "cls", "config", "data", "def", "elif", "fill_", "getattr", "hard_check_only", "if", "initializer_range", "inputs", "is", "isinstance", "kwargs", "mean", "module", "nn", "normal_", "not", "padding_idx", "return", "sdpa", "self", "std", "super", "supports_gradient_checkpointing", "transformer", "use_bettertransformer", "weight", "zero_"], "falcon/modeling_falcon.py:FalconModel": ["AttentionMaskConverter", "BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "False", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "alibi", "all_hidden_states", "all_self_attentions", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "block", "bool", "build_alibi_tensor", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "dtype", "else", "embed_dim", "enumerate", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "for", "forward", "full", "get_head_mask", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "h", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layer_past", "ln_f", "logger", "long", "mask", "mask_length", "masked_fill", "math", "min", "min_dtype", "must", "new_embeddings", "nn", "not", "npu", "num_attention_heads", "num_heads", "num_hidden_layers", "of", "one", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "reshape", "return", "return_dict", "rotary_emb", "sdpa", "self", "seq_length", "sequence_length", "set_input_embeddings", "shape", "specify", "sqrt", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_alibi", "use_cache", "use_return_dict", "using_static_cache", "v", "vocab_size", "warning_once", "with", "word_embeddings", "xpu"], "falcon/modeling_falcon.py:FalconForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "lm_logits", "logits", "logits_to_keep", "loss", "loss_function", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_output_embeddings", "slice", "slice_indices", "super", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "falcon/modeling_falcon.py:FalconForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "Linear", "LongTensor", "MSELoss", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "head_mask", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "tokens", "torch", "transformer", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "warning_once", "will", "with"], "falcon/modeling_falcon.py:FalconForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "Linear", "LongTensor", "ModelConfig", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "elif", "else", "forward", "getattr", "head_mask", "hidden_dropout", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "seq_length", "shape", "super", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view"], "falcon/modeling_falcon.py:FalconForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "lfm2/modeling_lfm2.py:Lfm2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "lfm2/modeling_lfm2.py:Lfm2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "lfm2/modeling_lfm2.py:Lfm2MLP": ["F", "False", "Linear", "ModelConfig", "ModelMLP", "Module", "None", "__init__", "bias", "block_auto_adjust_ff_dim", "block_ffn_dim_multiplier", "block_multiple_of", "class", "config", "def", "forward", "hidden_size", "if", "int", "intermediate_size", "is", "nn", "not", "return", "self", "silu", "super", "w1", "w2", "w3", "x"], "lfm2/modeling_lfm2.py:Lfm2HybridConvCache": ["Any", "False", "LongTensor", "ModelConfig", "ModelHybridConvCache", "None", "Optional", "Tensor", "Union", "_", "__getitem__", "__init__", "__len__", "_dtype", "_dynamo", "abs", "append", "beam_idx", "cache_kwargs", "cache_position", "cat", "class", "config", "conv_L_cache", "conv_cache", "conv_state", "crop", "def", "device", "dict", "dim", "dtype", "elif", "else", "first_attention_layer", "float32", "for", "full_attention", "full_mask_kv_offset", "get_mask_sizes", "get_seq_length", "hidden_size", "idx", "if", "in", "index", "index_select", "int", "is", "is_compileable", "key_cache", "key_states", "kv_length", "layer_idx", "layer_types", "len", "list", "mark_static_address", "max_batch_size", "max_length", "not", "num_hidden_layers", "numel", "or", "past_seen_tokens", "query_length", "range", "reorder_cache", "reset", "return", "self", "shape", "str", "tensor", "to", "torch", "tuple", "update", "value_cache", "value_states", "zero_", "zeros"], "lfm2/modeling_lfm2.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "lfm2/modeling_lfm2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "lfm2/modeling_lfm2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "lfm2/modeling_lfm2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "lfm2/modeling_lfm2.py:Lfm2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelHybridConvCache", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_layernorm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "norm_eps", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "out_proj", "output", "past_key_value", "past_key_values", "position_embeddings", "q_layernorm", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "lfm2/modeling_lfm2.py:apply_mask_to_padding_states": ["Model_mask_to_padding_states", "None", "and", "attention_mask", "def", "dtype", "hidden_states", "if", "is", "not", "return", "shape", "to"], "lfm2/modeling_lfm2.py:Lfm2ShortConv": ["B", "BCx", "Bx", "C", "Conv1d", "L_cache", "Linear", "LongTensor", "ModelConfig", "ModelHybridConvCache", "ModelShortConv", "Module", "None", "Optional", "Tensor", "__init__", "_dynamo", "activation", "and", "apply_mask_to_padding_states", "attention_mask", "bias", "cache_position", "causal_conv1d_fn", "causal_conv1d_update", "chunk", "clamp", "class", "config", "contiguous", "conv", "conv_L_cache", "conv_bias", "conv_cache", "conv_out", "conv_state", "conv_weights", "copy_", "cuda", "cuda_kernels_forward", "def", "deprecate_kwarg", "device", "dim", "dims", "dtype", "else", "forward", "functional", "groups", "hidden_size", "hidden_states", "if", "in", "in_channels", "in_proj", "int", "is", "is_compiling", "is_fast_path_available", "kernel_size", "layer_idx", "new_name", "nn", "not", "out_channels", "out_proj", "pad", "padding", "past_key_value", "past_key_values", "return", "roll", "self", "seqlen", "shape", "shifts", "size", "slow_forward", "squeeze", "sum", "super", "to", "torch", "transpose", "type", "unsqueeze", "version", "view", "weight", "x", "y"], "lfm2/modeling_lfm2.py:Lfm2DecoderLayer": ["GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelHybridConvCache", "ModelMLP", "ModelRMSNorm", "ModelShortConv", "None", "Optional", "Tensor", "_", "__init__", "attention_mask", "cache_position", "class", "config", "conv", "def", "deprecate_kwarg", "else", "eps", "feed_forward", "ffn_norm", "forward", "full_attention", "hidden_size", "hidden_states", "if", "int", "is_attention_layer", "kwargs", "layer_idx", "layer_types", "new_name", "norm_eps", "operator_norm", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "version"], "lfm2/modeling_lfm2.py:Lfm2PreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "lfm2/modeling_lfm2.py:Lfm2Model": ["BaseModelOutputWithPast", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelHybridConvCache", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "batch_size", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "dtype", "else", "embed_tokens", "embedding_norm", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "max_batch_size", "must", "nn", "norm_eps", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "pos_emb", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "lfm2/modeling_lfm2.py:Lfm2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "opt/modeling_opt.py:OPTLearnedPositionalEmbedding": ["Embedding", "LongTensor", "Model", "Modelional", "None", "__init__", "attention_mask", "class", "cumsum", "def", "dim", "embedding_dim", "forward", "if", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "super", "torch"], "opt/modeling_opt.py:eager_attention_forward": ["Model_attention_forward", "Modelional", "Module", "None", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "opt/modeling_opt.py:OPTAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Instantiating", "Linear", "Model", "ModelConfig", "Modelional", "Module", "None", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "enable_bias", "errors", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_head_mask", "layer_idx", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "size", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "opt/modeling_opt.py:OPTDecoderLayer": ["ACT2FN", "Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "LongTensor", "Model", "ModelAttention", "ModelConfig", "Modelional", "None", "Tensor", "Unpack", "__init__", "activation_fn", "activation_function", "attention_mask", "bias", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "do_layer_norm_before", "dropout", "elementwise_affine", "embed_dim", "enable_bias", "fc1", "fc2", "ffn_dim", "final_layer_norm", "forward", "functional", "hidden_size", "hidden_states", "hidden_states_shape", "if", "int", "kwargs", "layer_head_mask", "layer_idx", "layer_norm_elementwise_affine", "new_name", "nn", "not", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "position_ids", "reshape", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "shape", "size", "super", "torch", "training", "tuple", "use_cache", "version", "view"], "opt/modeling_opt.py:OPTPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoderLayer", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "init_std", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "opt/modeling_opt.py:OPTDecoder": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "Modelional", "ModuleList", "None", "Setting", "Tensor", "The", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_remove_final_layer_norm", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bias", "bool", "but", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "continue", "cuda", "cumsum", "decoder_layer", "def", "device", "diagonal", "dim", "do_layer_norm_before", "dropout", "dropout_probability", "dtype", "elementwise_affine", "else", "embed_positions", "embed_tokens", "enumerate", "exactly", "expand", "f", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "it", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm_elementwise_affine", "layer_outputs", "layerdrop", "layers", "len", "logger", "long", "make_flex_block_causal_mask", "mask_length", "mask_name", "masked_fill", "max_position_embeddings", "max_target_positions", "min", "min_dtype", "must", "nn", "not", "npu", "num_hidden_layers", "of", "one", "ones", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "pos_embeds", "position_ids", "post_init", "project_in", "project_out", "r", "raise", "rand", "range", "reshape", "return", "return_dict", "sdpa", "self", "seq_length", "sequence_length", "shape", "should", "size", "specified", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "using_compilable_cache", "view", "vocab_size", "warning_once", "with", "word_embed_proj_dim", "xpu", "zip"], "opt/modeling_opt.py:OPTModel": ["BaseModelOutputWithPast", "Cache", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoder", "ModelPreTrainedModel", "Modelional", "None", "Tensor", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "decoder_outputs", "def", "else", "embed_tokens", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "not", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "return", "return_dict", "self", "set_input_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "value"], "opt/modeling_opt.py:OPTForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "Modelional", "None", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "contiguous", "decoder", "def", "device", "else", "embed_tokens", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "logits", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight", "word_embed_proj_dim"], "opt/modeling_opt.py:OPTForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "Modelional", "None", "Results", "SequenceClassifierOutputWithPast", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "decoder", "def", "defined", "detect", "device", "dtype", "elif", "else", "embed_tokens", "f", "forward", "get_input_embeddings", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "model", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "sequence_length", "set_input_embeddings", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "tokens", "torch", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "value", "view", "warning_once", "will", "with", "word_embed_proj_dim"], "opt/modeling_opt.py:OPTForQuestionAnswering": ["Cache", "CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "Modelional", "None", "QuestionAnsweringModelOutput", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "decoder", "def", "device", "dim", "else", "embed_tokens", "end_logits", "end_loss", "end_positions", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "set_input_embeddings", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "to", "torch", "total_loss", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "value", "word_embed_proj_dim"], "m2m_100/modeling_m2m_100.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "m2m_100/modeling_m2m_100.py:M2M100ScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "m2m_100/modeling_m2m_100.py:M2M100SinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "contiguous", "cos", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "detach", "device", "dim", "dtype", "else", "emb", "emb_weights", "embedding_dim", "exp", "expand", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sequence_length", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "m2m_100/modeling_m2m_100.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "m2m_100/modeling_m2m_100.py:M2M100Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "m2m_100/modeling_m2m_100.py:M2M100EncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "m2m_100/modeling_m2m_100.py:M2M100DecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "m2m_100/modeling_m2m_100.py:M2M100PreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "def", "device", "diagonal", "dim", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "nn", "normal_", "not", "npu", "ones", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "m2m_100/modeling_m2m_100.py:M2M100Encoder": ["BaseModelOutput", "Embedding", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "ValueError", "You", "_", "__init__", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "skip_the_layer", "specified", "specify", "sqrt", "super", "synced_gpus", "the", "time", "to", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "weight"], "m2m_100/modeling_m2m_100.py:M2M100Decoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "should", "size", "skip_the_layer", "specified", "specify", "sqrt", "super", "synced_gpus", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with", "zip"], "m2m_100/modeling_m2m_100.py:M2M100Model": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "math", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "sqrt", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "m2m_100/modeling_m2m_100.py:M2M100ForConditionalGeneration": ["Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "masked_lm_loss", "model", "nn", "not", "num_embeddings", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "post_init", "r", "return", "return_dict", "self", "shared", "shift_tokens_right", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "extracted_states", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "decoder_classes", "decoder_coords", "encoder_class_logits", "encoder_coord_logits", "hidden_states", "init_reference_points", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboObjectDetectionOutput": ["FloatTensor", "LongTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "class", "classes_structure", "decoder_attentions", "decoder_class_logits", "decoder_coord_logits", "decoder_hidden_states", "encoder_attentions", "encoder_class_logits", "encoder_coord_logits", "encoder_extracted_states", "encoder_hidden_states", "init_reference_points", "intermediate_reference_points", "loss", "r", "torch", "tuple"], "omdet_turbo/modeling_omdet_turbo.py:MultiScaleDeformableAttention": ["False", "ModelScaleDeformableAttention", "Module", "Tensor", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "class", "contiguous", "def", "dim", "enumerate", "flatten", "for", "forward", "functional", "grid_sample", "height", "hidden_dim", "im2col_step", "in", "int", "level_id", "level_start_index", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "self", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "value_spatial_shapes_list", "view", "width", "zeros"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboLRUCache": ["False", "ModelLRUCache", "None", "OrderedDict", "__init__", "bool", "cache", "capacity", "class", "current_load", "def", "get", "has", "if", "in", "int", "key", "last", "move_to_end", "not", "popitem", "put", "return", "self", "value"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboLanguageBackbone": ["AutoModel", "ModelConfig", "ModelLanguageBackbone", "Module", "None", "Parameter", "ValueError", "__init__", "arange", "argmax", "class", "config", "def", "dim", "elif", "else", "encode_type", "encoding", "f", "for", "forward", "from_config", "hidden_states", "if", "is", "item", "mask", "max", "max_len", "max_pooled_output", "model", "nn", "not", "pooled_output", "projected_output", "raise", "required", "return", "self", "shape", "sum", "super", "supported", "task", "text_config", "text_outputs", "text_projection", "text_projection_in_dim", "text_projection_out_dim", "torch", "transpose", "truncated_mask", "truncated_output", "zeros"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboVisionBackbone": ["LayerNorm", "ModelConfig", "ModelVisionBackbone", "Module", "ModuleList", "__init__", "apply_layernorm_after_vision_backbone", "class", "config", "contiguous", "def", "encoder_in_channels", "eps", "feature_maps", "for", "forward", "if", "in", "in_channel_dim", "layer_norm", "layer_norm_eps", "layer_norms", "load_backbone", "nn", "output", "outputs", "permute", "pixel_values", "return", "self", "super", "vision_backbone", "zip"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboMultiscaleDeformableAttention": ["CUDA", "F", "False", "Last", "Linear", "Make", "ModelConfig", "ModelMultiscaleDeformableAttention", "Module", "MultiScaleDeformableAttention", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "attn", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "config", "d", "d_model", "def", "dim", "dim_per_head", "dimension", "disable_custom_kernels", "divisible", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "for", "forward", "got", "head", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "make", "masked_fill", "more", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_coordinates", "num_feature_levels", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "total_elements", "value", "value_proj", "view", "warn", "warnings", "which", "with", "with_pos_embed"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboConvNormLayer": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "Identity", "ModelConvNormLayer", "Module", "None", "__init__", "activation", "batch_norm_eps", "bias", "class", "config", "conv", "def", "else", "forward", "hidden_state", "if", "in_channels", "is", "kernel_size", "nn", "norm", "out_channels", "padding", "return", "self", "stride", "super"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboRepVggBlock": ["ACT2CLS", "Identity", "ModelConfig", "ModelConvNormLayer", "ModelRepVggBlock", "Module", "None", "__init__", "activation", "class", "config", "conv1", "conv2", "csp_activation", "def", "else", "encoder_hidden_dim", "forward", "hidden_channels", "hidden_expansion", "if", "int", "is", "nn", "padding", "return", "self", "super", "x", "y"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboCSPRepLayer": ["Identity", "ModelCSPRepLayer", "ModelConfig", "ModelConvNormLayer", "ModelRepVggBlock", "Module", "Sequential", "_", "__init__", "activation", "bottlenecks", "class", "config", "conv1", "conv2", "conv3", "csp_activation", "def", "else", "encoder_hidden_dim", "for", "forward", "hidden_channels", "hidden_expansion", "hidden_state", "hidden_state_1", "hidden_state_2", "if", "in", "in_channels", "int", "nn", "num_blocks", "out_channels", "range", "return", "self", "super"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboMultiheadAttention": ["Dropout", "False", "FloatTensor", "Linear", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_scores", "batch_size", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "f", "forward", "functional", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "keys", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "out_proj", "output_attentions", "outputs", "permute", "queries", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "values", "view"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboEncoderLayer": ["ACT2FN", "Dropout", "False", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderLayer", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_fn", "any", "attention_mask", "attentions", "bool", "clamp", "clamp_value", "class", "config", "def", "dropout", "dtype", "else", "encoder_dim_feedforward", "encoder_dropout", "encoder_feedforward_activation", "encoder_feedforward_dropout", "encoder_hidden_dim", "eps", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "hidden_size", "hidden_states", "if", "is", "isinf", "isnan", "key", "keys", "layer_norm_eps", "max", "min", "nn", "num_attention_heads", "or", "output_attentions", "pos_embed", "position_embeddings", "queries", "query", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "staticmethod", "super", "tensor", "torch", "training", "values", "with_pos_embed"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboEncoder": ["False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Tensor", "Union", "_", "__init__", "attention", "attention_mask", "bool", "class", "config", "def", "else", "encoder_layers", "for", "forward", "hidden_states", "if", "in", "layer", "layers", "nn", "output_attentions", "pos_embed", "position_embeddings", "range", "return", "self", "src", "src_mask", "super", "torch", "tuple"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboHybridEncoder": ["BatchNorm2d", "Conv2d", "Embed", "F", "False", "ModelCSPRepLayer", "ModelConfig", "ModelConvNormLayer", "ModelEncoder", "ModelEncoderOutput", "ModelHybridEncoder", "Module", "ModuleList", "None", "Sequential", "ValueError", "_", "__init__", "activation", "all_attentions", "append", "arange", "attentions", "be", "bias", "build_2d_sincos_position_embedding", "by", "channel_projection_layers", "class", "concat", "config", "contiguous", "conv_norm_activation", "cos", "cpu", "def", "device", "dim", "dimension", "divisible", "downsample_convs", "downsample_feat", "dtype", "else", "embed_dim", "embedding", "encoder", "encoder_hidden_dim", "encoder_in_channels", "encoder_layer_index", "encoder_projection_indices", "encoder_states", "enumerate", "eval_size", "extracted_states", "feat_high", "feat_low", "feature", "feature_to_project_index", "flatten", "float32", "for", "forward", "fpn_blocks", "fpn_feature_maps", "fpn_states", "fps_map", "grid_h", "grid_w", "height", "hidden_states", "i", "idx", "if", "ij", "in", "in_channel", "in_channels", "indexing", "inputs_embeddings", "insert", "int", "interpolate", "is", "kernel_size", "last_hidden_state", "lateral_convs", "layer_outputs", "len", "meshgrid", "mode", "must", "nearest", "nn", "not", "omega", "or", "out_channels", "out_h", "out_w", "output_attentions", "output_hidden_states", "pan_blocks", "permute", "pos_dim", "pos_embed", "position", "positional_encoding_temperature", "projected_features", "r", "raise", "range", "reshape", "return", "return_dict", "scale_factor", "self", "shape", "sin", "src_flatten", "staticmethod", "stride", "super", "temperature", "to", "torch", "training", "upsample_feat", "use_return_dict", "width"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboMLPWithDropout": ["ACT2FN", "Dropout", "Linear", "ModelMLPWithDropout", "Module", "__init__", "activation", "class", "class_embed_dim", "config", "decoder_activation", "decoder_dropout", "def", "dropout", "forward", "linear1", "linear2", "nn", "return", "self", "super", "task_encoder_hidden_dim", "x"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboMLP": ["F", "Linear", "ModelMLP", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "hidden_dim", "hidden_layers_dims", "i", "if", "in", "in_dim", "input_dim", "layer", "layers", "layers_dims", "nn", "num_layers", "out_dim", "output_dim", "relu", "return", "self", "super", "x", "zip"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboResidualLayer": ["Dropout", "LayerNorm", "ModelResidualLayer", "Module", "__init__", "class", "class_embed_dim", "config", "decoder_dropout", "def", "dropout", "eps", "forward", "layer_norm_eps", "nn", "norm1", "return", "self", "super", "x", "y"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboTaskEncoder": ["ModelMLPWithDropout", "ModelResidualLayer", "ModelTaskEncoder", "Module", "__init__", "class", "config", "def", "forward", "mlp", "mlp_out", "nn", "res1", "return", "self", "super", "x"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboDeformableTransformerDecoderLayer": ["ACT2FN", "Dropout", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelDeformableTransformerDecoderLayer", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "None", "__init__", "act", "attention_mask", "cat", "class", "config", "context", "cross_attention", "cross_attn", "decoder_activation", "decoder_dim_feedforward", "decoder_dropout", "decoder_embeddings", "decoder_hidden_dim", "decoder_num_heads", "decoder_num_points", "def", "dim", "dropout", "dropout1", "dropout2", "dropout3", "dropout4", "else", "encoder_hidden_states", "eps", "forward", "hidden_size", "hidden_states", "if", "is", "key", "layer_norm_eps", "level_start_index", "linear1", "linear2", "n_points", "nn", "norm1", "norm2", "norm3", "not", "num_attention_heads", "num_heads", "origin_embedding_len", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "pos", "query", "query_position", "reference_points", "residual", "return", "self", "self_attention", "self_attn", "shape", "spatial_shapes", "spatial_shapes_list", "staticmethod", "super", "task_features", "tensor", "torch", "transpose", "unsqueeze", "vision_features", "vision_shapes", "vision_shapes_list", "with_pos_embed"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboPreTrainedModel": ["BatchNorm2d", "Conv2d", "F", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoder", "ModelEncoderLayer", "ModelLanguageBackbone", "ModelPreTrainedModel", "None", "PreTrainedModel", "_", "_get_cache_key_at_index", "_init_weights", "_set_gradient_checkpointing", "and", "append", "attention_mask", "base_model_prefix", "batched_classes_embeddings", "bias", "bound", "cache_key", "cat", "channel_projection_layers", "class", "class_embeddings", "class_embeddings_regrouped", "classes_attention_mask", "classes_input_ids", "classes_structure", "config", "constant_", "cur_mask", "data", "decoder_bbox_head", "def", "device", "dim", "elif", "else", "emb", "embeddings", "encode_type", "encoder_bbox_head", "encoder_vision_features", "enumerate", "fc1", "fc2", "fill_", "for", "get", "get_cached_class_embeddings", "get_cached_task_embeddings", "get_language_embedding", "gradient_checkpointing", "has", "hasattr", "idx", "idx_to_put", "if", "in", "index", "init", "init_std", "input_ids", "input_mask", "is", "isinstance", "language_backbone", "language_cache_class", "language_cache_prompt", "layer", "layers", "learn_initial_query", "linear_init_", "main_input_name", "mask", "masks", "math", "max", "max_class_size", "max_len", "mean", "mlp", "model", "module", "module_to_init", "nn", "normal_", "not", "not_cached_classes", "not_cached_classes_ids", "not_cached_index", "not_cached_index_ids", "not_cached_mask", "not_cached_tasks", "pad", "pad_size", "pixel_values", "put", "query_position_head", "range", "return", "self", "shape", "size", "sqrt", "stack", "start", "staticmethod", "std", "task", "task_embeddings", "task_feature", "task_mask", "tasks_attention_mask", "tasks_input_ids", "text_projection", "text_projection_in_dim", "tgt_embed", "to", "tolist", "torch", "total_class_embs", "total_embeddings", "total_task_features", "total_task_masks", "tuple", "uniform_", "unsqueeze", "value", "weight", "xavier_uniform_", "zero_"], "omdet_turbo/modeling_omdet_turbo.py:_cosine_similarity_scaled": ["True", "_cosine_similarity_scaled", "a", "b", "bmm", "clamp_min", "def", "dim", "exp", "keepdim", "logit_scale", "logits_per_image", "norm", "return", "torch"], "omdet_turbo/modeling_omdet_turbo.py:get_class_similarity": ["Exception", "Model_class_similarity", "Unknown", "_cosine_similarity_scaled", "bmm", "class_distance_type", "class_logits", "class_proj", "cls_feature", "cosine", "def", "dot", "elif", "else", "f", "if", "log", "logit_scale", "raise", "return", "tensor", "torch"], "omdet_turbo/modeling_omdet_turbo.py:_inverse_sigmoid": ["_inverse_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboDecoder": ["BatchNorm2d", "Conv2d", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelDeformableTransformerDecoderLayer", "ModelMLP", "ModelPreTrainedModel", "ModelTaskEncoder", "ModuleList", "None", "Sequential", "True", "ValueError", "_", "__init__", "_get_decoder_input", "_get_encoder_input", "_inverse_sigmoid", "_prepare_4d_attention_mask", "all", "all_attns", "all_cross_attns", "all_hidden_states", "all_self_attns", "anchors", "append", "arange", "attention_mask", "attentions", "attn_mask_len", "batch_ind", "batch_size", "be", "bias", "bool", "break", "cat", "channel_projection_layers", "class", "class_distance_type", "class_embed_dim", "class_features", "class_proj", "cls_feature", "concat", "config", "cpu", "cross_attention", "cumsum", "decoder_bbox_head", "decoder_bboxes", "decoder_class_head", "decoder_classes", "decoder_coords", "decoder_embeddings", "decoder_hidden_dim", "decoder_num_layers", "def", "denoise_bboxes", "denoise_embeddings", "detach", "device", "dim", "dtype", "elif", "else", "embeddings", "encoder_bbox_head", "encoder_bboxes", "encoder_class_head", "encoder_class_logits", "encoder_class_similarity", "encoder_coord_logits", "encoder_outputs_bboxes", "encoder_vision_features", "end", "enumerate", "eps", "feat", "flatten", "float32", "for", "forward", "fusion_size", "generate_anchors", "get_class_similarity", "gradient_checkpointing", "grid_size", "grid_x", "grid_xy", "grid_y", "height", "hidden_dim", "hidden_states", "i", "if", "ij", "in", "indexing", "indices", "inf", "init_reference_points", "input_dim", "int64", "intermediate_reference_points", "is", "keepdim", "key_padding_mask", "last_hidden_state", "last_refined_bbox", "layer", "layer_norm_eps", "layers", "learn_initial_query", "len", "level", "level_start_index", "log", "lru_cache", "max", "maxsize", "meshgrid", "must", "new_vision_features", "new_vision_shapes", "new_vision_shapes_list", "new_zeros", "nn", "not", "num_layers", "num_queries", "ones_like", "original_class_projected", "output_attentions", "output_dim", "output_hidden_states", "permute", "post_init", "predicted_class_features", "prod", "provided", "query_position", "query_position_head", "raise", "range", "reference_points", "refined_bbox", "repeat", "reshape", "return", "return_dict", "self", "self_attention", "shape", "sigmoid", "spatial_shapes", "src_key_mask", "stack", "super", "task_encoder", "task_features", "task_mask", "task_project", "tensor", "tgt_embed", "to", "topk", "topk_ind", "torch", "training", "tuple", "unsqueeze", "use_return_dict", "valid_mask", "valid_wh", "values", "view", "vision_features", "vision_features_channels", "vision_shapes", "vision_shapes_list", "weight", "wh", "where", "width", "x", "zeros"], "omdet_turbo/modeling_omdet_turbo.py:OmDetTurboForObjectDetection": ["Embedding", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelForObjectDetection", "ModelHybridEncoder", "ModelLRUCache", "ModelLanguageBackbone", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModelVisionBackbone", "None", "NotImplementedError", "Optional", "Training", "True", "Union", "__init__", "attentions", "auto_docstring", "bool", "cache_size", "class", "class_features", "classes_attention_mask", "classes_input_ids", "classes_structure", "config", "decoder", "decoder_attentions", "decoder_class_logits", "decoder_classes", "decoder_coord_logits", "decoder_coords", "decoder_hidden_states", "decoder_outputs", "def", "else", "encoder", "encoder_attentions", "encoder_class_logits", "encoder_coord_logits", "encoder_extracted_states", "encoder_hidden_states", "encoder_outputs", "extracted_states", "for", "forward", "get_input_embeddings", "get_language_embedding", "hidden_states", "if", "image_features", "implemented", "in", "init_reference_points", "int", "intermediate_reference_points", "is", "labels", "language_backbone", "language_cache_class", "language_cache_prompt", "loss", "mean_resizing", "model", "model_embeds", "new_num_tokens", "nn", "not", "num_embeddings", "num_queries", "output", "output_attentions", "output_hidden_states", "pad_to_multiple_of", "pixel_values", "post_init", "r", "raise", "resize_token_embeddings", "return", "return_dict", "self", "set_input_embeddings", "super", "task_features", "task_mask", "tasks_attention_mask", "tasks_input_ids", "text_config", "torch", "tuple", "use_return_dict", "value", "vision_backbone", "vocab_size", "yet"], "blip/modeling_blip.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "blip/modeling_blip.py:blip_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "blip/modeling_blip.py:BlipForConditionalGenerationModelOutput": ["FloatTensor", "FutureWarning", "ModelForConditionalGenerationModelOutput", "ModelOutput", "None", "Optional", "Please", "Transformers", "and", "attentions", "attribute", "be", "class", "decoder_logits", "def", "deprecated", "final", "hidden_states", "image_embeds", "in", "instead", "is", "last_hidden_state", "logits", "loss", "of", "output", "property", "r", "removed", "retrieve", "return", "self", "the", "to", "torch", "tuple", "use", "version", "warn", "warnings", "will"], "blip/modeling_blip.py:BlipTextVisionModelOutput": ["FloatTensor", "ModelOutput", "ModelTextVisionModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "loss", "r", "torch", "tuple"], "blip/modeling_blip.py:BlipImageTextMatchingModelOutput": ["FloatTensor", "ModelImageTextMatchingModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "itm_score", "last_hidden_state", "loss", "question_embeds", "r", "torch", "tuple", "vision_pooler_output"], "blip/modeling_blip.py:BlipOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "blip/modeling_blip.py:BlipVisionEmbeddings": ["Conv2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "dtype", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embedding", "randn", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "view", "weight", "width"], "blip/modeling_blip.py:BlipTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "blip/modeling_blip.py:BlipAttention": ["Dropout", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "__init__", "_shape", "and", "attention_dropout", "attention_probs", "attention_scores", "be", "bsz", "by", "class", "config", "context_layer", "contiguous", "def", "dim", "divisible", "dropout", "embed_dim", "f", "forward", "functional", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "int", "is", "key_states", "kwargs", "matmul", "mixed_qkv", "must", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_heads", "output", "permute", "projection", "qkv", "query_states", "raise", "reshape", "return", "scale", "self", "seq_len", "size", "softmax", "super", "tensor", "tgt_len", "torch", "transpose", "tuple", "value_states", "view"], "blip/modeling_blip.py:BlipMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "blip/modeling_blip.py:BlipEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "blip/modeling_blip.py:BlipPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "and", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "factor", "fill_", "hasattr", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "past_key_values", "position_embedding", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "vision_config", "weight", "zero_"], "blip/modeling_blip.py:BlipEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "blip/modeling_blip.py:BlipVisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "attentions", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "get_input_embeddings", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "main_input_name", "nn", "pixel_values", "pooled_output", "pooler_output", "post_init", "post_layernorm", "raise", "return", "self", "specify", "super", "to", "torch", "tuple"], "blip/modeling_blip.py:BlipModel": ["False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelForImageTextRetrieval", "ModelForQuestionAnswering", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelVisionConfig", "ModelVisionModel", "Model_loss", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "True", "TypeError", "Union", "Unpack", "__init__", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "can_return_tuple", "class", "config", "def", "depending", "deprecated", "device", "dim", "dtype", "encoder_attention_mask", "encoder_hidden_states", "exp", "expected", "f", "forward", "future", "get_image_features", "get_input_embeddings", "get_multimodal_features", "get_text_features", "going", "hidden_size", "if", "image_atts", "image_embeds", "image_features", "in", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "kwargs", "logger", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "long", "loss", "matmul", "multimodal_features", "nn", "norm", "not", "of", "on", "ones", "or", "p", "pixel_values", "please", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "release", "return", "return_loss", "self", "set_input_embeddings", "size", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "use", "usecase", "value", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection", "warning", "your"], "blip/modeling_blip.py:BlipForConditionalGeneration": ["False", "FloatTensor", "GenerationMixin", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelForConditionalGenerationModelOutput", "ModelPreTrainedModel", "ModelTextLMHeadModel", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "bos_token_id", "can_return_tuple", "class", "cls", "config", "decoder", "decoder_input_ids", "decoder_pad_token_id", "def", "device", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "eos_token_id", "forward", "generate", "generate_kwargs", "get_input_embeddings", "hidden_states", "if", "image_attention_mask", "image_embeds", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "logits", "long", "loss", "main_input_name", "mean", "no_grad", "not", "ones", "outputs", "pad_token_id", "pixel_values", "post_init", "predictions", "r", "reduction", "repeat", "return", "self", "sep_token_id", "set_input_embeddings", "shape", "size", "super", "text_config", "text_decoder", "to", "torch", "tuple", "value", "vision_config", "vision_model", "vision_outputs"], "blip/modeling_blip.py:BlipForQuestionAnswering": ["Either", "False", "FloatTensor", "GenerationMixin", "LongTensor", "ModelConfig", "ModelForQuestionAnswering", "ModelPreTrainedModel", "ModelTextLMHeadModel", "ModelTextModel", "ModelTextVisionModelOutput", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "__init__", "_tied_weights_keys", "add_pooling_layer", "and", "answer_output", "are", "attention_mask", "attentions", "auto_docstring", "be", "bias", "bool", "bos_ids", "bos_token_id", "call", "calling", "can_return_tuple", "class", "cls", "config", "decoder", "decoder_attention_mask", "decoder_input_ids", "decoder_loss", "decoder_pad_token_id", "decoder_start_token_id", "def", "device", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "eos_token_id", "fill_value", "for", "forward", "full", "generate", "generate_kwargs", "get_input_embeddings", "hidden_states", "if", "image_attention_mask", "image_embeds", "inference", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "long", "loss", "make", "mean", "model", "no_grad", "not", "ones", "or", "outputs", "pad_token_id", "passed", "pixel_values", "post_init", "predictions", "question_attention_mask", "question_embeds", "question_outputs", "r", "raise", "reduction", "return", "return_dict", "self", "sep_token_id", "set_input_embeddings", "should", "size", "super", "sure", "text_config", "text_decoder", "text_encoder", "that", "the", "torch", "training", "tuple", "using", "value", "vision_config", "vision_model", "vision_outputs", "when", "with", "you"], "blip/modeling_blip.py:BlipForImageTextRetrieval": ["False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForImageTextRetrieval", "ModelImageTextMatchingModelOutput", "ModelPreTrainedModel", "ModelTextModel", "ModelTextVisionModelOutput", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "bos_token_id", "can_return_tuple", "class", "config", "decoder_pad_token_id", "decoder_start_token_id", "def", "dim", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "hasattr", "hidden_size", "hidden_states", "if", "image_atts", "image_embeds", "image_feat", "image_text_hidden_size", "input_ids", "interpolate_pos_encoding", "itm_head", "itm_score", "kwargs", "last_hidden_state", "long", "nn", "normalize", "not", "ones", "output", "pad_token_id", "pixel_values", "post_init", "question_embeds", "r", "return", "self", "set_input_embeddings", "size", "super", "t", "text_config", "text_encoder", "text_feat", "text_proj", "torch", "tuple", "use_itm_head", "value", "vision_config", "vision_model", "vision_outputs", "vision_proj"], "blip/modeling_blip_text.py:BlipTextEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "class", "config", "def", "dropout", "else", "embeddings", "eps", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "torch", "vocab_size", "word_embeddings"], "blip/modeling_blip_text.py:BlipTextSelfAttention": ["Cache", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "Linear", "ModelTextSelfAttention", "Module", "None", "Optional", "Softmax", "Tensor", "The", "True", "ValueError", "_", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_map", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_probs_dropped", "attention_scores", "attn_gradients", "batch_size", "bhld", "bhlr", "bhrd", "bool", "cache_position", "class", "config", "context_layer", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "d", "def", "deprecate_kwarg", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_size", "encoder_hidden_states", "forward", "get", "get_attention_map", "get_attn_gradients", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_updated", "isinstance", "key", "key_layer", "keys", "layer_idx", "layers", "long", "lrd", "math", "matmul", "max_position_embeddings", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "past_key_value", "past_key_values", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "save_attention_map", "save_attn_gradients", "self", "self_attention_cache", "seq_length", "shape", "size", "sqrt", "super", "the", "to", "torch", "transpose", "tuple", "update", "value", "value_layer", "values", "version", "view"], "blip/modeling_blip_text.py:BlipTextSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "blip/modeling_blip_text.py:BlipTextAttention": ["Cache", "False", "FloatTensor", "ModelTextAttention", "ModelTextSelfAttention", "ModelTextSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "cache_position", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_cross_attention", "key", "layer_idx", "len", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "version"], "blip/modeling_blip_text.py:BlipTextIntermediate": ["ACT2FN", "Linear", "ModelTextIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "blip/modeling_blip_text.py:BlipTextOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "blip/modeling_blip_text.py:BlipTextLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "ModelTextAttention", "ModelTextIntermediate", "ModelTextLayer", "ModelTextOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "cache_position", "chunk_size_feed_forward", "class", "config", "cross_attention_outputs", "crossattention", "def", "deprecate_kwarg", "encoder_attention_mask", "encoder_hidden_states", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "if", "intermediate", "intermediate_output", "is", "is_cross_attention", "is_decoder", "layer_idx", "layer_num", "layer_output", "new_name", "not", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple", "version"], "blip/modeling_blip_text.py:BlipTextEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "ModelTextEncoder", "ModelTextLayer", "Module", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "You", "__init__", "a", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "attention_mask", "attentions", "be", "bool", "cache_position", "checkpointing", "class", "config", "cross_attentions", "def", "deprecated", "e", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "from_legacy_cache", "g", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "logger", "nn", "not", "num_hidden_layers", "of", "output_attentions", "output_hidden_states", "pass", "past_key_values", "range", "removed", "return", "return_dict", "self", "should", "super", "torch", "training", "tuple", "use_cache", "v", "v4", "warning", "warning_once", "will", "with"], "blip/modeling_blip_text.py:BlipTextPooler": ["Linear", "ModelTextPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "blip/modeling_blip_text.py:BlipTextPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelTextPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "blip/modeling_blip_text.py:BlipTextLMPredictionHead": ["False", "Linear", "ModelTextLMPredictionHead", "ModelTextPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "blip/modeling_blip_text.py:BlipTextOnlyMLMHead": ["ModelTextLMPredictionHead", "ModelTextOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "blip/modeling_blip_text.py:BlipTextPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelTextConfig", "ModelTextPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "and", "base_model_prefix", "bert", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "self", "std", "weight", "zero_"], "blip/modeling_blip_text.py:BlipTextModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "False", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextModel", "ModelTextPooler", "ModelTextPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "Wrong", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "and", "arange", "at", "attention", "attention_mask", "attentions", "axis", "batch_size", "bool", "both", "cache_position", "cannot", "cat", "causal_mask", "class", "config", "cross_attentions", "def", "device", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_embeds", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "f", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "int", "invert_attention_mask", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "list", "mask", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_key_values_length", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prefix_seq_len", "prune_heads", "r", "raise", "repeat", "return", "return_dict", "same", "self", "seq_ids", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "the", "time", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings"], "blip/modeling_blip_text.py:BlipTextLMHeadModel": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "GenerationMixin", "ModelTextLMHeadModel", "ModelTextModel", "ModelTextOnlyMLMHead", "ModelTextPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "bert", "bias", "bool", "cache_position", "class", "cls", "config", "contiguous", "cross_attentions", "decoder", "def", "device", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "label_smoothing", "labels", "lm_loss", "logits", "loss", "loss_fct", "mean", "model_inputs", "model_kwargs", "new_embeddings", "none", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "reduction", "return", "return_dict", "return_logits", "self", "sequence_output", "set_input_embeddings", "set_output_embeddings", "shifted_prediction_scores", "size", "str", "sum", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "sew/modeling_sew.py:SEWNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "Model", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "sew/modeling_sew.py:SEWLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "Model", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "sew/modeling_sew.py:SEWGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "Model", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "sew/modeling_sew.py:SEWPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "Model", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "squeeze_factor", "stride", "super", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "sew/modeling_sew.py:SEWSamePadLayer": ["Model", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "sew/modeling_sew.py:SEWUpsampling": ["ACT2FN", "Linear", "Model", "Module", "__init__", "activation", "bsz", "class", "config", "def", "feat_extract_activation", "forward", "hidden_size", "hidden_states", "if", "nn", "projection", "reshape", "return", "self", "size", "squeeze_factor", "src_embed_dim", "src_len", "super", "tgt_embed_dim", "tgt_len"], "sew/modeling_sew.py:SEWFeatureEncoder": ["False", "Model", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "sew/modeling_sew.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "sew/modeling_sew.py:SEWAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "sew/modeling_sew.py:SEWFeedForward": ["ACT2FN", "Dropout", "Linear", "Model", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "sew/modeling_sew.py:SEWEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelFeedForward", "None", "_", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super"], "sew/modeling_sew.py:SEWEncoder": ["AvgPool1d", "BaseModelOutput", "Dropout", "False", "LayerNorm", "Model", "ModelLayer", "ModelPositionalConvEmbedding", "ModelUpsampling", "Module", "ModuleList", "None", "True", "_", "__init__", "_attn_implementation", "all_hidden_states", "all_self_attentions", "and", "arange", "attention_ids", "attention_mask", "attentions", "class", "config", "def", "device", "dropout", "dropout_probability", "dtype", "else", "eps", "expand", "expand_attention_mask", "finfo", "flash_attention_2", "for", "forward", "functional", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "input_lengths", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "long", "max_encoder_length", "min", "min_length", "n_input_timesteps", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "output_lengths", "pad", "pool", "pooled_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "self", "shape", "size", "skip_the_layer", "squeeze_factor", "sum", "super", "synced_gpus", "to", "torch", "training", "transpose", "tuple", "unsqueeze", "upsample", "v", "view"], "sew/modeling_sew.py:SEWPreTrainedModel": ["Conv1d", "False", "GatheredParameters", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelPositionalConvEmbedding", "None", "PreTrainedModel", "True", "Union", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "bool", "class", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "data", "deepspeed", "def", "device", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "hasattr", "if", "in", "in_channels", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "is_deepspeed_zero3_enabled", "isinstance", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "modifier_rank", "module", "nn", "normal_", "not", "output_lengths", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "sum", "supports_gradient_checkpointing", "to", "torch", "weight", "weight_g", "weight_v", "with", "zero", "zero_", "zeros", "zip"], "sew/modeling_sew.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "sew/modeling_sew.py:SEWModel": ["BaseModelOutput", "Dropout", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelEncoder", "ModelFeatureEncoder", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_get_feature_vector_attention_mask", "_mask_hidden_states", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "conv_dim", "def", "device", "dtype", "elif", "else", "encoder", "encoder_outputs", "eps", "expand", "extract_features", "feat_proj_dropout", "feature_dropout", "feature_extractor", "feature_projection", "forward", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "layer_norm", "layer_norm_eps", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "project_features", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "uniform_", "use_return_dict"], "sew/modeling_sew.py:SEWForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "sew/modeling_sew.py:SEWForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "gpt_oss/modeling_gpt_oss.py:GptOssRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "gpt_oss/modeling_gpt_oss.py:GptOssExperts": ["ModelExperts", "Module", "None", "Parameter", "Tensor", "_", "__init__", "alpha", "batch_size", "bmm", "clamp", "class", "config", "continue", "cpu", "current_state", "def", "device", "dim", "down_proj", "down_proj_bias", "dtype", "else", "empty", "expert_dim", "expert_hit", "expert_idx", "expert_mask", "for", "forward", "functional", "gate", "gate_up", "gate_up_proj", "gate_up_proj_bias", "gated_output", "glu", "greater", "hidden_size", "hidden_states", "if", "in", "index_add_", "intermediate_size", "limit", "max", "min", "next_states", "nn", "no_grad", "nonzero", "num_classes", "num_experts", "num_local_experts", "one_hot", "or", "out", "permute", "repeat", "reshape", "return", "router_indices", "routing_weights", "self", "shape", "sigmoid", "sum", "super", "to", "token_idx", "torch", "training", "transpose", "type", "up", "view", "weighted_output", "where", "with", "zeros_like"], "gpt_oss/modeling_gpt_oss.py:GptOssTopKRouter": ["F", "ModelTopKRouter", "Module", "Parameter", "__init__", "bias", "class", "config", "def", "dim", "dtype", "empty", "forward", "functional", "hidden_dim", "hidden_size", "hidden_states", "linear", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "reshape", "return", "router_indices", "router_logits", "router_scores", "router_top_value", "scatter_", "self", "softmax", "super", "top_k", "topk", "torch", "weight", "zeros_like"], "gpt_oss/modeling_gpt_oss.py:GptOssMLP": ["ModelExperts", "ModelMLP", "ModelTopKRouter", "Module", "__init__", "class", "config", "def", "experts", "forward", "hidden_states", "nn", "return", "routed_out", "router", "router_indices", "router_scores", "routing_weights", "self", "super"], "gpt_oss/modeling_gpt_oss.py:GptOssRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gpt_oss/modeling_gpt_oss.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "gpt_oss/modeling_gpt_oss.py:_apply_rotary_emb": ["Tensor", "_apply_rotary_emb", "cat", "chunk", "cos", "def", "dim", "first_", "first_half", "return", "second_", "second_half", "sin", "torch", "x"], "gpt_oss/modeling_gpt_oss.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "_Model_rotary_emb", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "sin", "unsqueeze", "unsqueeze_dim"], "gpt_oss/modeling_gpt_oss.py:eager_attention_forward": ["F", "Model_attention_forward", "Module", "None", "Optional", "Tensor", "True", "attention_mask", "attn_output", "attn_weights", "cat", "causal_mask", "combined_logits", "contiguous", "def", "dim", "dropout", "dtype", "expand", "float", "functional", "if", "is", "keepdim", "key", "key_states", "kwargs", "matmul", "max", "module", "nn", "not", "num_key_value_groups", "p", "probs", "query", "repeat_kv", "reshape", "return", "scaling", "scores", "shape", "sinks", "softmax", "torch", "training", "transpose", "value", "value_states", "values"], "gpt_oss/modeling_gpt_oss.py:GptOssAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "empty", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "s_aux", "scaling", "self", "shape", "sin", "sinks", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "gpt_oss/modeling_gpt_oss.py:GptOssDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "gpt_oss/modeling_gpt_oss.py:GptOssPreTrainedModel": ["Embedding", "False", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelExperts", "ModelPreTrainedModel", "ModelRMSNorm", "ModelTopKRouter", "None", "OutputRecorder", "Parameter", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_keep_in_fp32_modules", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attention", "_supports_flash_attn", "_supports_flex_attention", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "down_proj", "down_proj_bias", "elif", "fill_", "gate_up_proj", "gate_up_proj_bias", "hidden_states", "if", "index", "initializer_range", "input_layernorm", "is", "isinstance", "mean", "model", "module", "nn", "norm", "normal_", "not", "padding_idx", "past_key_values", "post_attention_layernorm", "router_logits", "self", "sinks", "std", "supports_gradient_checkpointing", "weight", "zero_"], "gpt_oss/modeling_gpt_oss.py:GptOssModel": ["Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_no_split_modules", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "gpt_oss/modeling_gpt_oss.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "gpt_oss/modeling_gpt_oss.py:GptOssForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "gpt_oss/modeling_gpt_oss.py:GptOssForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "gpt_oss/modeling_gpt_oss.py:GptOssForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "hubert/modeling_hubert.py:HubertPositionalConvEmbedding": ["ACT2FN", "BatchNorm1d", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "None", "__init__", "activation", "batch_norm", "class", "config", "conv", "conv_pos_batch_norm", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "not", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "hubert/modeling_hubert.py:HubertSamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "hubert/modeling_hubert.py:HubertNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "hubert/modeling_hubert.py:HubertLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "hubert/modeling_hubert.py:HubertGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "hubert/modeling_hubert.py:HubertFeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "hubert/modeling_hubert.py:HubertFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "feat_proj_layer_norm", "forward", "hidden_size", "hidden_states", "if", "layer_norm", "layer_norm_eps", "nn", "projection", "return", "self", "super"], "hubert/modeling_hubert.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "hubert/modeling_hubert.py:HubertAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "hubert/modeling_hubert.py:HubertFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "hubert/modeling_hubert.py:HubertEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelFeedForward", "None", "_", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super"], "hubert/modeling_hubert.py:HubertEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalConvEmbedding", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "tensor", "torch", "training", "tuple", "unsqueeze", "v"], "hubert/modeling_hubert.py:HubertAttnAdapterLayer": ["FloatTensor", "LayerNorm", "Linear", "ModelAttnAdapterLayer", "Module", "ReLU", "__init__", "act_fn", "adapter_attn_dim", "class", "config", "def", "forward", "hidden_dim", "hidden_size", "hidden_states", "input_dim", "linear_1", "linear_2", "nn", "norm", "return", "self", "super", "torch"], "hubert/modeling_hubert.py:HubertEncoderLayerStableLayerNorm": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelAttnAdapterLayer", "ModelEncoderLayerStableLayerNorm", "ModelFeedForward", "None", "Optional", "Tensor", "_", "__init__", "adapter_attn_dim", "adapter_layer", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "bool", "class", "config", "def", "dropout", "else", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "getattr", "hidden_dropout", "hidden_size", "hidden_states", "if", "is", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super", "torch"], "hubert/modeling_hubert.py:HubertEncoderStableLayerNorm": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoderLayerStableLayerNorm", "ModelEncoderStableLayerNorm", "ModelPositionalConvEmbedding", "Module", "ModuleList", "None", "Tensor", "True", "Union", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "unsqueeze", "v"], "hubert/modeling_hubert.py:HubertPreTrainedModel": ["BatchNorm1d", "Conv1d", "GatheredParameters", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "Union", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "bool", "class", "config", "conv_kernel", "conv_stride", "cumsum", "data", "deepspeed", "def", "device", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "hasattr", "if", "in", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "is_deepspeed_zero3_enabled", "isinstance", "kaiming_normal_", "kernel_size", "layer_weights", "long", "main_input_name", "masked_spec_embed", "mean", "modifier_rank", "module", "nn", "normal_", "not", "num_hidden_layers", "output_lengths", "return", "rounding_mode", "self", "shape", "std", "stride", "sum", "supports_gradient_checkpointing", "to", "torch", "uniform_", "weight", "weight_g", "weight_v", "with", "zero", "zero_", "zeros", "zip"], "hubert/modeling_hubert.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "hubert/modeling_hubert.py:HubertModel": ["BaseModelOutput", "FloatTensor", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderStableLayerNorm", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_get_feature_vector_attention_mask", "_mask_hidden_states", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "do_stable_layer_norm", "dtype", "elif", "else", "encoder", "encoder_outputs", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "uniform_", "use_return_dict"], "hubert/modeling_hubert.py:HubertForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "hubert/modeling_hubert.py:HubertForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "swin/modeling_swin.py:SwinEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "reshaped_hidden_states", "torch", "tuple"], "swin/modeling_swin.py:SwinModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "reshaped_hidden_states", "torch", "tuple"], "swin/modeling_swin.py:SwinMaskedImageModelingOutput": ["FloatTensor", "FutureWarning", "ModelMaskedImageModelingOutput", "ModelOutput", "None", "Optional", "Please", "Transformers", "and", "attentions", "attribute", "be", "class", "def", "deprecated", "final", "hidden_states", "in", "instead", "is", "logits", "loss", "of", "output", "property", "r", "reconstruction", "removed", "reshaped_hidden_states", "retrieve", "return", "self", "the", "to", "torch", "tuple", "use", "version", "warn", "warnings", "will"], "swin/modeling_swin.py:SwinImageClassifierOutput": ["FloatTensor", "ModelImageClassifierOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "swin/modeling_swin.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "height", "input_feature", "num_channels", "permute", "return", "shape", "view", "width"], "swin/modeling_swin.py:window_reverse": ["Model_reverse", "Model_size", "Models", "contiguous", "def", "height", "num_channels", "permute", "return", "shape", "view", "width"], "swin/modeling_swin.py:SwinEmbeddings": ["BoolTensor", "Dropout", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "config", "def", "dim", "dropout", "else", "embed_dim", "embeddings", "expand", "forward", "functional", "grid_size", "height", "hidden_dropout_prob", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "norm", "not", "num_channels", "num_patches", "num_positions", "output_dimensions", "patch_embeddings", "patch_grid", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "tuple", "type_as", "unsqueeze", "use_absolute_embeddings", "use_mask_token", "view", "width", "zeros"], "swin/modeling_swin.py:SwinPatchEmbeddings": ["Conv2d", "FloatTensor", "Iterable", "ModelPatchEmbeddings", "Module", "Optional", "Tensor", "_", "__init__", "abc", "class", "collections", "config", "def", "else", "embed_dim", "embeddings", "flatten", "forward", "functional", "grid_size", "height", "hidden_size", "if", "image_size", "int", "isinstance", "kernel_size", "maybe_pad", "nn", "num_channels", "num_patches", "output_dimensions", "pad", "pad_values", "patch_size", "pixel_values", "projection", "return", "self", "shape", "stride", "super", "torch", "transpose", "tuple", "width"], "swin/modeling_swin.py:SwinPatchMerging": ["False", "LayerNorm", "Linear", "ModelPatchMerging", "Module", "None", "Tensor", "__init__", "batch_size", "bias", "cat", "class", "def", "dim", "forward", "functional", "height", "if", "input_dimensions", "input_feature", "input_feature_0", "input_feature_1", "input_feature_2", "input_feature_3", "input_resolution", "int", "maybe_pad", "nn", "norm", "norm_layer", "num_channels", "or", "pad", "pad_values", "reduction", "return", "self", "shape", "should_pad", "super", "torch", "tuple", "view", "width"], "swin/modeling_swin.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "swin/modeling_swin.py:SwinDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "swin/modeling_swin.py:SwinSelfAttention": ["Dropout", "False", "FloatTensor", "Iterable", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "The", "ValueError", "__init__", "a", "abc", "all_head_size", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "collections", "config", "context_layer", "contiguous", "coords", "coords_flatten", "coords_h", "coords_w", "def", "dim", "dropout", "else", "f", "flatten", "forward", "functional", "head_mask", "heads", "hidden", "hidden_shape", "hidden_states", "if", "ij", "indexing", "int", "is", "isinstance", "key", "key_layer", "mask_shape", "math", "matmul", "meshgrid", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "num_channels", "num_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "register_buffer", "relative_coords", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "return", "self", "shape", "size", "softmax", "sqrt", "stack", "sum", "super", "the", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view", "window_size", "zeros"], "swin/modeling_swin.py:SwinSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "swin/modeling_swin.py:SwinAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "swin/modeling_swin.py:SwinIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "swin/modeling_swin.py:SwinOutput": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "swin/modeling_swin.py:SwinLayer": ["False", "FloatTensor", "Identity", "LayerNorm", "ModelAttention", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "always_partition", "attention", "attention_output", "attention_outputs", "attention_windows", "attn_mask", "batch_size", "bool", "channels", "chunk_size_feed_forward", "class", "config", "contiguous", "count", "def", "device", "dim", "dims", "drop_path", "drop_path_rate", "dtype", "else", "eps", "for", "forward", "functional", "get_attn_mask", "head_mask", "height", "height_pad", "height_slice", "height_slices", "hidden_states", "hidden_states_windows", "if", "img_mask", "in", "input_dimensions", "input_resolution", "int", "intermediate", "is_tracing", "jit", "layer_norm_eps", "layer_output", "layer_outputs", "layernorm_after", "layernorm_before", "mask_windows", "masked_fill", "maybe_pad", "min", "nn", "not", "num_heads", "or", "output", "output_attentions", "pad", "pad_bottom", "pad_right", "pad_values", "pass", "return", "roll", "self", "set_shift_and_window_size", "shape", "shift_size", "shifted_hidden_states", "shifted_windows", "shifts", "shortcut", "size", "slice", "super", "tensor", "torch", "torch_int", "tuple", "unsqueeze", "view", "was_padded", "width", "width_pad", "width_slice", "width_slices", "window_partition", "window_reverse", "window_size", "zeros"], "swin/modeling_swin.py:SwinStage": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelLayer", "ModelStage", "ModuleList", "None", "Optional", "Tensor", "__init__", "always_partition", "blocks", "bool", "class", "config", "def", "depth", "dim", "downsample", "drop_path", "drop_path_rate", "else", "enumerate", "for", "forward", "head_mask", "height", "height_downsampled", "hidden_states", "hidden_states_before_downsampling", "i", "if", "in", "input_dimensions", "input_resolution", "int", "is", "layer_head_mask", "layer_module", "layer_outputs", "nn", "norm_layer", "not", "num_heads", "output_attentions", "output_dimensions", "pointing", "range", "return", "self", "shift_size", "stage_outputs", "super", "torch", "tuple", "width", "width_downsampled", "window_size"], "swin/modeling_swin.py:SwinEncoder": ["False", "FloatTensor", "ModelEncoder", "ModelEncoderOutput", "ModelPatchMerging", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "always_partition", "and", "attentions", "batch_size", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "dim", "downsample", "dpr", "drop_path", "drop_path_rate", "elif", "else", "embed_dim", "enumerate", "for", "forward", "gradient_checkpointing", "grid_size", "head_mask", "hidden_size", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "input_dimensions", "input_resolution", "int", "is", "item", "last_hidden_state", "layer_head_mask", "layer_module", "layer_outputs", "layers", "len", "linspace", "nn", "not", "num_heads", "num_layers", "output_attentions", "output_dimensions", "output_hidden_states", "output_hidden_states_before_downsampling", "permute", "range", "reshaped_hidden_state", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "sum", "super", "torch", "tuple", "v", "view", "x"], "swin/modeling_swin.py:SwinPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "ModelSelfAttention", "ModelStage", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "relative_position_bias_table", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "swin/modeling_swin.py:SwinModel": ["AdaptiveAvgPool1d", "BoolTensor", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flatten", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_dimensions", "int", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "len", "nn", "not", "num_features", "num_layers", "output", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_grid", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "tuple", "use_mask_token", "use_return_dict"], "swin/modeling_swin.py:SwinForMaskedImageModeling": ["BoolTensor", "Conv2d", "False", "FloatTensor", "Model", "ModelForMaskedImageModeling", "ModelMaskedImageModelingOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PixelShuffle", "Sequential", "True", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "batch_size", "bool", "bool_masked_pos", "class", "config", "contiguous", "decoder", "def", "else", "embed_dim", "encoder_stride", "floor", "forward", "functional", "head_mask", "height", "hidden_states", "if", "image_size", "in_channels", "int", "interpolate_pos_encoding", "is", "kernel_size", "l1_loss", "loss", "mask", "masked_im_loss", "math", "nn", "none", "not", "num_channels", "num_features", "num_layers", "out_channels", "output", "output_attentions", "output_hidden_states", "outputs", "patch_size", "pixel_values", "post_init", "r", "reconstructed_pixel_values", "reconstruction", "reconstruction_loss", "reduction", "repeat_interleave", "reshape", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_length", "sequence_output", "shape", "size", "sum", "super", "torch", "transpose", "tuple", "unsqueeze", "use_mask_token", "use_return_dict", "width"], "swin/modeling_swin.py:SwinForImageClassification": ["False", "FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelImageClassifierOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_features", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "post_init", "r", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "swin/modeling_swin.py:SwinBackbone": ["BackboneMixin", "BackboneOutput", "LayerNorm", "ModelBackbone", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "ModuleDict", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "_out_features", "always_partition", "attentions", "batch_size", "bool", "channels", "class", "config", "contiguous", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "get_input_embeddings", "head_mask", "height", "hidden_state", "hidden_states", "hidden_states_norms", "i", "if", "in", "input_dimensions", "int", "is", "len", "nn", "not", "num_channels", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "output_hidden_states_before_downsampling", "outputs", "patch_embeddings", "patch_grid", "permute", "pixel_values", "post_init", "range", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "stage", "stage_names", "super", "torch", "use_return_dict", "view", "width", "zip"], "squeezebert/modeling_squeezebert.py:SqueezeBertEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "arange", "class", "config", "def", "device", "dropout", "dtype", "else", "embedding_size", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "squeezebert/modeling_squeezebert.py:MatMulWrapper": ["Model1", "Model2", "ModelMulWrapper", "Modelmul", "Module", "__init__", "class", "def", "forward", "nn", "return", "self", "super", "torch"], "squeezebert/modeling_squeezebert.py:SqueezeBertLayerNorm": ["LayerNorm", "ModelLayerNorm", "__init__", "class", "def", "eps", "forward", "hidden_size", "nn", "normalized_shape", "permute", "return", "self", "x"], "squeezebert/modeling_squeezebert.py:ConvDropoutLayerNorm": ["Dropout", "Model1d", "ModelDropoutLayerNorm", "ModelLayerNorm", "Module", "__init__", "cin", "class", "cout", "def", "dropout", "dropout_prob", "forward", "groups", "hidden_states", "in_channels", "input_tensor", "kernel_size", "layernorm", "nn", "out_channels", "return", "self", "super", "x"], "squeezebert/modeling_squeezebert.py:ConvActivation": ["ACT2FN", "Model1d", "ModelActivation", "Module", "__init__", "act", "cin", "class", "cout", "def", "forward", "groups", "in_channels", "kernel_size", "nn", "out_channels", "output", "return", "self", "super", "x"], "squeezebert/modeling_squeezebert.py:SqueezeBertSelfAttention": ["Conv1d", "Dropout", "MatMulWrapper", "ModelSelfAttention", "Module", "Softmax", "ValueError", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_score", "cin", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "f", "forward", "groups", "heads", "hidden_states", "if", "in_channels", "int", "is", "k_groups", "kernel_size", "key", "key_layer", "math", "matmul_qk", "matmul_qkv", "mixed_key_layer", "mixed_query_layer", "mixed_value_layer", "multiple", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "out_channels", "output_attentions", "permute", "q_groups", "query", "query_layer", "raise", "result", "return", "self", "size", "softmax", "sqrt", "super", "the", "transpose_for_scores", "transpose_key_for_scores", "transpose_output", "v_groups", "value", "value_layer", "view", "x"], "squeezebert/modeling_squeezebert.py:SqueezeBertModule": ["ConvActivation", "ConvDropoutLayerNorm", "ModelModule", "ModelSelfAttention", "Module", "__init__", "act", "att", "attention", "attention_mask", "attention_output", "attention_score", "c0", "c1", "c2", "c3", "cin", "class", "config", "context_layer", "cout", "def", "dropout_prob", "feature_map", "forward", "groups", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "intermediate", "intermediate_groups", "intermediate_output", "intermediate_size", "k_groups", "layer_output", "nn", "output", "output_attentions", "output_dict", "output_groups", "post_attention", "post_attention_groups", "post_attention_output", "q_groups", "return", "self", "super", "v_groups"], "squeezebert/modeling_squeezebert.py:SqueezeBertEncoder": ["BaseModelOutput", "Conv1d", "False", "If", "Model", "ModelEncoder", "ModelModule", "Module", "ModuleList", "None", "True", "_", "__init__", "a", "adjust", "all_attentions", "all_hidden_states", "assert", "attention_mask", "attention_score", "attentions", "before", "channels", "class", "config", "count", "def", "elif", "else", "embedding_size", "feature_map", "first", "for", "forward", "head_mask", "head_mask_is_all_none", "hidden_size", "hidden_states", "if", "implementation", "in", "insert", "intermediate", "is", "last_hidden_state", "layer", "layer_output", "layers", "len", "nn", "not", "num_hidden_layers", "number", "of", "output_attentions", "output_hidden_states", "permute", "please", "range", "return", "return_dict", "self", "super", "supported", "the", "to", "tuple", "v", "want", "yet", "you"], "squeezebert/modeling_squeezebert.py:SqueezeBertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super"], "squeezebert/modeling_squeezebert.py:SqueezeBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "squeezebert/modeling_squeezebert.py:SqueezeBertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "squeezebert/modeling_squeezebert.py:SqueezeBertOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super"], "squeezebert/modeling_squeezebert.py:SqueezeBertPreTrainedModel": ["Conv1d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "transformer", "weight", "zero_"], "squeezebert/modeling_squeezebert.py:SqueezeBertModel": ["BaseModelOutputWithPooling", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "new_embeddings", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "squeezebert/modeling_squeezebert.py:SqueezeBertForMaskedLM": ["CrossEntropyLoss", "MaskedLMOutput", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "transformer", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "squeezebert/modeling_squeezebert.py:SqueezeBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "super", "token_type_ids", "torch", "transformer", "tuple", "use_return_dict", "view"], "squeezebert/modeling_squeezebert.py:SqueezeBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "transformer", "tuple", "use_return_dict", "view"], "squeezebert/modeling_squeezebert.py:SqueezeBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "transformer", "tuple", "use_return_dict", "view"], "squeezebert/modeling_squeezebert.py:SqueezeBertForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "transformer", "tuple", "use_return_dict"], "lfm2_vl/modeling_lfm2_vl.py:Lfm2VlMultiModalProjector": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "Tensor", "__init__", "act", "batch_size", "bias", "channels", "class", "config", "def", "downsample_factor", "factor", "forward", "height", "hidden_size", "hidden_states", "image_features", "in_channels", "layer_norm", "linear_1", "linear_2", "nn", "permute", "pixel_unshuffle", "projector_bias", "projector_hidden_act", "projector_hidden_size", "reshape", "return", "self", "size", "super", "text_config", "torch", "vision_config", "width"], "lfm2_vl/modeling_lfm2_vl.py:Lfm2VlPreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "past_key_values", "supports_gradient_checkpointing"], "lfm2_vl/modeling_lfm2_vl.py:Lfm2VlCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "lfm2_vl/modeling_lfm2_vl.py:Lfm2VlModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "lfm2_vl/modeling_lfm2_vl.py:Lfm2VlModel": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "append", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "device", "dim", "do", "dtype", "else", "exactly", "expand_as", "f", "feature", "feature_org_h", "feature_org_w", "features", "for", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "img_embedding", "img_feature_lengths", "img_idx", "in", "input_ids", "inputs_embeds", "is", "kwargs", "language_model", "last_hidden_state", "list", "long", "masked_scatter", "match", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "outputs", "past_key_values", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "r", "raise", "range", "reshape", "return", "self", "set_decoder", "set_input_embeddings", "shape", "size", "spatial_shapes", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "value", "vision_config", "vision_tower"], "lfm2_vl/modeling_lfm2_vl.py:Lfm2VlForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "spatial_shapes", "super", "text_config", "torch", "tuple", "use_cache", "value", "vision_tower", "vocab_size", "weight"], "superpoint/modeling_superpoint.py:remove_keypoints_from_borders": ["Model_keypoints_from_borders", "Tensor", "border", "def", "height", "int", "keypoints", "mask", "mask_h", "mask_w", "return", "scores", "torch", "tuple", "width"], "superpoint/modeling_superpoint.py:top_k_keypoints": ["Model_k_keypoints", "Modelk", "Tensor", "def", "dim", "if", "indices", "int", "k", "keypoints", "len", "return", "scores", "torch", "tuple"], "superpoint/modeling_superpoint.py:simple_nms": ["Expected", "Model_nms", "Tensor", "ValueError", "_", "def", "float", "for", "functional", "if", "in", "int", "kernel_size", "max_mask", "max_pool", "max_pool2d", "new_max_mask", "nms_radius", "nn", "padding", "positive", "raise", "range", "return", "scores", "stride", "supp_mask", "supp_scores", "torch", "values", "where", "x", "zeros", "zeros_like"], "superpoint/modeling_superpoint.py:SuperPointKeypointDescriptionOutput": ["BoolTensor", "FloatTensor", "IntTensor", "ModelKeypointDescriptionOutput", "ModelOutput", "None", "Optional", "class", "descriptors", "hidden_states", "keypoints", "loss", "mask", "r", "scores", "torch", "tuple"], "superpoint/modeling_superpoint.py:SuperPointConvBlock": ["Conv2d", "False", "MaxPool2d", "Model", "ModelConfig", "ModelConvBlock", "Module", "None", "ReLU", "Tensor", "True", "__init__", "add_pooling", "bool", "class", "config", "conv_a", "conv_b", "def", "else", "forward", "hidden_states", "if", "in_channels", "inplace", "int", "is", "kernel_size", "nn", "not", "out_channels", "padding", "pool", "relu", "return", "self", "stride", "torch"], "superpoint/modeling_superpoint.py:SuperPointEncoder": ["BaseModelOutputWithNoAttention", "False", "Model", "ModelConfig", "ModelConvBlock", "ModelEncoder", "Module", "ModuleList", "None", "Optional", "True", "Union", "__init__", "add_pooling", "all_hidden_states", "append", "bool", "class", "config", "conv_block", "conv_blocks", "def", "else", "encoder_hidden_sizes", "for", "forward", "hidden_states", "i", "if", "in", "input", "input_dim", "is", "last_hidden_state", "len", "nn", "not", "output", "output_hidden_states", "range", "return", "return_dict", "self", "tuple", "v"], "superpoint/modeling_superpoint.py:SuperPointInterestPointDecoder": ["Conv2d", "MaxPool2d", "Model", "ModelConfig", "ModelInterestPointDecoder", "Module", "None", "ReLU", "Tensor", "True", "_", "__init__", "_extract_keypoints", "_get_pixel_scores", "batch_size", "border_removal_distance", "class", "config", "conv_score_a", "conv_score_b", "decoder_hidden_size", "def", "dtype", "encoded", "encoder_hidden_sizes", "flip", "forward", "functional", "height", "if", "inplace", "kernel_size", "keypoint_decoder_dim", "keypoint_threshold", "keypoints", "max_keypoints", "nms_radius", "nn", "nonzero", "padding", "permute", "pool", "relu", "remove_keypoints_from_borders", "reshape", "return", "scores", "self", "shape", "simple_nms", "softmax", "stride", "t", "to", "top_k_keypoints", "torch", "tuple", "width"], "superpoint/modeling_superpoint.py:SuperPointDescriptorDecoder": ["Conv2d", "MaxPool2d", "Model", "ModelConfig", "ModelDescriptorDecoder", "Module", "None", "ReLU", "Tensor", "True", "__init__", "_sample_descriptors", "align_corners", "batch_size", "bilinear", "class", "config", "conv_descriptor_a", "conv_descriptor_b", "decoder_hidden_size", "def", "descriptor_decoder_dim", "descriptors", "dim", "divisor", "encoded", "encoder_hidden_sizes", "forward", "functional", "grid_sample", "height", "inplace", "int", "kernel_size", "keypoints", "kwargs", "mode", "nn", "normalize", "num_channels", "p", "padding", "pool", "relu", "reshape", "return", "scale", "self", "shape", "staticmethod", "stride", "tensor", "to", "torch", "transpose", "view", "width"], "superpoint/modeling_superpoint.py:SuperPointPreTrainedModel": ["Conv2d", "False", "FloatTensor", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "Union", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "extract_one_channel_pixel_values", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "return", "self", "std", "supports_gradient_checkpointing", "torch", "weight", "zero_"], "superpoint/modeling_superpoint.py:SuperPointForKeypointDetection": ["FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDescriptorDecoder", "ModelEncoder", "ModelForKeypointDetection", "ModelInterestPointDecoder", "ModelKeypointDescriptionOutput", "ModelPreTrainedModel", "None", "Optional", "Union", "ValueError", "_", "__init__", "_descriptors", "_keypoints", "_scores", "auto_docstring", "batch_size", "bool", "class", "config", "def", "descriptor_decoder", "descriptor_decoder_dim", "descriptors", "device", "does", "dtype", "else", "encoder", "encoder_outputs", "enumerate", "extract_one_channel_pixel_values", "for", "forward", "height", "hidden_states", "i", "if", "in", "int", "is", "keypoint_decoder", "keypoints", "keypoints_scores", "labels", "last_hidden_state", "list_descriptors", "list_keypoints", "list_keypoints_scores", "list_scores", "loss", "mask", "max", "maximum_num_keypoints", "not", "now", "output_hidden_states", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "scores", "self", "shape", "support", "tensor", "torch", "training", "tuple", "use_return_dict", "v", "width", "zeros", "zip"], "gemma2/modeling_gemma2.py:Gemma2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "gemma2/modeling_gemma2.py:Gemma2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_activation", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "gemma2/modeling_gemma2.py:Gemma2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "gemma2/modeling_gemma2.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "gemma2/modeling_gemma2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "gemma2/modeling_gemma2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "gemma2/modeling_gemma2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_dim", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softcap", "softmax", "tanh", "to", "torch", "training", "transpose", "tuple", "value", "value_states"], "gemma2/modeling_gemma2.py:Gemma2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_logit_softcapping", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_pre_attn_scalar", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "softcap", "super", "torch", "training", "transpose", "tuple", "update", "use_bidirectional_attention", "v_proj", "value_states", "version", "view"], "gemma2/modeling_gemma2.py:Gemma2DecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "pre_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "gemma2/modeling_gemma2.py:Gemma2PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "RMSNorm", "True", "__class__", "__name__", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "in", "model", "module", "past_key_values", "self", "super", "supports_gradient_checkpointing", "weight", "zero_"], "gemma2/modeling_gemma2.py:Gemma2Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "checkpointing", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "dtype", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "logger", "mask_kwargs", "must", "nn", "norm", "normalizer", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "tensor", "torch", "training", "unsqueeze", "use_cache", "vocab_size", "warning_once", "with"], "gemma2/modeling_gemma2.py:Gemma2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "final_logit_softcapping", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "tanh", "torch", "use_cache", "vocab_size", "weight"], "gemma2/modeling_gemma2.py:Gemma2ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "gemma2/modeling_gemma2.py:Gemma2ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "git/modeling_git.py:GitVisionModelOutput": ["FloatTensor", "ModelOutput", "ModelVisionModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "git/modeling_git.py:GitEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "class", "config", "def", "dropout", "else", "embeddings", "eps", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "torch", "vocab_size", "word_embeddings"], "git/modeling_git.py:GitSelfAttention": ["Cache", "Dropout", "Embedding", "False", "FloatTensor", "Instantiating", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "ValueError", "_", "__class__", "__init__", "__name__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bhld", "bhlr", "bhrd", "bool", "caching", "call", "cat", "class", "config", "context_layer", "contiguous", "creating", "cutoff", "def", "deprecate_kwarg", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "during", "einsum", "elif", "else", "embedding_size", "errors", "f", "forward", "functional", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "image_patch_tokens", "image_size", "int", "is", "key", "key_layer", "key_layer_past", "key_length", "layer_idx", "lead", "logger", "long", "lrd", "make", "math", "matmul", "max_position_embeddings", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "num_image_with_embedding", "number", "of", "or", "output_attentions", "passing", "past_key_value", "past_key_values", "patch_size", "permute", "pixel_values_present", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "provide", "query", "query_layer", "query_length", "raise", "recommended", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "sure", "tensor", "the", "this", "to", "torch", "transpose", "tuple", "update", "used", "value", "value_layer", "value_layer_past", "version", "view", "vision_config", "warning_once", "when", "will", "without"], "git/modeling_git.py:GitSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "git/modeling_git.py:GitAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelSelfOutput", "Model_SELF_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "attn_output", "bool", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "layer_idx", "len", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "past_key_value", "past_key_values", "pixel_values_present", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_weights", "set", "super", "torch", "tuple", "union", "value", "version"], "git/modeling_git.py:GitIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "git/modeling_git.py:GitOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "git/modeling_git.py:GitLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "deprecate_kwarg", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "layer_idx", "layer_output", "new_name", "output", "output_attentions", "past_key_value", "past_key_values", "pixel_values_present", "return", "self", "self_attention_weights", "seq_len_dim", "super", "torch", "tuple", "version"], "git/modeling_git.py:GitEncoder": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "checkpointing", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "logger", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "past_key_values", "pixel_values_present", "range", "return", "return_dict", "self", "super", "torch", "training", "tuple", "use_cache", "v", "warning_once", "with"], "git/modeling_git.py:GitPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "patch_embedding", "position_embedding", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "git/modeling_git.py:GitVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "git/modeling_git.py:GitVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "git/modeling_git.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "git/modeling_git.py:GitVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "git/modeling_git.py:GitVisionEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEncoderLayer", "ModelVisionMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "git/modeling_git.py:GitVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "git/modeling_git.py:GitVisionTransformer": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "ValueError", "You", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_layernorm", "pre_layrnorm", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "git/modeling_git.py:GitVisionModel": ["BaseModelOutput", "False", "FloatTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embeddings", "forward", "get_input_embeddings", "if", "interpolate_pos_encoding", "is", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "vision_model"], "git/modeling_git.py:GitProjection": ["LayerNorm", "Linear", "ModelConfig", "ModelProjection", "Module", "Sequential", "Tensor", "__init__", "class", "config", "def", "embeddings", "eps", "forward", "hidden_size", "layer_norm_eps", "nn", "return", "self", "super", "torch", "vision_config", "visual_projection"], "git/modeling_git.py:GitModel": ["BaseModelOutputWithPast", "BaseModelOutputWithPooling", "Cache", "False", "FloatTensor", "Memory", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelProjection", "ModelVisionModel", "None", "Optional", "Parameter", "ParameterList", "Tensor", "Union", "ValueError", "You", "_", "__init__", "_generate_future_mask", "_prepare_4d_attention_mask", "_prune_heads", "a", "and", "append", "at", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bool", "boolean", "both", "bottom_left", "cannot", "cat", "class", "clone", "combined_attention_mask", "config", "create_attention_mask", "def", "device", "diagonal", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "expanded_attn_mask", "fill_value", "float", "for", "forward", "frame_idx", "full", "full_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "image_encoder", "img_temporal_embedding", "in", "inf", "input_ids", "input_shape", "inputs_embeds", "int", "interpolate_pos_encoding", "is", "isinstance", "items", "key", "last_hidden_state", "layer", "left", "list", "mask", "masked_fill", "memory", "memory_key_padding_mask", "must", "ndim", "nn", "not", "num_hidden_layers", "num_image_with_embedding", "num_memory", "num_tgt", "of", "ones", "or", "origin_left", "output_attentions", "output_hidden_states", "padding", "past_key_values", "past_key_values_length", "pixel_values", "pixel_values_present", "position_ids", "post_init", "projected_visual_features", "prune_heads", "r", "raise", "range", "rank", "repeat", "return", "return_dict", "right", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "tensor", "tgt", "tgt_len", "tgt_mask", "the", "time", "to", "top_left", "top_right", "torch", "triu", "tuple", "update", "use_cache", "use_return_dict", "value", "vision_config", "visual_features", "visual_features_frame", "visual_projection", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zero_negative_infinity", "zeros", "zeros_like"], "git/modeling_git.py:GitForCausalLM": ["Cache", "CausalLMOutputWithPast", "False", "GenerationMixin", "Linear", "Model", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "contiguous", "def", "else", "encoder", "for", "forward", "get", "get_output_embeddings", "get_seq_length", "head_mask", "hidden_size", "hidden_states", "if", "image_patch_tokens", "in", "input_ids", "input_shape", "inputs_embeds", "interpolate_pos_encoding", "is", "items", "key", "kwargs", "labels", "layer", "list", "loModels", "loss", "loss_function", "model_inputs", "new_embeddings", "new_ones", "nn", "not", "num_image_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "past_length", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "remove_prefix_length", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "shifted_loModels", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Identity", "Model", "Module", "None", "Tensor", "__init__", "activation", "bias", "class", "convolution", "def", "else", "forward", "hidden_state", "if", "in_channels", "input", "int", "is", "kernel_size", "nn", "normalization", "not", "out_channels", "padding", "relu", "return", "self", "str", "stride", "super"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetEmbeddings": ["Make", "MaxPool2d", "Model", "ModelResNetConfig", "ModelResNetConvLayer", "Module", "Sequential", "Tensor", "ValueError", "__init__", "activation", "channel", "class", "config", "configuration", "def", "dimension", "embedder", "embedding", "embedding_size", "forward", "hidden_act", "if", "in", "kernel_size", "match", "nn", "num_channels", "of", "one", "padding", "pixel", "pixel_values", "pooler", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "values", "with"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetShortCut": ["BatchNorm2d", "Conv2d", "False", "Model", "Module", "Tensor", "__init__", "bias", "class", "convolution", "def", "forward", "hidden_state", "in_channels", "input", "int", "kernel_size", "nn", "normalization", "out_channels", "return", "self", "stride", "super"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetBasicLayer": ["ACT2FN", "AvgPool2d", "False", "Identity", "Model", "ModelResNetConfig", "ModelResNetConvLayer", "ModelResNetShortCut", "Module", "None", "Sequential", "True", "__init__", "activation", "bool", "ceil_mode", "class", "config", "def", "else", "forward", "hidden_act", "hidden_state", "if", "in_channels", "int", "layer", "nn", "out_channels", "residual", "return", "self", "shortcut", "should_apply_shortcut", "stride", "super"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetBottleNeckLayer": ["ACT2FN", "AvgPool2d", "Identity", "Model", "ModelResNetConfig", "ModelResNetConvLayer", "ModelResNetShortCut", "Module", "None", "Sequential", "True", "__init__", "activation", "ceil_mode", "class", "config", "def", "downsample_in_bottleneck", "else", "forward", "hidden_act", "hidden_state", "if", "in_channels", "int", "kernel_size", "layer", "nn", "not", "or", "out_channels", "reduces_channels", "reduction", "residual", "return", "self", "shortcut", "should_apply_shortcut", "stride", "super"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetStage": ["Model", "ModelResNetBasicLayer", "ModelResNetBottleNeckLayer", "ModelResNetConfig", "Module", "Sequential", "Tensor", "True", "_", "__init__", "bottleneck", "class", "config", "def", "depth", "else", "first_layer", "for", "forward", "hidden_state", "if", "in", "in_channels", "input", "int", "layer", "layer_type", "layers", "nn", "out_channels", "range", "return", "self", "should_apply_shortcut", "stride", "super"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetEncoder": ["BaseModelOutputWithNoAttention", "False", "Model", "ModelResNetConfig", "ModelResNetStage", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "append", "bool", "class", "config", "def", "depth", "depths", "downsample_in_first_stage", "else", "embedding_size", "for", "forward", "hidden_sizes", "hidden_state", "hidden_states", "if", "in", "in_channels", "in_out_channels", "is", "last_hidden_state", "nn", "not", "out_channels", "output_hidden_states", "return", "return_dict", "self", "stage_module", "stages", "stride", "super", "tuple", "v", "zip"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetPreTrainedModel": ["BatchNorm2d", "Conv2d", "GroupNorm", "Linear", "Model", "ModelResNetConfig", "ModelResNetConvLayer", "ModelResNetShortCut", "None", "PreTrainedModel", "_", "_calculate_fan_in_and_fan_out", "_init_weights", "_no_split_modules", "a", "base_model_prefix", "bias", "bound", "class", "config", "constant_", "def", "elif", "else", "fan_in", "fan_out", "if", "init", "is", "isinstance", "kaiming_normal_", "kaiming_uniform_", "main_input_name", "math", "mode", "module", "nn", "nonlinearity", "not", "pixel_values", "relu", "resnet", "self", "sqrt", "uniform_", "weight"], "rt_detr/modeling_rt_detr_resnet.py:RTDetrResNetBackbone": ["BackboneMixin", "BackboneOutput", "False", "Model", "ModelResNetEmbeddings", "ModelResNetEncoder", "ModelResNetPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedder", "embedding_output", "embedding_size", "encoder", "enumerate", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_features", "out_features", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "stage", "stage_names", "super", "use_return_dict"], "rt_detr/modeling_rt_detr.py:MultiScaleDeformableAttention": ["False", "ModelScaleDeformableAttention", "Module", "Tensor", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "class", "contiguous", "def", "dim", "enumerate", "flatten", "for", "forward", "functional", "grid_sample", "height", "hidden_dim", "im2col_step", "in", "int", "level_id", "level_start_index", "list", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "self", "shape", "split", "stack", "sum", "torch", "transpose", "tuple", "value", "value_l_", "value_list", "value_spatial_shapes", "value_spatial_shapes_list", "view", "width", "zeros"], "rt_detr/modeling_rt_detr.py:RTDetrDecoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "rt_detr/modeling_rt_detr.py:RTDetrModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "denoising_meta_values", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "rt_detr/modeling_rt_detr.py:RTDetrObjectDetectionOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "denoising_meta_values", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "rt_detr/modeling_rt_detr.py:_get_clones": ["ModuleList", "N", "_get_clones", "def", "for", "i", "in", "nn", "partial_module", "range", "return"], "rt_detr/modeling_rt_detr.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "rt_detr/modeling_rt_detr.py:RTDetrFrozenBatchNorm2d": ["Model", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "rt_detr/modeling_rt_detr.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "rt_detr/modeling_rt_detr.py:get_contrastive_denoising_training_group": ["Model_contrastive_denoising_training_group", "None", "attn_mask", "batch_size", "bool", "box_noise_scale", "boxes", "center_to_corners_format", "class_embed", "class_labels", "clip_", "corners_to_center_format", "def", "denoise_positive_idx", "denoising_meta_values", "device", "diff", "dn_num_group", "dn_num_split", "dn_positive_idx", "dtype", "else", "float", "for", "full", "i", "idx_block_end", "idx_block_start", "if", "in", "inf", "input_query_bbox", "input_query_class", "int32", "inverse_sigmoid", "known_bbox", "label_noise_ratio", "len", "mask", "max", "max_gt_num", "min", "n", "negative_gt_mask", "new_label", "nonzero", "num_classes", "num_denoising_queries", "num_ground_truths", "num_groups_denoising_queries", "num_gt", "num_queries", "pad_gt_mask", "positive_gt_mask", "rand_like", "rand_part", "rand_sign", "randint_like", "range", "return", "split", "squeeze", "t", "tarModel_size", "tarModels", "tile", "torch", "torch_int", "where", "zeros"], "rt_detr/modeling_rt_detr.py:RTDetrConvEncoder": ["Model", "Module", "None", "Tensor", "__init__", "append", "backbone", "bool", "channels", "class", "config", "def", "feature_map", "feature_maps", "features", "float", "for", "forward", "freeze_backbone_batch_norms", "functional", "if", "in", "intermediate_channel_sizes", "interpolate", "load_backbone", "mask", "model", "nn", "no_grad", "out", "pixel_mask", "pixel_values", "replace_batch_norm", "return", "self", "shape", "size", "super", "to", "torch", "with"], "rt_detr/modeling_rt_detr.py:RTDetrConvNormLayer": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "Identity", "Model", "Module", "None", "__init__", "activation", "batch_norm_eps", "bias", "class", "config", "conv", "def", "else", "forward", "hidden_state", "if", "in_channels", "is", "kernel_size", "nn", "norm", "out_channels", "padding", "return", "self", "stride", "super"], "rt_detr/modeling_rt_detr.py:RTDetrEncoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "def", "dropout", "dtype", "embed_dim", "encoder_activation_function", "encoder_ffn_dim", "encoder_hidden_dim", "eps", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "kwargs", "layer_norm_eps", "max", "min", "nn", "normalize_before", "not", "num_attention_heads", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "rt_detr/modeling_rt_detr.py:RTDetrRepVggBlock": ["ACT2CLS", "Identity", "Model", "ModelConfig", "ModelConvNormLayer", "Module", "None", "__init__", "activation", "activation_function", "class", "config", "conv1", "conv2", "def", "else", "encoder_hidden_dim", "forward", "hidden_channels", "hidden_expansion", "if", "int", "is", "nn", "padding", "return", "self", "super", "x", "y"], "rt_detr/modeling_rt_detr.py:RTDetrCSPRepLayer": ["Identity", "Model", "ModelConfig", "ModelConvNormLayer", "ModelRepVggBlock", "Module", "Sequential", "_", "__init__", "activation", "activation_function", "bottlenecks", "class", "config", "conv1", "conv2", "conv3", "def", "else", "encoder_hidden_dim", "for", "forward", "hidden_channels", "hidden_expansion", "hidden_state", "hidden_state_1", "hidden_state_2", "if", "in", "in_channels", "int", "nn", "num_blocks", "out_channels", "range", "return", "self", "super"], "rt_detr/modeling_rt_detr.py:RTDetrMultiscaleDeformableAttention": ["CUDA", "F", "False", "Last", "Linear", "Make", "Model", "ModelConfig", "Module", "MultiScaleDeformableAttention", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "attn", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "config", "d", "d_model", "def", "dim", "dim_per_head", "dimension", "disable_custom_kernels", "divisible", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "for", "forward", "got", "head", "height", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "int", "is", "length", "level_start_index", "make", "masked_fill", "more", "must", "n_heads", "n_levels", "n_points", "nn", "not", "num_coordinates", "num_feature_levels", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "reference_points", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "total_elements", "value", "value_proj", "view", "warn", "warnings", "which", "width", "with", "with_pos_embed"], "rt_detr/modeling_rt_detr.py:RTDetrMultiheadAttention": ["Attention", "False", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_reshape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "expand", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "k_proj", "key_states", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "rt_detr/modeling_rt_detr.py:RTDetrDecoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_activation_function", "decoder_attention_heads", "decoder_ffn_dim", "decoder_n_points", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "eps", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "layer_norm_eps", "level_start_index", "n_points", "nn", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "second_residual", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training"], "rt_detr/modeling_rt_detr.py:RTDetrPreTrainedModel": ["BatchNorm2d", "Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelForObjectDetection", "ModelHybridEncoder", "ModelModel", "ModelMultiscaleDeformableAttention", "None", "Parameter", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "abs", "and", "arange", "attention_weights", "base_model_prefix", "bbox_embed", "bias", "class", "class_embed", "config", "constant_", "cos", "data", "def", "default_dtype", "denoising_class_embed", "dtype", "elif", "enc_score_head", "fill_", "float", "for", "get_default_dtype", "grid_init", "hasattr", "i", "if", "in", "init", "initializer_bias_prior_prob", "initializer_range", "int64", "is", "isinstance", "keepdim", "layer", "layers", "learn_initial_query", "log", "main_input_name", "math", "max", "mean", "module", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "num_denoising", "num_labels", "or", "output_proj", "pi", "pixel_values", "prior_prob", "r", "range", "repeat", "sampling_offsets", "self", "sin", "stack", "std", "thetas", "to", "torch", "value_proj", "view", "weight", "weight_embedding", "with", "xavier_uniform_", "zero_"], "rt_detr/modeling_rt_detr.py:RTDetrEncoder": ["False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "attention_mask", "bool", "class", "config", "def", "encoder_layers", "for", "forward", "hidden_states", "in", "layer", "layers", "nn", "output_attentions", "pos_embed", "position_embeddings", "range", "return", "self", "src", "src_mask", "super", "torch"], "rt_detr/modeling_rt_detr.py:RTDetrHybridEncoder": ["BaseModelOutput", "Embed", "F", "Model", "ModelCSPRepLayer", "ModelConfig", "ModelConvNormLayer", "ModelEncoder", "Module", "ModuleList", "None", "ValueError", "_", "__init__", "activation", "activation_function", "all_attentions", "append", "arange", "attention_mask", "attentions", "backbone_feature_map", "be", "build_2d_sincos_position_embedding", "by", "class", "concat", "config", "contiguous", "cos", "cpu", "def", "device", "dim", "dimension", "divisible", "downsample_conv", "downsample_convs", "downsampled_feature_map", "dtype", "else", "embed_dim", "embedding", "enc_ind", "encode_proj_layers", "encoder", "encoder_hidden_dim", "encoder_in_channels", "encoder_layers", "encoder_states", "enumerate", "eval_size", "feat_strides", "flatten", "float32", "for", "forward", "fpn_block", "fpn_blocks", "fpn_feature_map", "fpn_feature_maps", "fused_feature_map", "grid_h", "grid_w", "height", "hidden_states", "i", "idx", "if", "ij", "in", "in_channels", "indexing", "inputs_embeds", "interpolate", "is", "kernel_size", "last_hidden_state", "lateral_conv", "lateral_convs", "layer_outputs", "len", "level_start_index", "meshgrid", "mode", "must", "nearest", "new_fpn_feature_map", "new_pan_feature_map", "nn", "not", "num_fpn_stages", "num_pan_stages", "omega", "or", "out_channels", "out_h", "out_strides", "out_w", "output_attentions", "output_hidden_states", "pan_block", "pan_blocks", "pan_feature_maps", "permute", "pos_dim", "pos_embed", "position", "position_embeddings", "positional_encoding_temperature", "r", "raise", "range", "reshape", "return", "return_dict", "scale_factor", "self", "shape", "sin", "spatial_shapes", "src_flatten", "staticmethod", "stride", "super", "temperature", "to", "top_fpn_feature_map", "top_pan_feature_map", "torch", "torch_int", "training", "tuple", "use_return_dict", "v", "valid_ratios", "width", "zip"], "rt_detr/modeling_rt_detr.py:RTDetrDecoder": ["F", "Model", "ModelConfig", "ModelLayer", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "class", "class_embed", "config", "cross_attentions", "d_model", "decoder_layer", "decoder_layers", "def", "detach", "dim", "dropout", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "hidden_states", "idx", "if", "in", "inputs_embeds", "intermediate", "intermediate_hidden_states", "intermediate_logits", "intermediate_reference_points", "inverse_sigmoid", "is", "last_hidden_state", "layer_outputs", "layers", "level_start_index", "logits", "new_reference_points", "nn", "not", "num_layers", "output_attentions", "output_hidden_states", "position_embeddings", "post_init", "predicted_corners", "query_pos_head", "r", "range", "reference_points", "reference_points_input", "return", "return_dict", "self", "sigmoid", "spatial_shapes", "spatial_shapes_list", "stack", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "v", "valid_ratios"], "rt_detr/modeling_rt_detr.py:RTDetrMLPPredictionHead": ["Linear", "Model", "Module", "ModuleList", "__init__", "class", "config", "d_model", "def", "else", "enumerate", "for", "forward", "functional", "h", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "rt_detr/modeling_rt_detr.py:RTDetrModel": ["BaseModelOutput", "BatchNorm2d", "Conv2d", "Embedding", "F", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelConvEncoder", "ModelDecoder", "ModelHybridEncoder", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Sequential", "True", "Union", "_", "__init__", "_len_sources", "all", "anchor_image_size", "anchors", "and", "append", "arange", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_norm_eps", "batch_size", "bias", "bool", "box_noise_scale", "cat", "class", "class_embed", "compile_compatible_method_lru_cache", "concat", "config", "cpu", "cross_attentions", "cumsum", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_in_channels", "decoder_input_proj", "decoder_input_proj_list", "decoder_inputs_embeds", "decoder_outputs", "def", "denoising_bbox_unact", "denoising_class", "denoising_class_embed", "denoising_meta_values", "detach", "device", "dict", "dim", "dn_outputs", "dtype", "elif", "else", "empty", "enc_bbox_head", "enc_output", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "enc_score_head", "enc_topk_bboxes", "enc_topk_logits", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_dim", "encoder_hidden_states", "encoder_input_proj", "encoder_input_proj_list", "encoder_last_hidden_state", "encoder_outputs", "end", "enumerate", "eps", "feat_strides", "features", "finfo", "flatten", "float32", "for", "forward", "freeze_backbone", "gather", "generate_anchors", "get_contrastive_denoising_training_group", "get_encoder", "grid_size", "grid_x", "grid_xy", "grid_y", "height", "hidden_states", "i", "if", "ij", "in", "in_channels", "index", "indexing", "init_reference_points", "initial_reference_points", "inputs_embeds", "int", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "is", "isinstance", "keepdim", "kernel_size", "label_noise_ratio", "labels", "last_hidden_state", "layer_norm_eps", "learn_initial_query", "len", "level", "level_start_index", "list", "log", "long", "mask", "max", "maxsize", "memory", "meshgrid", "new_zeros", "nn", "not", "num_backbone_outs", "num_channels", "num_classes", "num_denoising", "num_denoising_queries", "num_feature_levels", "num_labels", "num_layers", "num_queries", "ones", "ones_like", "or", "output_attentions", "output_hidden_states", "output_memory", "padding", "padding_idx", "param", "parameters", "pixel_mask", "pixel_values", "post_init", "prod", "proj_feats", "r", "range", "reference_points", "reference_points_unact", "repeat", "requires_grad_", "reshape", "return", "return_dict", "s", "self", "shape", "sigmoid", "source", "source_flatten", "sources", "spatial_shapes", "spatial_shapes_list", "spatial_shapes_tuple", "stack", "stride", "super", "target", "targets", "tensor", "tile", "to", "topk", "topk_ind", "torch", "training", "transpose", "tuple", "tuple_outputs", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_mask", "value", "values", "weight_embedding", "wh", "where", "width"], "rt_detr/modeling_rt_detr.py:RTDetrForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Union", "_", "__init__", "_get_clones", "_no_split_modules", "_set_aux_loss", "_tied_weights_keys", "a", "auto_docstring", "auxiliary_outputs", "b", "bbox_embed", "bool", "class", "class_embed", "config", "cross_attentions", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_layers", "def", "denoising_meta_values", "device", "dict", "else", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "if", "in", "init_reference_points", "initial_reference_points", "inputs_embeds", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "is", "jit", "kwargs", "labels", "last_hidden_state", "list", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_labels", "num_layers", "num_pred", "output", "output_attentions", "output_hidden_states", "outputs", "outputs_class", "outputs_coord", "partial", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "predicted_corners", "r", "range", "return", "return_dict", "self", "super", "torch", "training", "tuple", "unused", "use_return_dict", "with_box_refine", "zip"], "idefics3/modeling_idefics3.py:Idefics3BaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "idefics3/modeling_idefics3.py:Idefics3CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "idefics3/modeling_idefics3.py:Idefics3VisionEmbeddings": ["BoolTensor", "Conv2d", "Embedding", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "True", "_", "__init__", "arange", "batch_idx", "batch_size", "boundaries", "bucket_coords_h", "bucket_coords_w", "bucketize", "class", "config", "def", "device", "dtype", "embed_dim", "embeddings", "enumerate", "fill_value", "flatten", "for", "forward", "fractional_coords_h", "fractional_coords_w", "full", "h_indices", "hidden_size", "image_size", "in", "in_channels", "kernel_size", "max_im_h", "max_im_w", "max_nb_patches_h", "max_nb_patches_w", "nb_patches_h", "nb_patches_w", "nn", "num_channels", "num_patches", "num_patches_per_side", "num_positions", "out_channels", "p_attn_mask", "padding", "patch_attention_mask", "patch_embedding", "patch_embeds", "patch_size", "pixel_values", "pos_ids", "position_embedding", "position_ids", "return", "right", "self", "shape", "size", "stride", "sum", "super", "torch", "transpose", "valid", "view", "w_indices"], "idefics3/modeling_idefics3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "idefics3/modeling_idefics3.py:Idefics3VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "idefics3/modeling_idefics3.py:Idefics3VisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "idefics3/modeling_idefics3.py:Idefics3SimpleMLP": ["False", "Linear", "ModelSimpleMLP", "Module", "__init__", "bias", "class", "config", "def", "forward", "hidden_size", "input_size", "nn", "output_size", "proj", "return", "scale_factor", "self", "super", "text_config", "vision_config", "x"], "idefics3/modeling_idefics3.py:Idefics3EncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelEncoderLayer", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "idefics3/modeling_idefics3.py:Idefics3Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "last_hidden_state", "layer_outputs", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "idefics3/modeling_idefics3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "idefics3/modeling_idefics3.py:Idefics3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "idefics3/modeling_idefics3.py:Idefics3Connector": ["ModelConnector", "ModelSimpleMLP", "Module", "__init__", "bsz", "class", "config", "def", "embed_dim", "forward", "height", "image_hidden_states", "int", "modality_projection", "nn", "permute", "pixel_shuffle", "reshape", "return", "scale_factor", "self", "seq", "size", "super", "view", "width", "x"], "idefics3/modeling_idefics3.py:Idefics3PreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionAttention", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "get_text_config", "getattr", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "idefics3/modeling_idefics3.py:Idefics3VisionTransformer": ["BaseModelOutput", "BoolTensor", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_attn_implementation", "_can_record_outputs", "_prepare_4d_attention_mask", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "any", "attention_mask", "attentions", "batch_size", "bool", "check_model_inputs", "class", "config", "def", "device", "dtype", "elif", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "flash_attention_2", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "ones", "patch_attention_mask", "patch_size", "pixel_values", "post_layernorm", "return", "self", "set_input_embeddings", "size", "super", "to", "torch", "tuple", "value", "view"], "idefics3/modeling_idefics3.py:Idefics3Model": ["AutoModel", "BoolTensor", "Cache", "DynamicCache", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelBaseModelOutputWithPast", "ModelConfig", "ModelConnector", "ModelModel", "ModelPreTrainedModel", "ModelVisionTransformer", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_from_config", "_text_require_grads_hook", "_vision_require_grads_hook", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cache_position", "can_return_tuple", "cannot", "checkpointing", "children", "class", "config", "connector", "contiguous", "custom_intro", "def", "device", "dim", "dimension", "disable_input_require_grads", "dtype", "either", "elif", "else", "enable_input_require_grads", "expand_as", "forward", "from_config", "get_image_features", "get_input_embeddings", "get_lowest_module", "gradient", "gradient_checkpointing", "have", "height", "hidden_states", "if", "image_hidden_states", "image_seq_len", "image_size", "image_token_id", "incompatible", "input", "input_ids", "inputs_embeds", "inputs_merger", "int", "is", "kwargs", "last_hidden_state", "len", "list", "logger", "long", "make_inputs_require_grads", "masked_scatter", "module", "nb_values_per_image", "not", "num_channels", "num_images", "numel", "ones", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "padding_idx", "past_key_values", "patch_attention_mask", "patch_size", "patches_subgrid", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "r", "raise", "real_images_inds", "register_forward_hook", "remove", "requires_grad_", "return", "return_dict", "same", "scale_factor", "self", "seq_length", "set_input_embeddings", "shape", "size", "special_image_mask", "specify", "step", "sum", "super", "tensor", "text_config", "text_model", "the", "time", "to", "torch", "training", "tuple", "unfold", "unsqueeze", "use_cache", "use_return_dict", "value", "view", "vision_config", "vision_model", "vocab_size", "warning_once", "width", "with"], "idefics3/modeling_idefics3.py:Idefics3ForConditionalGeneration": ["BoolTensor", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_text_require_grads_hook", "_tied_weights_keys", "_vision_require_grads_hook", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "disable_input_require_grads", "else", "enable_input_require_grads", "forward", "get_image_features", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_token_id", "input", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "make_inputs_require_grads", "model", "model_inputs", "module", "nn", "not", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "register_forward_hook", "remove", "requires_grad_", "return", "return_dict", "self", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "text_model", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_model", "vocab_size", "weight"], "idefics2/modeling_idefics2.py:Idefics2BaseModelOutputWithPast": ["Cache", "FloatTensor", "ModelBaseModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "idefics2/modeling_idefics2.py:Idefics2CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "idefics2/modeling_idefics2.py:Idefics2VisionEmbeddings": ["BoolTensor", "Conv2d", "Embedding", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "True", "_", "__init__", "arange", "batch_idx", "batch_size", "boundaries", "bucket_coords_h", "bucket_coords_w", "bucketize", "class", "config", "def", "device", "dtype", "embed_dim", "embeddings", "enumerate", "fill_value", "flatten", "for", "forward", "fractional_coords_h", "fractional_coords_w", "full", "h_indices", "hidden_size", "image_size", "in", "in_channels", "kernel_size", "max_im_h", "max_im_w", "max_nb_patches_h", "max_nb_patches_w", "nb_patches_h", "nb_patches_w", "nn", "num_channels", "num_patches", "num_patches_per_side", "num_positions", "out_channels", "p_attn_mask", "padding", "patch_attention_mask", "patch_embedding", "patch_embeds", "patch_size", "pixel_values", "pos_ids", "position_embedding", "position_ids", "return", "right", "self", "shape", "size", "stride", "sum", "super", "torch", "transpose", "valid", "view", "w_indices"], "idefics2/modeling_idefics2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "hasattr", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value"], "idefics2/modeling_idefics2.py:Idefics2VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelVisionAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "idefics2/modeling_idefics2.py:Idefics2VisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "idefics2/modeling_idefics2.py:Idefics2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "int", "intermediate_size", "nn", "output_size", "return", "self", "str", "super", "up_proj", "x"], "idefics2/modeling_idefics2.py:Idefics2MultiheadAttentionPoolingHead": ["LayerNorm", "ModelMLP", "ModelMultiheadAttentionPoolingHead", "ModelVisionConfig", "Module", "MultiheadAttention", "Parameter", "True", "__init__", "attention", "batch_first", "batch_size", "class", "config", "def", "eps", "forward", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "layer_norm_eps", "layernorm", "mlp", "nn", "num_attention_heads", "output_size", "probe", "randn", "repeat", "residual", "return", "self", "shape", "super", "torch"], "idefics2/modeling_idefics2.py:Idefics2EncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelEncoderLayer", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "idefics2/modeling_idefics2.py:Idefics2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "idefics2/modeling_idefics2.py:Idefics2PreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelMultiheadAttentionPoolingHead", "ModelPerceiverLayer", "ModelPerceiverResampler", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionAttention", "MultiheadAttention", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_reset_parameters", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "get_text_config", "getattr", "if", "initializer_range", "is", "isinstance", "latents", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "probe", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "idefics2/modeling_idefics2.py:Idefics2VisionTransformer": ["BaseModelOutput", "BoolTensor", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_attn_implementation", "_can_record_outputs", "_prepare_4d_attention_mask", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "any", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "check_model_inputs", "class", "config", "def", "device", "dtype", "elif", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "flash_attention_2", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "ones", "patch_attention_mask", "patch_size", "pixel_values", "post_layernorm", "r", "return", "self", "set_input_embeddings", "size", "super", "to", "torch", "tuple", "value", "view"], "idefics2/modeling_idefics2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "idefics2/modeling_idefics2.py:Idefics2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "idefics2/modeling_idefics2.py:Idefics2PerceiverAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelPerceiverAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_attn_implementation", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bsz", "class", "concat", "config", "context", "def", "deprecate_kwarg", "dim", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "keys", "kv_seq_len", "kwargs", "latents", "layer_idx", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "queries", "resampler_head_dim", "resampler_n_heads", "reshape", "return", "scaling", "self", "size", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "values", "version", "view"], "idefics2/modeling_idefics2.py:Idefics2PerceiverLayer": ["Cache", "FloatTensor", "LongTensor", "ModelMLP", "ModelPerceiverAttention", "ModelPerceiverLayer", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "class", "config", "context", "def", "deprecate_kwarg", "depth", "eps", "forward", "hidden_act", "hidden_size", "input_context_norm", "input_latents_norm", "int", "intermediate_size", "kwargs", "latents", "layer_idx", "mlp", "n_latents", "new_name", "nn", "output_size", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "resampler_depth", "resampler_n_latents", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "version"], "idefics2/modeling_idefics2.py:Idefics2PerceiverResampler": ["ModelPerceiverConfig", "ModelPerceiverLayer", "ModelPerceiverResampler", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Parameter", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_supports_flash_attention_2", "_supports_flex_attn", "_supports_sdpa", "attention_mask", "auto_docstring", "cat", "class", "compressed_context", "config", "context", "def", "depth", "device", "dim", "dtype", "else", "eps", "expand", "flash_attention_2", "for", "forward", "hidden_act", "hidden_size", "idx", "if", "in", "kwargs", "latent_attention_mask", "latents", "layers", "n_latents", "nn", "norm", "ones", "perceiver_layer", "position_ids", "r", "range", "resampler_depth", "resampler_n_latents", "return", "rms_norm_eps", "self", "shape", "size", "super", "tgt_len", "torch", "unsqueeze"], "idefics2/modeling_idefics2.py:Idefics2Connector": ["ModelConnector", "ModelMLP", "ModelPerceiverResampler", "Module", "__init__", "_from_config", "attention_mask", "class", "config", "context", "def", "forward", "hidden_act", "hidden_size", "image_hidden_states", "intermediate_size", "modality_projection", "nn", "output_size", "perceiver_config", "perceiver_resampler", "return", "self", "super", "text_config", "vision_config"], "idefics2/modeling_idefics2.py:Idefics2Model": ["AutoModel", "BoolTensor", "Cache", "DynamicCache", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelBaseModelOutputWithPast", "ModelConfig", "ModelConnector", "ModelModel", "ModelPreTrainedModel", "ModelVisionTransformer", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_from_config", "_text_require_grads_hook", "_vision_require_grads_hook", "all", "and", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "cache_position", "can_return_tuple", "cannot", "checkpointing", "children", "class", "config", "connector", "contiguous", "custom_intro", "def", "device", "dim", "dimension", "disable_input_require_grads", "dtype", "either", "elif", "else", "enable_input_require_grads", "expand_as", "forward", "from_config", "get_image_features", "get_input_embeddings", "get_lowest_module", "gradient", "gradient_checkpointing", "have", "height", "hidden_states", "if", "image_hidden_states", "image_seq_len", "image_token_id", "incompatible", "input", "input_ids", "inputs_embeds", "inputs_merger", "is", "kwargs", "last_hidden_state", "len", "list", "logger", "long", "make_inputs_require_grads", "masked_scatter", "module", "nb_values_per_image", "not", "num_channels", "num_images", "numel", "ones", "or", "output", "outputs", "pad_token_id", "padding_idx", "past_key_values", "patch_attention_mask", "patch_size", "patches_subgrid", "perceiver_config", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "r", "raise", "real_images_inds", "register_forward_hook", "remove", "requires_grad_", "resampler_n_latents", "return", "return_dict", "same", "self", "seq_length", "set_input_embeddings", "shape", "size", "special_image_mask", "specify", "step", "sum", "super", "tensor", "text_config", "text_model", "the", "time", "to", "torch", "training", "tuple", "unfold", "unsqueeze", "use_cache", "value", "view", "vision_config", "vision_model", "vocab_size", "warning_once", "width", "with"], "idefics2/modeling_idefics2.py:Idefics2ForConditionalGeneration": ["BoolTensor", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_text_require_grads_hook", "_tied_weights_keys", "_vision_require_grads_hook", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "disable_input_require_grads", "else", "enable_input_require_grads", "forward", "get_image_features", "get_input_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_token_id", "input", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "make_inputs_require_grads", "model", "model_inputs", "module", "nn", "not", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_attention_mask", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "register_forward_hook", "remove", "requires_grad_", "return", "return_dict", "self", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "text_model", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_model", "vocab_size", "weight"], "d_fine/modeling_d_fine.py:multi_scale_deformable_attention_v2": ["False", "Model_scale_deformable_attention_v2", "Tensor", "_", "align_corners", "append", "arange", "attention_weights", "batch_size", "bilinear", "clamp", "concat", "contiguous", "def", "default", "device", "dim", "discrete", "elif", "enumerate", "flatten", "for", "functional", "grid_sample", "height", "hidden_dim", "if", "in", "int", "int64", "level_id", "list", "method", "mode", "nn", "num_heads", "num_levels", "num_points", "num_points_list", "num_queries", "output", "padding_mode", "permute", "repeat", "reshape", "return", "sampling_coord", "sampling_coord_x", "sampling_coord_y", "sampling_grid_l_", "sampling_grids", "sampling_idx", "sampling_locations", "sampling_value_l_", "sampling_value_list", "shape", "split", "stack", "sum", "tensor", "to", "torch", "transpose", "unsqueeze", "value", "value_l_", "value_list", "value_spatial_shapes", "view", "width", "zeros"], "d_fine/modeling_d_fine.py:DFineMultiscaleDeformableAttention": ["F", "Last", "Linear", "Make", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "ValueError", "_", "__init__", "align", "and", "attention_mask", "attention_weights", "batch_size", "be", "but", "class", "config", "d_model", "decoder_attention_heads", "decoder_method", "decoder_n_points", "decoder_offset_scale", "def", "dim", "dtype", "elif", "else", "encoder", "encoder_hidden_states", "f", "flip", "float", "float32", "for", "forward", "get", "hidden", "hidden_states", "if", "in", "instead", "is", "is_torchdynamo_compiling", "isinstance", "length", "list", "masked_fill", "ms_deformable_attn_core", "multi_scale_deformable_attention_v2", "must", "n", "n_heads", "n_levels", "n_points", "nn", "not", "num_feature_levels", "num_points_list", "num_points_scale", "num_queries", "of", "offset", "offset_normalizer", "offset_scale", "or", "output", "raise", "range", "reference_points", "register_buffer", "reshape", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "total_points", "tuple", "unsqueeze", "value", "with"], "d_fine/modeling_d_fine.py:DFineGate": ["LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "cat", "chunk", "class", "d_model", "def", "dim", "forward", "gate", "gate1", "gate2", "gate_input", "gates", "hidden_states", "int", "nn", "norm", "return", "second_residual", "self", "sigmoid", "super", "torch"], "d_fine/modeling_d_fine.py:DFineMultiheadAttention": ["Attention", "False", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_reshape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "expand", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "k_proj", "key_states", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "d_fine/modeling_d_fine.py:DFineDecoderLayer": ["ACT2FN", "Any", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelGate", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "attention_dropout", "attention_mask", "bool", "clamp", "class", "config", "cross_attn_weights", "d_model", "decoder_activation_function", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_hidden_states", "eps", "fc1", "fc2", "final_layer_norm", "forward", "functional", "gateway", "hidden_states", "hidden_states_2", "if", "is", "layer_norm_eps", "max", "min", "nn", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training", "tuple"], "d_fine/modeling_d_fine.py:DFinePreTrainedModel": ["BatchNorm2d", "Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelForObjectDetection", "ModelGate", "ModelHybridEncoder", "ModelLQE", "ModelModel", "ModelMultiscaleDeformableAttention", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "abs", "and", "arange", "attention_weights", "base_model_prefix", "bbox_embed", "bias", "class", "class_embed", "concat", "config", "constant_", "cos", "data", "def", "default_dtype", "denoising_class_embed", "dtype", "enc_score_head", "fill_", "flatten", "float", "for", "gate", "get_default_dtype", "grid_init", "hasattr", "if", "in", "init", "initializer_bias_prior_prob", "initializer_range", "int64", "is", "isinstance", "keepdim", "layer", "layers", "learn_initial_query", "log", "main_input_name", "math", "max", "mean", "module", "n", "n_heads", "nn", "no_grad", "normal_", "not", "num_denoising", "num_labels", "num_points_list", "or", "pi", "pixel_values", "prior_prob", "r", "reg_conf", "reg_scale", "reshape", "sampling_offsets", "scaling", "self", "sin", "stack", "std", "sum", "thetas", "tile", "to", "torch", "up", "values", "weight", "weight_embedding", "with", "xavier_uniform_", "zero_"], "d_fine/modeling_d_fine.py:DFineIntegral": ["F", "Model", "ModelConfig", "Module", "Tensor", "_", "__init__", "batch_size", "class", "config", "def", "device", "dim", "forward", "linear", "max_num_bins", "nn", "num_queries", "pred_corners", "project", "reshape", "return", "self", "shape", "softmax", "super", "to", "torch"], "d_fine/modeling_d_fine.py:DFineDecoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "d_fine/modeling_d_fine.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "d_fine/modeling_d_fine.py:weighting_function": ["Model_function", "None", "Tensor", "abs", "cat", "def", "for", "i", "in", "int", "left_values", "max_num_bins", "range", "reg_scale", "return", "right_values", "step", "torch", "up", "upper_bound1", "upper_bound2", "values", "zeros_like"], "d_fine/modeling_d_fine.py:distance2bbox": ["Model", "Tensor", "abs", "bboxes", "bottom_right_x", "bottom_right_y", "corners_to_center_format", "def", "distance", "float", "points", "reg_scale", "return", "stack", "top_left_x", "top_left_y", "torch"], "d_fine/modeling_d_fine.py:DFineDecoder": ["F", "False", "Model", "ModelConfig", "ModelIntegral", "ModelLQE", "ModelLayer", "ModelMLP", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Parameter", "Tensor", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "clamp", "class", "class_embed", "config", "cross_attentions", "d_model", "decoder_attention_heads", "decoder_layer", "decoder_layers", "def", "detach", "dim", "distance2bbox", "dropout", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eval_idx", "for", "forward", "hidden_size", "hidden_states", "i", "if", "in", "initial_reference_points", "inputs_embeds", "integral", "inter_ref_bbox", "intermediate", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "inverse_sigmoid", "is", "last_hidden_state", "layer_scale", "layers", "level_start_index", "lqe_layers", "max", "max_num_bins", "memory_mask", "min", "new_reference_points", "nn", "not", "num_head", "num_layers", "or", "output", "output_attentions", "output_detach", "output_hidden_states", "position_embeddings", "post_init", "pre_bbox_head", "pred_corners", "pred_corners_undetach", "project", "query_pos_embed", "query_pos_head", "r", "range", "ref_points_detach", "ref_points_initial", "ref_points_input", "reference_points", "reg_scale", "requires_grad", "return", "return_dict", "scores", "self", "sigmoid", "spatial_shapes", "spatial_shapes_list", "stack", "super", "tensor", "torch", "training", "tuple", "unsqueeze", "up", "use_return_dict", "v", "weighting_function"], "d_fine/modeling_d_fine.py:DFineModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "denoising_meta_values", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "d_fine/modeling_d_fine.py:DFineFrozenBatchNorm2d": ["Model", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "d_fine/modeling_d_fine.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "d_fine/modeling_d_fine.py:DFineConvEncoder": ["Model", "Module", "None", "Tensor", "__init__", "append", "backbone", "bool", "channels", "class", "config", "def", "feature_map", "feature_maps", "features", "float", "for", "forward", "freeze_backbone_batch_norms", "functional", "if", "in", "intermediate_channel_sizes", "interpolate", "load_backbone", "mask", "model", "nn", "no_grad", "out", "pixel_mask", "pixel_values", "replace_batch_norm", "return", "self", "shape", "size", "super", "to", "torch", "with"], "d_fine/modeling_d_fine.py:get_contrastive_denoising_training_group": ["Model_contrastive_denoising_training_group", "None", "attn_mask", "batch_size", "bool", "box_noise_scale", "boxes", "center_to_corners_format", "class_embed", "class_labels", "clip_", "corners_to_center_format", "def", "denoise_positive_idx", "denoising_meta_values", "device", "diff", "dn_num_group", "dn_num_split", "dn_positive_idx", "dtype", "else", "float", "for", "full", "i", "idx_block_end", "idx_block_start", "if", "in", "inf", "input_query_bbox", "input_query_class", "int32", "inverse_sigmoid", "known_bbox", "label_noise_ratio", "len", "mask", "max", "max_gt_num", "min", "n", "negative_gt_mask", "new_label", "nonzero", "num_classes", "num_denoising_queries", "num_ground_truths", "num_groups_denoising_queries", "num_gt", "num_queries", "pad_gt_mask", "positive_gt_mask", "rand_like", "rand_part", "rand_sign", "randint_like", "range", "return", "split", "squeeze", "t", "tarModel_size", "tarModels", "tile", "torch", "torch_int", "where", "zeros"], "d_fine/modeling_d_fine.py:DFineModel": ["BaseModelOutput", "BatchNorm2d", "Conv2d", "Embedding", "F", "False", "FloatTensor", "Identity", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelConvEncoder", "ModelDecoder", "ModelHybridEncoder", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Sequential", "True", "Union", "_", "__init__", "_len_sources", "all", "anchor_image_size", "anchors", "and", "append", "arange", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_norm_eps", "batch_size", "batchnorm", "bias", "bool", "box_noise_scale", "cat", "class", "class_embed", "compile_compatible_method_lru_cache", "concat", "config", "conv", "cpu", "cross_attentions", "cumsum", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_in_channels", "decoder_input_proj", "decoder_input_proj_list", "decoder_inputs_embeds", "decoder_outputs", "def", "denoising_bbox_unact", "denoising_class", "denoising_class_embed", "denoising_meta_values", "detach", "device", "dict", "dim", "dn_outputs", "dtype", "elif", "else", "empty", "enc_bbox_head", "enc_output", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "enc_score_head", "enc_topk_bboxes", "enc_topk_logits", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_dim", "encoder_hidden_states", "encoder_input_proj", "encoder_input_proj_list", "encoder_last_hidden_state", "encoder_outputs", "end", "enumerate", "eps", "feat_strides", "features", "finfo", "flatten", "float32", "for", "forward", "freeze_backbone", "gather", "generate_anchors", "get_contrastive_denoising_training_group", "get_encoder", "grid_size", "grid_x", "grid_xy", "grid_y", "height", "hidden_size", "hidden_states", "i", "if", "ij", "in", "in_channels", "index", "indexing", "init_reference_points", "initial_reference_points", "inputs_embeds", "int", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "is", "isinstance", "keepdim", "kernel_size", "label_noise_ratio", "labels", "last_hidden_state", "layer_norm_eps", "learn_initial_query", "len", "level", "level_start_index", "list", "log", "long", "mask", "max", "maxsize", "memory", "meshgrid", "new_zeros", "nn", "not", "num_backbone_outs", "num_channels", "num_classes", "num_denoising", "num_denoising_queries", "num_feature_levels", "num_labels", "num_layers", "num_queries", "ones", "ones_like", "or", "output_attentions", "output_hidden_states", "output_memory", "padding", "padding_idx", "param", "parameters", "pixel_mask", "pixel_values", "post_init", "prod", "proj_feats", "r", "range", "reference_points", "reference_points_unact", "repeat", "requires_grad_", "reshape", "return", "return_dict", "s", "self", "shape", "sigmoid", "source", "source_flatten", "sources", "spatial_shapes", "spatial_shapes_list", "spatial_shapes_tuple", "stack", "stride", "super", "target", "targets", "tensor", "tile", "to", "topk", "topk_ind", "torch", "training", "transpose", "tuple", "tuple_outputs", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_mask", "value", "values", "weight_embedding", "wh", "where", "width"], "d_fine/modeling_d_fine.py:DFineObjectDetectionOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "denoising_meta_values", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "d_fine/modeling_d_fine.py:DFineForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelMLP", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Union", "_", "__init__", "_no_split_modules", "_set_aux_loss", "_tied_weights_keys", "a", "auto_docstring", "auxiliary_outputs", "b", "bbox_embed", "bool", "class", "class_embed", "config", "cross_attentions", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_layers", "def", "denoising_meta_values", "device", "dict", "else", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eval_idx", "for", "forward", "hidden_size", "if", "in", "init_reference_points", "initial_reference_points", "inputs_embeds", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "is", "jit", "kwargs", "labels", "last_hidden_state", "layer_scale", "list", "logits", "loss", "loss_dict", "loss_function", "max_num_bins", "model", "nn", "not", "num_labels", "num_pred", "output", "output_attentions", "output_hidden_states", "outputs", "outputs_class", "outputs_coord", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "predicted_corners", "r", "range", "return", "return_dict", "round", "scaled_dim", "self", "super", "torch", "training", "tuple", "unused", "use_return_dict", "zip"], "d_fine/modeling_d_fine.py:DFineMLPPredictionHead": ["Linear", "Model", "Module", "ModuleList", "__init__", "class", "config", "d_model", "def", "else", "enumerate", "for", "forward", "functional", "h", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "d_fine/modeling_d_fine.py:DFineMLP": ["ACT2CLS", "Linear", "Model", "Module", "ModuleList", "Tensor", "__init__", "act", "class", "def", "else", "enumerate", "for", "forward", "hidden_dim", "hidden_dims", "i", "if", "in", "in_dim", "input_dim", "input_dims", "int", "layer", "layers", "nn", "num_layers", "out_dim", "output_dim", "output_dims", "relu", "return", "self", "stat_features", "str", "super", "torch", "zip"], "d_fine/modeling_d_fine.py:DFineLQE": ["F", "Model", "ModelConfig", "ModelMLP", "Module", "Tensor", "True", "_", "__init__", "batch_size", "cat", "class", "config", "def", "dim", "forward", "keepdim", "length", "lqe_hidden_dim", "lqe_layers", "max_num_bins", "mean", "nn", "pred_corners", "prob", "prob_topk", "quality_score", "reg_conf", "reshape", "return", "scores", "self", "size", "softmax", "stat", "super", "top_prob_values", "topk", "torch"], "d_fine/modeling_d_fine.py:DFineConvNormLayer": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "Identity", "Model", "ModelConfig", "Module", "None", "Optional", "__init__", "activation", "batch_norm_eps", "bias", "class", "config", "conv", "def", "else", "forward", "groups", "hidden_state", "if", "in_channels", "int", "is", "kernel_size", "nn", "norm", "out_channels", "padding", "return", "self", "str", "stride", "super"], "d_fine/modeling_d_fine.py:DFineRepVggBlock": ["ACT2CLS", "Identity", "Model", "ModelConfig", "ModelConvNormLayer", "Module", "None", "__init__", "activation", "activation_function", "class", "config", "conv1", "conv2", "def", "else", "forward", "hidden_channels", "if", "in_channels", "int", "is", "nn", "out_channels", "padding", "return", "self", "super", "x", "y"], "d_fine/modeling_d_fine.py:DFineCSPRepLayer": ["Identity", "Model", "ModelConfig", "ModelConvNormLayer", "ModelRepVggBlock", "Module", "ModuleList", "Tensor", "_", "__init__", "activation", "activation_function", "bottleneck", "bottlenecks", "class", "config", "conv1", "conv2", "conv3", "def", "else", "expansion", "float", "for", "forward", "hidden_channels", "hidden_state", "hidden_state_1", "hidden_state_2", "hidden_state_3", "if", "in", "in_channels", "int", "nn", "num_blocks", "out_channels", "range", "return", "self", "super", "torch"], "d_fine/modeling_d_fine.py:DFineRepNCSPELAN4": ["Model", "ModelCSPRepLayer", "ModelConfig", "ModelConvNormLayer", "Module", "Tensor", "__init__", "act", "activation", "branch1", "branch2", "cat", "class", "config", "conv1", "conv1_dim", "conv2", "conv2_dim", "conv3", "conv3_dim", "conv4", "conv4_dim", "conv_dim", "csp_rep1", "csp_rep2", "def", "encoder_hidden_dim", "extend", "forward", "hidden_expansion", "input_features", "int", "list", "merged_features", "nn", "num_blocks", "numb_blocks", "return", "round", "self", "silu", "split", "split_features", "str", "super", "torch"], "d_fine/modeling_d_fine.py:DFineSCDown": ["Model", "ModelConfig", "ModelConvNormLayer", "Module", "Tensor", "__init__", "class", "config", "conv1", "conv2", "def", "encoder_hidden_dim", "forward", "input_features", "int", "kernel_size", "nn", "return", "self", "stride", "super", "torch"], "d_fine/modeling_d_fine.py:DFineEncoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "def", "dropout", "dtype", "embed_dim", "encoder_activation_function", "encoder_ffn_dim", "encoder_hidden_dim", "eps", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "kwargs", "layer_norm_eps", "max", "min", "nn", "normalize_before", "not", "num_attention_heads", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "d_fine/modeling_d_fine.py:DFineEncoder": ["False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "attention_mask", "bool", "class", "config", "def", "encoder_layers", "for", "forward", "hidden_states", "in", "layer", "layers", "nn", "output_attentions", "pos_embed", "position_embeddings", "range", "return", "self", "src", "src_mask", "super", "torch"], "d_fine/modeling_d_fine.py:DFineHybridEncoder": ["BaseModelOutput", "Embed", "F", "Model", "ModelConfig", "ModelConvNormLayer", "ModelEncoder", "ModelRepNCSPELAN4", "ModelSCDown", "Module", "ModuleList", "None", "ValueError", "_", "__init__", "all_attentions", "append", "arange", "attention_mask", "attentions", "backbone_feature_map", "be", "build_2d_sincos_position_embedding", "by", "class", "concat", "config", "contiguous", "cos", "cpu", "def", "depth_mult", "device", "dim", "dimension", "divisible", "downsample_conv", "downsample_convs", "downsampled_feature_map", "dtype", "else", "embed_dim", "embedding", "enc_ind", "encode_proj_layers", "encoder", "encoder_hidden_dim", "encoder_in_channels", "encoder_layers", "encoder_states", "enumerate", "eval_size", "feat_strides", "flatten", "float32", "for", "forward", "fpn_block", "fpn_blocks", "fpn_feature_map", "fpn_feature_maps", "fpn_layer", "fused_feature_map", "grid_h", "grid_w", "height", "hidden_states", "i", "idx", "if", "ij", "in", "in_channels", "indexing", "inputs_embeds", "interpolate", "is", "last_hidden_state", "lateral_conv", "lateral_convs", "lateral_layer", "layer_outputs", "len", "level_start_index", "meshgrid", "mode", "must", "nearest", "new_fpn_feature_map", "new_pan_feature_map", "nn", "not", "num_blocks", "num_fpn_stages", "numb_blocks", "omega", "or", "out_channels", "out_h", "out_strides", "out_w", "output_attentions", "output_hidden_states", "pan_block", "pan_blocks", "pan_feature_maps", "permute", "pos_dim", "pos_embed", "position", "position_embeddings", "positional_encoding_temperature", "r", "raise", "range", "reshape", "return", "return_dict", "round", "scale_factor", "self", "shape", "sin", "spatial_shapes", "src_flatten", "staticmethod", "super", "temperature", "to", "top_fpn_feature_map", "top_pan_feature_map", "torch", "torch_int", "training", "tuple", "use_return_dict", "v", "valid_ratios", "width", "zip"], "mistral3/modeling_mistral3.py:Mistral3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "mistral3/modeling_mistral3.py:Mistral3PatchMerger": ["False", "Linear", "ModelConfig", "ModelPatchMerger", "Module", "Tensor", "__init__", "append", "bias", "cat", "class", "config", "d", "def", "dim", "enumerate", "for", "forward", "functional", "grid", "h", "hidden_size", "image_features", "image_grid", "image_index", "image_size", "image_sizes", "image_tokens", "in", "kernel_size", "merging_layer", "nn", "patch_size", "permute", "permuted_tensor", "return", "self", "shape", "spatial_merge_size", "split", "stride", "super", "t", "tokens_per_image", "torch", "unfold", "unsqueeze", "view", "vision_config", "w"], "mistral3/modeling_mistral3.py:Mistral3MultiModalProjector": ["ACT2FN", "Linear", "ModelConfig", "ModelMultiModalProjector", "ModelPatchMerger", "ModelRMSNorm", "Module", "Tensor", "__init__", "act", "bias", "class", "config", "def", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "image_features", "image_sizes", "int", "isinstance", "len", "linear_1", "linear_2", "multimodal_projector_bias", "nn", "norm", "num_feature_layers", "patch_merger", "projector_hidden_act", "return", "rms_norm_eps", "self", "super", "text_config", "torch", "vision_config", "vision_feature_layer"], "mistral3/modeling_mistral3.py:Mistral3CausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "mistral3/modeling_mistral3.py:Mistral3ModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "mistral3/modeling_mistral3.py:Mistral3PreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "past_key_values", "supports_gradient_checkpointing"], "mistral3/modeling_mistral3.py:Mistral3Model": ["AutoModel", "Cache", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "device", "dim", "do", "downsample_ratio", "dtype", "else", "exactly", "expand_as", "f", "features", "for", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "height", "hidden_states", "hs_pool", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_sizes", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "k", "kwargs", "language_model", "last_hidden_state", "layer_idx", "list", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "patch_size", "pixel_values", "position_ids", "post_init", "raise", "return", "return_dict", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "specify", "split", "split_sizes", "squeeze", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "value", "vision_config", "vision_feature_layer", "vision_tower", "width"], "mistral3/modeling_mistral3.py:Mistral3ForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "use_return_dict", "value", "vision_feature_layer", "vision_tower", "vocab_size", "weight"], "imagegpt/modeling_imagegpt.py:ImageGPTLayerNorm": ["ModelLayerNorm", "Module", "Parameter", "Tensor", "True", "__init__", "axis", "class", "def", "eps", "float", "forward", "hidden_size", "int", "keepdim", "mean", "nn", "return", "self", "sqrt", "square", "super", "tensor", "torch", "tuple", "weight"], "imagegpt/modeling_imagegpt.py:ImageGPTAttention": ["Cache", "Conv1D", "Dropout", "EncoderDecoderCache", "Error", "False", "If", "ModelAttention", "Module", "None", "Optional", "Please", "RuntimeError", "Softmax", "Tensor", "True", "ValueError", "_", "__init__", "_attn", "_merge_heads", "_split_heads", "_upcast_and_reordered_attn", "alpha", "and", "as", "attention", "attention_mask", "attn_dropout", "attn_head_size", "attn_output", "attn_pdrop", "attn_weights", "autocast", "baddbmm", "be", "beta", "bias", "bool", "bsz", "by", "c_attn", "c_proj", "cache_position", "cat", "causal_mask", "class", "config", "contiguous", "cross", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "defined", "device", "dim", "divisible", "dk", "does", "dtype", "else", "embed_dim", "empty", "enabled", "encoder_attention_mask", "encoder_hidden_states", "f", "find_pruneable_heads_and_indices", "finfo", "float", "float32", "forward", "get", "got", "hasattr", "have", "head_dim", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "index_attn", "instantiate", "int", "is", "is_cross_attention", "is_updated", "isinstance", "k", "k_seq_len", "key", "key_length", "keys", "layer_idx", "layer_past", "layers", "len", "make", "mask_value", "masked_bias", "matmul", "max_position_embeddings", "max_positions", "min", "must", "new_shape", "nn", "not", "num_attention_heads", "num_heads", "ones", "output_attentions", "permute", "persistent", "prune_conv1d_layer", "prune_heads", "pruned_heads", "q", "q_attn", "q_seq_len", "query", "query_length", "raise", "register_buffer", "reorder_and_upcast_attn", "reshape", "resid_dropout", "resid_pdrop", "return", "scale_attn_by_inverse_layer_idx", "scale_attn_weights", "scale_factor", "self", "self_attention_cache", "seq_len", "set", "shape", "size", "split", "split_size", "super", "sure", "tensor", "the", "to", "torch", "torch_float", "transpose", "tril", "tuple", "type", "union", "upcasting", "update", "use_cache", "used", "value", "values", "view", "weights", "where", "with"], "imagegpt/modeling_imagegpt.py:ImageGPTMLP": ["ACT2FN", "Conv1D", "Dropout", "ModelMLP", "Module", "Tensor", "__init__", "act", "activation_function", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "resid_pdrop", "return", "self", "super", "torch"], "imagegpt/modeling_imagegpt.py:ImageGPTBlock": ["Cache", "False", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelBlock", "ModelLayerNorm", "ModelMLP", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "add_cross_attention", "are", "attention", "attention_mask", "attn", "attn_output", "attn_outputs", "be", "bool", "by", "cache_position", "class", "config", "cross", "cross_attn_outputs", "crossattention", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_hidden_states", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "inner_dim", "instantiated", "is", "is_cross_attention", "layer_idx", "layer_norm_epsilon", "layer_past", "layers", "ln_1", "ln_2", "ln_cross_attn", "mlp", "n_inner", "not", "output_attentions", "outputs", "passed", "raise", "residual", "return", "self", "setting", "super", "to", "torch", "tuple", "use_cache", "with"], "imagegpt/modeling_imagegpt.py:ImageGPTPreTrainedModel": ["Conv1D", "Embedding", "Linear", "ModelBlock", "ModelConfig", "ModelLayerNorm", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "__init__", "_init_weights", "_no_split_modules", "and", "base_model_prefix", "bias", "c_proj", "class", "config", "data", "def", "elif", "fill_", "for", "if", "in", "initializer_range", "input_ids", "inputs", "is", "isinstance", "kwargs", "main_input_name", "math", "mean", "module", "n_layer", "name", "named_parameters", "nn", "normal_", "not", "p", "padding_idx", "self", "sqrt", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "imagegpt/modeling_imagegpt.py:ImageGPTModel": ["Any", "BaseModelOutputWithPastAndCrossAttentions", "Cache", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "ModelBlock", "ModelConfig", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "arange", "at", "attention_mask", "attentions", "attn", "auto_docstring", "batch_size", "be", "block", "bool", "both", "cache_position", "cannot", "checkpointing", "class", "config", "cross_attentions", "cuda", "def", "defined", "deprecated", "device", "device_map", "drop", "dtype", "e", "either", "elif", "else", "embd_pdrop", "embed_dim", "encoder_attention_mask", "encoder_batch_size", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "enumerate", "eps", "finfo", "for", "forward", "from_legacy_cache", "g", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "h", "has", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "isinstance", "items", "k", "kwargs", "last_device", "last_hidden_state", "layer", "layer_idx", "layer_norm_epsilon", "ln_f", "logger", "long", "max_position_embeddings", "min", "model_parallel", "n_layer", "new_embeddings", "nn", "not", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "pass", "past_key_values", "past_length", "position_embeds", "position_ids", "post_init", "prune_heads", "r", "raise", "range", "removed", "return", "return_dict", "same", "self", "set_device", "set_input_embeddings", "shape", "should", "size", "specify", "str", "super", "the", "time", "to", "token_type_embeds", "token_type_ids", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "warning_once", "will", "with", "wpe", "wte"], "imagegpt/modeling_imagegpt.py:ImageGPTForCausalImageModeling": ["Any", "Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "GenerationMixin", "Linear", "ModelConfig", "ModelForCausalModelModeling", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "contiguous", "cross_attentions", "def", "device_map", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "model_parallel", "n_embd", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "shift_labels", "shift_logits", "size", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "imagegpt/modeling_imagegpt.py:ImageGPTForImageClassification": ["Any", "Cache", "False", "Linear", "ModelConfig", "ModelForModelClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutputWithPast", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "def", "dim", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_function", "mean", "n_embd", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "pooled_hidden_states", "position_ids", "post_init", "r", "return", "return_dict", "score", "self", "super", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict"], "moshi/modeling_moshi.py:MoshiConditionalGenerationGenerateOutput": ["Cache", "FloatTensor", "LongTensor", "ModelConditionalGenerationGenerateOutput", "ModelOutput", "None", "Optional", "Tensor", "attentions", "audio_codes", "audio_sequences", "beam_indices", "class", "hidden_states", "logits", "past_key_values", "r", "scores", "sequences", "sequences_scores", "torch", "tuple"], "moshi/modeling_moshi.py:MoshiCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "moshi/modeling_moshi.py:MoshiConditionalGenerationOutputWithPast": ["Cache", "FloatTensor", "ModelConditionalGenerationOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "audio_logits", "class", "depth_attentions", "depth_hidden_states", "depth_loss", "depth_past_key_values", "hidden_states", "last_hidden_state", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "moshi/modeling_moshi.py:MoshiUnconditionalInput": ["LongTensor", "ModelOutput", "ModelUnconditionalInput", "Model_audio_codes", "None", "Optional", "Tensor", "attention_mask", "class", "input_ids", "r", "torch", "user_audio_codes"], "moshi/modeling_moshi.py:MoshiRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "ones", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x"], "moshi/modeling_moshi.py:MoshiFlexibleLinear": ["ModelFlexibleLinear", "Module", "None", "Parameter", "__init__", "class", "def", "else", "forward", "if", "index_select", "input_size", "is", "layer_idx", "matmul", "nn", "not", "num_layers", "output_size", "randn", "return", "selected_weights", "self", "squeeze", "super", "torch", "transpose", "weight", "x"], "moshi/modeling_moshi.py:MoshiLinear": ["False", "Linear", "ModelFlexibleLinear", "ModelLinear", "Module", "None", "__init__", "bias", "class", "def", "else", "forward", "if", "input_dim", "layer_idx", "linear", "nn", "not", "num_codebooks", "num_layers", "output_dim", "return", "self", "super", "use_flexible_linear", "x"], "moshi/modeling_moshi.py:MoshiRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "moshi/modeling_moshi.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "moshi/modeling_moshi.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "moshi/modeling_moshi.py:MoshiGatingMLP": ["ACT2FN", "False", "Linear", "ModelFlexibleLinear", "ModelGatingMLP", "Module", "None", "Optional", "Tensor", "_", "__init__", "activation_fn", "batch_size", "bias", "class", "config", "def", "else", "fc1", "fc2", "ffn_dim", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "int", "is", "layer_idx", "nn", "num_codebooks", "num_layers", "return", "self", "sequence_length", "shape", "super", "torch", "use_flexible_linear", "view"], "moshi/modeling_moshi.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "moshi/modeling_moshi.py:MoshiAttention": ["Cache", "False", "Instantiating", "LongTensor", "ModelAttention", "ModelConfig", "ModelLinear", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bool", "bsz", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "else", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "lead", "logger", "make", "math", "matmul", "max_position_embeddings", "must", "new_name", "nn", "not", "num_attention_heads", "num_codebooks", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "repeat_kv", "return", "rope_theta", "rotary_emb", "scaling", "self", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "use_flexible_linear", "use_rope", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "moshi/modeling_moshi.py:MoshiFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "StaticCache", "Tensor", "The", "ValueError", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "an", "and", "apply_rotary_pos_emb", "args", "at", "attention_dropout", "attention_mask", "attn_implementation", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache", "cache_kwargs", "cache_position", "cast", "casted", "class", "com", "compatible", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attention_2", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "github", "hasattr", "have", "head_dim", "hidden", "hidden_states", "https", "huggingface", "if", "implementation", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "isinstance", "issue", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "make", "mean", "might", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "o_proj", "open", "or", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "raise", "related", "reshape", "return", "rotary_emb", "sdpa", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "static", "super", "sure", "target_dtype", "the", "this", "time", "to", "torch", "training", "transformers", "transpose", "tuple", "type", "upcasted", "update", "use", "use_cache", "use_top_left_mask", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "with", "you"], "moshi/modeling_moshi.py:MoshiSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "rotary_emb", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "moshi/modeling_moshi.py:MoshiDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelGatingMLP", "ModelRMSNorm", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "sliding_window", "super", "torch", "tuple", "use_cache", "use_flexible_linear", "use_rope", "version"], "moshi/modeling_moshi.py:MoshiPreTrainedModel": ["Embedding", "Linear", "MimiTransformerLayer", "ModelConfig", "ModelDecoderLayer", "ModelFlexibleLinear", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "input_ids", "is", "isinstance", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "moshi/modeling_moshi.py:MoshiDepthDecoder": ["Attention", "AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "BoolTensor", "Cache", "CausalLMOutputWithPast", "CrossEntropyLoss", "DynamicCache", "Embedding", "False", "Flash", "FloatTensor", "GenerationMixin", "LongTensor", "Make", "Model", "ModelDecoderLayer", "ModelDepthConfig", "ModelDepthDecoder", "ModelFlexibleLinear", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all", "all_hidden_states", "all_self_attns", "and", "append", "arange", "are", "attempting", "attention_mask", "attentions", "audio_vocab_size", "batch_size", "batched", "before", "behaviour", "bitwise_or_", "bool", "cache_position", "call", "cat", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal_attend_mask", "dim", "dtype", "else", "embed_tokens", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "float", "for", "forward", "from_legacy_cache", "full", "generation", "get_max_cache_shape", "get_seq_length", "get_text_config", "getattr", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input", "input_ids", "input_projections", "input_size", "input_tensor", "inputs_embeds", "int", "is", "is_padding_right", "is_sliding", "is_static_sliding_cache", "is_training", "isinstance", "item", "labels", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "lead", "left", "lm_heads", "logger", "logits", "loss", "loss_fct", "make_flex_block_causal_mask", "mask_length", "masked_fill", "may", "min", "min_dtype", "nn", "not", "npu", "num_codebooks", "num_hidden_layers", "of", "or", "output_attentions", "output_hidden_states", "padding_mask", "padding_side", "past_key_values", "past_key_values_length", "past_seen_tokens", "perform", "position_ids", "position_idx", "raise", "range", "reshape", "return", "return_dict", "right", "sdpa", "self", "sequence_length", "shape", "size", "sliding_attend_mask", "sliding_window", "staticmethod", "sum", "super", "sure", "target_length", "text_config", "text_embed_tokens", "the", "this", "to", "tokenizer", "tokenizing", "torch", "training", "tuple", "type", "unexpected", "unsqueeze", "use_cache", "use_flexible_linear", "use_return_dict", "use_rope", "use_sliding_window", "using_static_cache", "v", "version", "vocab_size", "warning_once", "with", "xpu"], "moshi/modeling_moshi.py:MoshiModel": ["Attention", "AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "Flash", "FloatTensor", "LongTensor", "Make", "Model", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "Setting", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all", "all_hidden_states", "all_self_attns", "and", "arange", "are", "attempting", "attention_mask", "attentions", "auto_docstring", "batch_size", "batched", "before", "behaviour", "bitwise_or_", "bool", "cache_position", "call", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal_attend_mask", "dim", "dtype", "else", "embed_tokens", "eps", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "generation", "get_max_cache_shape", "get_seq_length", "get_text_config", "getattr", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_padding_right", "is_sliding", "is_static_sliding_cache", "is_training", "isinstance", "item", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "lead", "left", "list", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "may", "min", "min_dtype", "nn", "norm", "not", "npu", "num_hidden_layers", "of", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "padding_side", "past_key_values", "past_key_values_length", "past_seen_tokens", "perform", "position_ids", "post_init", "raise", "range", "reshape", "return", "return_dict", "right", "rms_norm_eps", "sdpa", "self", "sequence_length", "shape", "size", "sliding_attend_mask", "sliding_window", "staticmethod", "sum", "super", "sure", "target_length", "text_config", "the", "this", "to", "tokenizer", "tokenizing", "torch", "training", "tuple", "type", "unexpected", "unsqueeze", "use_cache", "use_flexible_linear", "use_return_dict", "use_sliding_window", "using_static_cache", "v", "version", "vocab_size", "warning_once", "with", "xpu"], "moshi/modeling_moshi.py:MoshiForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "contiguous", "def", "device", "else", "embed_tokens", "float", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "shift_labels", "shift_logits", "slice", "slice_indices", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "moshi/modeling_moshi.py:MoshiForConditionalGeneration": ["Any", "At", "AutoModel", "BoolTensor", "Cache", "Check", "Embedding", "False", "FloatTensor", "GenerationConfig", "GenerationMixin", "LongTensor", "Make", "Model", "ModelConditionalGenerationGenerateOutput", "ModelConditionalGenerationOutputWithPast", "ModelConfig", "ModelDepthDecoder", "ModelForCausalLM", "ModelForConditionalGeneration", "ModelOutput", "ModelPreTrainedModel", "ModelUnconditionalInput", "Model_audio_codes", "Model_delay_pattern_mask", "Model_input", "Model_input_values", "Model_seq_length", "ModuleList", "No", "None", "Optional", "Seq2SeqLMOutput", "StaticCache", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_check_and_maybe_initialize_inputs", "_from_config", "_prepare_4d_causal_attention_mask_with_cache_position", "_prepare_attention_mask_for_generation", "_prepare_generated_length", "_prepare_generation_config", "_prepare_inputs_embeds_for_generation", "_requires_grad", "_supports_flash_attn", "_supports_sdpa", "_tied_weights_keys", "_update_model_kwargs_for_generation", "all", "alongside", "and", "any", "apply_delay_pattern_mask", "argument", "as", "at", "attention_mask", "attention_mask_from_padding", "attentions", "audio", "audio_codes", "audio_encoder", "audio_encoder_", "audio_encoder_config", "audio_inputs_embeds", "audio_labels", "audio_logits", "audio_sequences", "audio_values", "audio_vocab_size", "auto_docstring", "batch_size", "beam_indices", "been", "blank_input_values", "blank_user_audio_codes", "bool", "bos_token_id", "bsz", "build_delay_pattern_mask", "cache_implementation", "cache_position", "can_infer_attention_mask", "cat", "ceil", "class", "codebook", "concat_unconditional_inputs", "config", "cpu", "d", "decode", "decoder", "decoder_", "decoder_last_hidden_state", "decoder_outputs", "decoder_pad_token_mask", "def", "default_attention_mask", "depth_attentions", "depth_decoder", "depth_decoder_", "depth_decoder_config", "depth_decoder_generation_config", "depth_decoder_kwargs_keys", "depth_decoder_outputs", "depth_hidden_states", "depth_input_ids", "depth_loss", "depth_past_key_values", "device", "dict", "dim", "disk", "docstrings", "doesn", "dtype", "either", "elif", "else", "embed_tokens", "encode", "eos_token_id", "expand", "expanded_beam_indices", "final_loss", "for", "forward", "frame_rate", "freeze_audio_encoder", "freeze_depth_decoder", "from_config", "gather", "generate", "generated_audio_codes", "generation_config", "get", "get_audio_encoder", "get_depth_decoder", "get_input_embeddings", "get_max_cache_shape", "get_output_embeddings", "get_unconditional_inputs", "has_default_max_length", "has_default_min_length", "hasattr", "have", "hf_device_map", "hidden_size", "hidden_states", "if", "in", "index", "index_select", "information", "input_ids", "input_ids_length", "input_ids_shifted", "input_values", "inputs", "inputs_embeds", "inputs_tensor", "int", "int64", "is", "is_encoder_decoder", "is_pad_token_in_inputs", "is_pad_token_not_equal_to_eos_token_id", "isin", "isinstance", "items", "key", "kwargs", "kwargs_audio_encoder", "kwargs_decoder", "kwargs_depth_decoder", "labels", "last_generated_audio_codes", "last_hidden_state", "least", "len", "length", "lm_head", "logits", "logits_to_keep", "long", "loss", "main_device", "main_input_name", "mask", "math", "max", "max_length", "min", "min_length", "model", "model_input_name", "model_inputs", "model_kwargs", "more", "must", "ndim", "ne", "new_embeddings", "nn", "no_grad", "not", "num_beams", "num_codebooks", "num_new_tokens", "num_quantizers", "num_return_sequences", "num_samples", "of", "one", "one_input_has_been_passed", "ones", "or", "other", "others", "output_attentions", "output_audio_codes", "output_hidden_states", "output_scores", "output_text_ids", "output_values", "outputs", "pad_token_id", "param", "parameters", "passed", "past_key_values", "pattern_mask", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "provide", "r", "raise", "range", "ratio", "repeat_interleave", "requires_grad", "reshape", "return", "return_audio_codes", "return_audio_waveforms", "return_dict", "return_dict_in_generate", "same", "sampling_rate", "self", "seq_len", "seq_len_to_keep", "sequence", "sequence_length", "sequences", "set_input_embeddings", "set_output_embeddings", "shape", "startswith", "static", "staticmethod", "str", "sum", "super", "supports_gradient_checkpointing", "sure", "t", "target_length", "text_labels", "that", "the", "they", "to", "to_tuple", "tokens_seq_length", "torch", "transpose", "tuple", "unconditional_inputs", "unsqueeze", "update", "use", "use_cache", "use_return_dict", "user", "user_audio_codes", "user_delay_pattern_mask", "user_input", "user_input_values", "user_seq_length", "value", "values", "view", "vocab_size", "weight", "where", "zeros"], "shieldgemma2/modeling_shieldgemma2.py:ShieldGemma2ImageClassifierOutputWithNoAttention": ["ImageClassifierOutputWithNoAttention", "ModelImageClassifierOutputWithNoAttention", "None", "Optional", "Tensor", "class", "probabilities", "torch"], "shieldgemma2/modeling_shieldgemma2.py:ShieldGemma2ForImageClassification": ["AutoModelForImageTextToText", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelForImageClassification", "ModelImageClassifierOutputWithNoAttention", "None", "Optional", "PreTrainedModel", "Tensor", "Union", "__init__", "_checkpoint_conversion_mapping", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "config", "decoder", "def", "dim", "forward", "from_config", "get_decoder", "get_input_embeddings", "get_output_embeddings", "getattr", "input_ids", "inputs_embeds", "int", "labels", "language_model", "lm_head", "lm_kwargs", "logits", "logits_to_keep", "model", "multi_modal_projector", "new_embeddings", "no_token_index", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "probabilities", "r", "return", "return_dict", "selected_logits", "self", "set_decoder", "set_input_embeddings", "set_output_embeddings", "softmax", "super", "tie_weights", "token_type_ids", "torch", "use_cache", "value", "vision_tower", "yes_token_index"], "vision_text_dual_encoder/modeling_vision_text_dual_encoder.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "vision_text_dual_encoder/modeling_vision_text_dual_encoder.py:clip_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "vision_text_dual_encoder/modeling_vision_text_dual_encoder.py:VisionTextDualEncoderModel": ["AutoConfig", "AutoModel", "BaseModelOutputWithPooling", "CLIPModelConfig", "CLIPModelModel", "CLIPOutput", "Either", "False", "FloatTensor", "If", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "Model_", "Model_config", "Model_embed_dim", "Model_model", "Model_model_name_or_path", "Model_model_output", "Model_outputs", "None", "Optional", "Parameter", "PreTrainedModel", "T", "TRAIN", "Tensor", "The", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_supports_flash_attn", "_supports_sdpa", "a", "able", "an", "and", "are", "argument", "as", "attention_mask", "auto_docstring", "base_model_prefix", "be", "bias", "bool", "class", "classmethod", "clip", "clip_loss", "cls", "config", "config_class", "configuration", "def", "defined", "del", "dim", "down", "else", "exp", "f", "filter_out_non_signature_kwargs", "for", "forward", "from_Model_text_configs", "from_Model_text_pretrained", "from_config", "from_pretrained", "get_image_features", "get_text_features", "has", "hidden_size", "if", "image_embeds", "image_features", "in", "inference", "initialized", "input_ids", "is", "isinstance", "it", "items", "keepdim", "key", "kwargs", "kwargs_Model", "kwargs_text", "layer", "len", "logger", "logit", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "model", "model_args", "model_type", "newly", "nn", "norm", "not", "of", "on", "or", "output", "output_attentions", "output_hidden_states", "pixel_values", "pooler_output", "pop", "position_ids", "predictions", "probably", "projection", "projection_dim", "provided", "r", "raise", "return", "return_dict", "return_loss", "scale", "self", "should", "startswith", "str", "stream", "super", "t", "task", "tensor", "text", "text_", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_name_or_path", "text_model_output", "text_outputs", "text_projection", "this", "to", "token_type_ids", "torch", "tuple", "type", "use", "value", "visual_projection", "warning", "weight", "weights"], "distilbert/modeling_distilbert.py:create_sinusoidal_embeddings": ["GatheredParameters", "Model_sinusoidal_embeddings", "Tensor", "_Model_sinusoidal_embeddings", "deepspeed", "def", "dim", "distributed", "else", "get_rank", "if", "int", "is_deepspeed_zero3_enabled", "modifier_rank", "n_pos", "out", "torch", "with", "zero"], "distilbert/modeling_distilbert.py:_create_sinusoidal_embeddings": ["False", "FloatTensor", "Tensor", "_create_sinusoidal_embeddings", "array", "cos", "def", "detach_", "dim", "for", "in", "int", "j", "n_pos", "np", "out", "pos", "position_enc", "power", "range", "requires_grad", "sin", "torch"], "distilbert/modeling_distilbert.py:Embeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Model", "Module", "None", "Optional", "PretrainedConfig", "Tensor", "__init__", "arange", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "eps", "expand", "expand_as", "forward", "hasattr", "if", "input_embeds", "input_ids", "is", "long", "max_position_Model", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_Model", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "torch", "unsqueeze", "vocab_size", "word_Model"], "distilbert/modeling_distilbert.py:MultiHeadSelfAttention": ["Dropout", "False", "Linear", "ModelHeadSelfAttention", "Module", "None", "Optional", "PretrainedConfig", "Tensor", "ValueError", "__init__", "attention_dropout", "attention_head_size", "bool", "bs", "class", "config", "context", "contiguous", "def", "dim", "dim_per_head", "divide", "dropout", "dtype", "else", "evenly", "expand_as", "f", "find_pruneable_heads_and_indices", "finfo", "forward", "functional", "head_mask", "heads", "if", "in_features", "index", "int", "is", "is_causal", "k", "k_length", "k_lin", "key", "len", "list", "mask", "mask_reshp", "masked_fill", "math", "matmul", "min", "must", "n_heads", "nn", "not", "out_features", "out_lin", "output_attentions", "p", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "q_length", "q_lin", "query", "raise", "return", "scores", "self", "set", "shape", "size", "softmax", "sqrt", "super", "tensor", "torch", "transpose", "tuple", "union", "unshape", "v", "v_lin", "value", "view", "weights", "x"], "distilbert/modeling_distilbert.py:DistilBertFlashAttention2": ["False", "ModelFlashAttention2", "MultiHeadSelfAttention", "None", "Optional", "Tensor", "The", "We", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "args", "attention_dropout", "attn_dropout", "attn_output", "attn_weights", "attn_weights_reshaped", "back", "batch_size", "be", "bool", "cast", "casted", "class", "config", "cpu", "def", "device", "device_type", "dim", "dim_per_head", "dropout", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "have", "head_mask", "hidden", "if", "in", "input", "is_autocast_enabled", "is_causal", "k_lin", "key", "key_states", "kwargs", "layer", "layers", "logger", "mask", "might", "mps", "n_heads", "norm", "or", "out_lin", "output_attentions", "q_length", "q_lin", "query", "query_states", "related", "reshape", "return", "seems", "self", "silently", "size", "states", "super", "target_dtype", "the", "this", "to", "torch", "training", "tuple", "type", "upcasted", "use_top_left_mask", "v_lin", "value", "value_states", "view", "warning_once", "weight", "will", "x", "you"], "distilbert/modeling_distilbert.py:DistilBertSdpaAttention": ["Falling", "False", "ModelSdpaAttention", "MultiHeadSelfAttention", "None", "Optional", "PretrainedConfig", "Tensor", "This", "Transformers", "True", "_", "__init__", "argument", "attention", "attention_dropout", "attn_implementation", "attn_mask", "attn_output", "back", "batch_size", "be", "bool", "but", "can", "class", "config", "contiguous", "def", "dim", "dim_per_head", "does", "dropout_p", "dropout_prob", "eager", "else", "forward", "from", "functional", "head_mask", "if", "implementation", "is", "is_causal", "k", "k_lin", "key", "loading", "logger", "manual", "mask", "model", "n_heads", "nn", "not", "onwards", "or", "out_lin", "output_attentions", "q", "q_lin", "query", "removed", "required", "return", "scaled_dot_product_attention", "self", "shape", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "unshape", "used", "using", "v", "v5", "v_lin", "value", "version", "view", "warning", "warning_once", "when", "will", "x"], "distilbert/modeling_distilbert.py:FFN": ["Dropout", "Linear", "Model", "Module", "PretrainedConfig", "Tensor", "__init__", "activation", "apply_chunking_to_forward", "chunk_size_feed_forward", "class", "config", "def", "dim", "dropout", "ff_chunk", "forward", "get_activation", "hidden_dim", "in_features", "input", "lin1", "lin2", "nn", "out_features", "p", "return", "self", "seq_len_dim", "super", "torch", "x"], "distilbert/modeling_distilbert.py:TransformerBlock": ["FFN", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelBlock", "Model_ATTENTION_CLASSES", "None", "Optional", "PretrainedConfig", "Tensor", "TypeError", "ValueError", "__init__", "_attn_implementation", "a", "attention", "attn_mask", "be", "bool", "but", "class", "config", "def", "dim", "divide", "else", "eps", "evenly", "f", "ffn", "ffn_output", "forward", "head_mask", "if", "is", "it", "key", "mask", "must", "n_heads", "nn", "normalized_shape", "not", "output", "output_attentions", "output_layer_norm", "query", "raise", "return", "sa_layer_norm", "sa_output", "sa_weights", "self", "super", "torch", "tuple", "type", "value", "x"], "distilbert/modeling_distilbert.py:Transformer": ["BaseModelOutput", "False", "Model", "ModelBlock", "Module", "ModuleList", "None", "Optional", "PretrainedConfig", "Tensor", "The", "Union", "ValueError", "_", "__init__", "all_attentions", "all_hidden_states", "attentions", "attn_mask", "be", "bool", "but", "class", "config", "def", "else", "enumerate", "f", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_state", "hidden_states", "i", "if", "in", "is", "it", "last_hidden_state", "layer", "layer_module", "layer_outputs", "len", "length", "n_layers", "nn", "not", "of", "output_attentions", "output_hidden_states", "raise", "range", "return", "return_dict", "self", "should", "super", "the", "torch", "tuple", "v", "x"], "distilbert/modeling_distilbert.py:DistilBertPreTrainedModel": ["Embedding", "Embeddings", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_supports_flash_attn", "_supports_sdpa", "and", "base_model_prefix", "bias", "class", "config", "create_sinusoidal_embeddings", "data", "def", "dim", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "max_position_embeddings", "mean", "module", "nn", "normal_", "not", "padding_idx", "position_embeddings", "self", "sinusoidal_pos_embds", "std", "supports_gradient_checkpointing", "weight", "zero_"], "distilbert/modeling_distilbert.py:DistilBertModel": ["BaseModelOutput", "Embedding", "Embeddings", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "PretrainedConfig", "Setting", "Tensor", "Transformer", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "and", "at", "attention", "attention_mask", "attn_mask", "auto_docstring", "bool", "both", "cannot", "class", "clone", "config", "create_sinusoidal_embeddings", "def", "device", "dict", "dim", "dtype", "either", "elif", "else", "embeddings", "f", "flash_attention_2", "for", "forward", "get_head_mask", "get_input_embeddings", "get_position_embeddings", "have", "head_mask", "head_mask_is_none", "heads", "heads_to_prune", "if", "in", "info", "input_ids", "input_shape", "inputs_embeds", "int", "is", "items", "layer", "list", "logger", "max_position_embeddings", "n_pos", "new_embeddings", "new_num_position_embeddings", "nn", "no_grad", "not", "num_hidden_layers", "num_position_embeds_diff", "old_position_embeddings_weight", "ones", "or", "out", "output_attentions", "output_hidden_states", "position_embeddings", "post_init", "prune_heads", "r", "raise", "resize_position_embeddings", "return", "return_dict", "same", "sdpa", "self", "set_input_embeddings", "sinusoidal_pos_embds", "size", "specify", "super", "tgt_len", "the", "time", "to", "torch", "transformer", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "weight", "with", "word_embeddings", "x"], "distilbert/modeling_distilbert.py:DistilBertForMaskedLM": ["CrossEntropyLoss", "Embedding", "LayerNorm", "Linear", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "PretrainedConfig", "Tensor", "Union", "__init__", "_tied_weights_keys", "activation", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "dim", "dlbrt_output", "else", "eps", "forward", "get_activation", "get_output_embeddings", "get_position_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "loss", "mlm_loss", "mlm_loss_fct", "new_embeddings", "new_num_position_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "post_init", "prediction_logits", "r", "resize_position_embeddings", "return", "return_dict", "self", "set_output_embeddings", "size", "super", "torch", "tuple", "use_return_dict", "view", "vocab_layer_norm", "vocab_projector", "vocab_size", "vocab_transform", "weight"], "distilbert/modeling_distilbert.py:DistilBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Embedding", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "Model_output", "None", "Optional", "PretrainedConfig", "ReLU", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dim", "dropout", "dtype", "elif", "else", "forward", "get_position_embeddings", "head_mask", "hidden_state", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "new_num_position_embeddings", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pooled_output", "post_init", "pre_classifier", "problem_type", "r", "regression", "resize_position_embeddings", "return", "return_dict", "self", "seq_classif_dropout", "single_label_classification", "squeeze", "super", "torch", "tuple", "use_return_dict", "view"], "distilbert/modeling_distilbert.py:DistilBertForQuestionAnswering": ["CrossEntropyLoss", "Dropout", "Embedding", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "Model_output", "None", "Optional", "PretrainedConfig", "QuestionAnsweringModelOutput", "Tensor", "Union", "ValueError", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "but", "clamp", "class", "config", "contiguous", "def", "dim", "dropout", "else", "end_logits", "end_loss", "end_positions", "f", "forward", "get_position_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "int", "is", "it", "len", "logits", "loss", "loss_fct", "new_num_position_embeddings", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "post_init", "qa_dropout", "qa_outputs", "r", "raise", "resize_position_embeddings", "return", "return_dict", "self", "should", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_return_dict"], "distilbert/modeling_distilbert.py:DistilBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "Embedding", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PretrainedConfig", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "get_position_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "loss", "loss_fct", "new_num_position_embeddings", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "post_init", "r", "resize_position_embeddings", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict", "view"], "distilbert/modeling_distilbert.py:DistilBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Embedding", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "PretrainedConfig", "ReLU", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dim", "dropout", "else", "forward", "get_position_embeddings", "head_mask", "hidden_state", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "loss", "loss_fct", "new_num_position_embeddings", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "post_init", "pre_classifier", "r", "reshaped_logits", "resize_position_embeddings", "return", "return_dict", "self", "seq_classif_dropout", "shape", "size", "super", "torch", "tuple", "use_return_dict", "view"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderEmbeddings": ["Dropout", "Embedding", "LayerNorm", "LongTensor", "ModelConfig", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "True", "__init__", "bias", "class", "compile", "compiled_embeddings", "config", "def", "drop", "dynamic", "else", "embedding_dropout", "eps", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "nn", "norm", "norm_bias", "norm_eps", "not", "pad_token_id", "padding_idx", "reference_compile", "return", "self", "super", "tok_embeddings", "torch", "vocab_size"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderMLP": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "Wi", "Wo", "__init__", "act", "bias", "chunk", "class", "config", "def", "dim", "drop", "forward", "gate", "hidden_activation", "hidden_size", "hidden_states", "input", "int", "intermediate_size", "mlp_bias", "mlp_dropout", "nn", "return", "self", "super", "torch"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "modernbert_decoder/modeling_modernbert_decoder.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "modernbert_decoder/modeling_modernbert_decoder.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "modernbert_decoder/modeling_modernbert_decoder.py:eager_attention_forward": ["ModelAttention", "Model_attention_forward", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_dim", "if", "int", "is", "key", "kwargs", "matmul", "module", "nn", "p", "query", "return", "scaling", "shape", "sliding_window", "softmax", "to", "torch", "training", "transpose", "tuple", "value"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "Wo", "__init__", "_attn_implementation", "a", "all_head_size", "apply_rotary_pos_emb", "attention", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "f", "forward", "head_dim", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_sliding", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "multiple", "new_name", "nn", "not", "num_attention_heads", "num_heads", "number", "of", "out_drop", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "sin", "size", "sliding_attention", "sliding_window", "super", "the", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "Identity", "LayerNorm", "LongTensor", "ModelAttention", "ModelConfig", "ModelLayer", "ModelMLP", "None", "Optional", "Tensor", "__init__", "attention_mask", "attention_type", "attn", "attn_norm", "attn_outputs", "bias", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "int", "is_sliding", "kwargs", "layer_idx", "layer_types", "mlp", "mlp_norm", "mlp_output", "new_name", "nn", "norm_bias", "norm_eps", "past_key_value", "past_key_values", "position_embeddings", "position_embeddings_global", "position_embeddings_local", "residual", "return", "self", "super", "torch", "tuple", "use_cache", "version"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderPredictionHead": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelPredictionHead", "Module", "Tensor", "__init__", "act", "bias", "class", "classifier_activation", "classifier_bias", "config", "def", "dense", "eps", "forward", "hidden_size", "hidden_states", "nn", "norm", "norm_bias", "norm_eps", "return", "self", "super", "torch"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderPreTrainedModel": ["LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelForCausalLM", "ModelForSequenceClassification", "ModelLayer", "ModelMLP", "ModelPreTrainedModel", "ModelPredictionHead", "Module", "None", "PreTrainedModel", "True", "Wi", "Wo", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "attentions", "b", "base_model_prefix", "bias", "class", "classifier", "config", "cutoff_factor", "data", "decoder", "def", "dense", "elif", "embedding", "fill_", "final_out", "float", "hidden_size", "hidden_states", "if", "in", "init", "init_weight", "initializer_cutoff_factor", "initializer_range", "is", "isinstance", "k_proj", "math", "mean", "model", "module", "nn", "not", "num_hidden_layers", "out", "past_key_values", "q_proj", "self", "sqrt", "std", "stds", "supports_gradient_checkpointing", "tok_embeddings", "trunc_normal_", "v_proj", "weight", "zero_", "zeros_"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "False", "LayerNorm", "LongTensor", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embeddings", "enumerate", "eps", "exactly", "expand", "final_norm", "for", "forward", "full_attention", "get_input_embeddings", "get_seq_length", "global_rotary_emb", "gradient_checkpointing", "hidden_size", "hidden_states", "idx", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layers", "local_rotary_emb", "mask_kwargs", "must", "nn", "norm_bias", "norm_eps", "not", "num_hidden_layers", "of", "one", "or", "past_key_values", "past_seen_tokens", "position_embeddings_global", "position_embeddings_local", "position_ids", "post_init", "raise", "range", "return", "self", "seq_length", "set_input_embeddings", "shape", "sliding_attention", "specify", "super", "tok_embeddings", "torch", "training", "tuple", "unsqueeze", "use_cache", "value", "warn_if_padding_and_no_attention_mask"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderForCausalLM": ["Cache", "CausalLMOutputWithPast", "CrossEntropyLoss", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "can_return_tuple", "class", "config", "contiguous", "decoder", "decoder_bias", "def", "device", "forward", "get_output_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "new_embeddings", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "set_output_embeddings", "shift_labels", "shift_logits", "super", "to", "torch", "tuple", "use_cache", "view", "vocab_size", "weight"], "modernbert_decoder/modeling_modernbert_decoder.py:ModernBertDecoderForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "Dropout", "Linear", "LongTensor", "MSELoss", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "blab", "bool", "can_return_tuple", "checkpoint", "class", "classifier", "classifier_bias", "classifier_dropout", "config", "conjunction", "dec", "def", "defined", "detect", "device", "drop", "dtype", "elif", "else", "f", "forward", "handle", "head", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "jhu", "kwargs", "labels", "last_non_pad_token", "logger", "logits", "long", "loss", "loss_fct", "may", "model", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "self", "sequence_length", "shape", "single_label_classification", "sizes", "squeeze", "super", "test", "to", "token", "token_indices", "tokens", "torch", "transformer_outputs", "tuple", "unexpected", "use_cache", "using", "view", "warning_once", "will", "with"], "deit/modeling_deit.py:DeiTEmbeddings": ["BoolTensor", "Dropout", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_and_dist_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "distillation_token", "distillation_tokens", "dropout", "else", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embedding", "position_embeddings", "reshape", "return", "self", "seq_length", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_mask_token", "view", "width", "zeros"], "deit/modeling_deit.py:DeiTPatchEmbeddings": ["Conv2d", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "flatten", "forward", "height", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "width", "with", "x"], "deit/modeling_deit.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "deit/modeling_deit.py:DeiTSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "deit/modeling_deit.py:DeiTSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "deit/modeling_deit.py:DeiTAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "deit/modeling_deit.py:DeiTIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deit/modeling_deit.py:DeiTOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "deit/modeling_deit.py:DeiTLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "deit/modeling_deit.py:DeiTEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "deit/modeling_deit.py:DeiTPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "distillation_token", "dtype", "elif", "fill_", "float32", "hidden_states", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "deit/modeling_deit.py:DeiTModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "BoolTensor", "False", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dtype", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "expected_dtype", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "interpolate_pos_encoding", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "nn", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "projection", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch", "use_mask_token", "weight"], "deit/modeling_deit.py:DeiTPooler": ["ACT2FN", "Linear", "ModelConfig", "ModelPooler", "Module", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "pooler_act", "pooler_output_size", "return", "self", "super", "torch"], "deit/modeling_deit.py:DeiTForMaskedImageModeling": ["BaseModelOutputWithPooling", "BoolTensor", "Conv2d", "False", "MaskedImageModelingOutput", "Model", "ModelConfig", "ModelForMaskedImageModeling", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PixelShuffle", "Sequential", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "batch_size", "bool", "bool_masked_pos", "can_return_tuple", "class", "config", "contiguous", "decoder", "def", "encoder_stride", "forward", "functional", "head_mask", "height", "hidden_size", "hidden_states", "if", "image_size", "in_channels", "int", "interpolate_pos_encoding", "is", "kernel_size", "kwargs", "l1_loss", "last_hidden_state", "loss", "mask", "masked_im_loss", "nn", "none", "not", "num_channels", "out_channels", "outputs", "patch_size", "permute", "pixel_values", "post_init", "r", "reconstructed_pixel_values", "reconstruction", "reconstruction_loss", "reduction", "repeat_interleave", "reshape", "return", "self", "sequence_length", "sequence_output", "shape", "size", "sum", "super", "torch", "unsqueeze", "use_mask_token", "width"], "deit/modeling_deit.py:DeiTForImageClassification": ["BaseModelOutputWithPooling", "False", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch"], "deit/modeling_deit.py:DeiTForImageClassificationWithTeacherOutput": ["FloatTensor", "ModelForImageClassificationWithTeacherOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cls_logits", "distillation_logits", "hidden_states", "logits", "r", "torch", "tuple"], "deit/modeling_deit.py:DeiTForImageClassificationWithTeacher": ["BaseModelOutputWithPooling", "False", "Identity", "Linear", "Model", "ModelConfig", "ModelForImageClassificationWithTeacher", "ModelForImageClassificationWithTeacherOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "cls_classifier", "cls_logits", "config", "def", "distillation_classifier", "distillation_logits", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "kwargs", "last_hidden_state", "logits", "nn", "num_labels", "outputs", "pixel_values", "post_init", "return", "self", "sequence_output", "super", "torch"], "aria/modeling_aria.py:AriaTextRMSNorm": ["ModelTextRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "vModelnce", "vModelnce_epsilon", "weight"], "aria/modeling_aria.py:AriaProjectorMLP": ["ACT2FN", "False", "Linear", "ModelProjectorMLP", "Module", "__init__", "act", "bias", "class", "def", "forward", "gelu_new", "hidden_features", "hidden_states", "in_features", "linear_in", "linear_out", "nn", "output_dim", "return", "self", "super"], "aria/modeling_aria.py:AriaCrossAttention": ["Dropout", "False", "LayerNorm", "Linear", "ModelConfig", "ModelCrossAttention", "Module", "MultiheadAttention", "None", "True", "_", "__init__", "attn_mask", "attn_output", "batch_first", "bias", "class", "config", "def", "dropout", "dropout_rate", "float", "forward", "hidden_size", "hidden_states", "k_proj", "key", "key_value_states", "layer_norm", "layer_norm_kv", "linear", "multihead_attn", "nn", "num_attention_heads", "num_heads", "q_proj", "query", "return", "self", "super", "v_proj", "value", "vision_config"], "aria/modeling_aria.py:AriaProjector": ["KeyError", "LayerNorm", "ModelConfig", "ModelCrossAttention", "ModelProjector", "ModelProjectorMLP", "Module", "None", "Number", "Optional", "Parameter", "Tensor", "__init__", "amongst", "attention_out", "attn_mask", "batch_size", "class", "config", "cross_attn", "def", "expand", "f", "feed_forward", "forward", "found", "hidden_features", "hidden_size", "if", "in", "in_features", "is", "key_value_states", "keys", "kv_dim", "layer_norm", "max_value_projector_patch_to_query_dict", "nn", "not", "num_attention_heads", "num_heads", "num_patches", "of", "out", "output_dim", "patch_to_query_dict", "patches", "possible", "projector_patch_to_query_dict", "queries", "query", "query_num", "raise", "repeat", "repeat_interleave", "return", "self", "shape", "size", "super", "text_config", "torch", "unsqueeze", "values", "vision_config", "zeros"], "aria/modeling_aria.py:AriaSharedExpertsMLP": ["ACT2FN", "Linear", "ModelSharedExpertsMLP", "ModelTextConfig", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "moe_num_shared_experts", "nn", "return", "self", "super", "up_proj", "x"], "aria/modeling_aria.py:sequential_experts_gemm": ["Model_experts_gemm", "cat", "cumsum", "cumsum_num_tokens", "def", "device", "dim", "dtype", "end", "expert_num", "expert_weights", "for", "in", "long", "matmul", "num_tokens", "out", "out_features", "output", "range", "return", "shape", "start", "token_states", "tokens", "tokens_per_expert", "torch", "zero_tensor", "zeros"], "aria/modeling_aria.py:AriaGroupedExpertsGemm": ["ModelGroupedExpertsGemm", "Module", "Parameter", "__init__", "class", "cpu", "def", "empty", "forward", "groups", "in_features", "input", "nn", "out_features", "return", "self", "sequential_experts_gemm", "super", "tokens_per_expert", "torch", "weight"], "aria/modeling_aria.py:AriaGroupedExpertsMLP": ["ModelGroupedExpertsGemm", "ModelGroupedExpertsMLP", "ModelTextConfig", "Module", "None", "__init__", "chunk", "class", "config", "def", "dim", "fc1", "fc1_output", "fc2", "fc2_output", "forward", "functional", "gate", "hidden_size", "intermediate_size", "moe_num_experts", "nn", "permuted_tokens", "projection", "return", "self", "silu", "super", "tokens_per_expert", "torch"], "aria/modeling_aria.py:AriaTextMoELayer": ["False", "Linear", "ModelGroupedExpertsMLP", "ModelSharedExpertsMLP", "ModelTextConfig", "ModelTextMoELayer", "Module", "Tensor", "__init__", "argsort", "bias", "bins", "class", "config", "def", "device", "dim", "dtype", "expert_output", "experts", "flatten", "flatten_indices", "float32", "forward", "functional", "hidden_size", "hidden_states", "histc", "index_copy_", "index_select", "indices", "k", "logits", "max", "min", "moe_num_experts", "moe_topk", "nn", "original_dtype", "original_shape", "output", "permuted_tokens", "return", "router", "scores", "self", "shape", "shared_expert_output", "shared_experts", "size", "softmax", "sorted_indices", "sum", "super", "to", "tokens_per_expert", "top_indices", "top_logits", "topk", "torch", "unpermuted_tokens", "unsqueeze", "view", "zeros"], "aria/modeling_aria.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "aria/modeling_aria.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "aria/modeling_aria.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "aria/modeling_aria.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "aria/modeling_aria.py:AriaTextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelTextAttention", "ModelTextConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "aria/modeling_aria.py:AriaTextDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextMoELayer", "ModelTextRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "aria/modeling_aria.py:AriaTextPreTrainedModel": ["ModelGroupedExpertsGemm", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextPreTrainedModel", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "aria/modeling_aria.py:AriaPreTrainedModel": ["False", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelProjector", "ModelTextAttention", "ModelTextDecoderLayer", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "def", "hidden_states", "if", "init", "initializer_range", "isinstance", "module", "nn", "past_key_values", "query", "self", "std", "super", "supports_gradient_checkpointing", "trunc_normal_"], "aria/modeling_aria.py:AriaTextRotaryEmbedding": ["False", "ModelTextConfig", "ModelTextRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "aria/modeling_aria.py:AriaTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextDecoderLayer", "ModelTextModel", "ModelTextPreTrainedModel", "ModelTextRMSNorm", "ModelTextRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "aria/modeling_aria.py:AriaTextForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelTextConfig", "ModelTextForCausalLM", "ModelTextModel", "ModelTextPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "aria/modeling_aria.py:AriaCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "aria/modeling_aria.py:AriaModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "aria/modeling_aria.py:AriaModel": ["AutoModel", "Cache", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelProjector", "None", "Optional", "Tensor", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_create_patch_attention_mask", "all", "and", "attention_mask", "attentions", "attn_mask", "auto_docstring", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "device", "dim", "dimension", "do", "dtype", "else", "expand_as", "f", "features", "flatten", "flattened_mask", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "if", "image", "image_attn_mask", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "input_ids", "inputs_embeds", "int", "is", "kwargs", "language_model", "last_hidden_state", "logical_not", "long", "masked_scatter", "match", "model", "multi_modal_projector", "n_image_features", "n_image_tokens", "not", "numel", "output_hidden_states", "outputs", "past_key_values", "patch_attention_mask", "patch_size", "patches_subgrid", "pixel_mask", "pixel_values", "position_ids", "post_init", "raise", "return", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "size", "special_image_mask", "step", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unfold", "unsqueeze", "use_cache", "value", "vision_config", "vision_feature_layer", "vision_tower"], "aria/modeling_aria.py:AriaForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_mask", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "value", "vision_feature_layer", "vision_tower", "vocab_size", "weight"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1RMSNorm": ["ModelYuanDenseV1RMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1MLP": ["ACT2FN", "False", "Linear", "ModelYuanDenseV1Config", "ModelYuanDenseV1MLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "is_shared_mlp", "layer_idx", "nn", "return", "self", "super", "up_proj", "x"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelYuanDenseV1Attention", "ModelYuanDenseV1Config", "ModelYuanDenseV1RMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_layernorm", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_layernorm", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelYuanDenseV1Attention", "ModelYuanDenseV1Config", "ModelYuanDenseV1DecoderLayer", "ModelYuanDenseV1MLP", "ModelYuanDenseV1RMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1PreTrainedModel": ["Embedding", "Linear", "ModelYuanDenseV1Attention", "ModelYuanDenseV1Config", "ModelYuanDenseV1DecoderLayer", "ModelYuanDenseV1PreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1RotaryEmbedding": ["False", "ModelYuanDenseV1Config", "ModelYuanDenseV1RotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "alpha", "and", "arange", "attention_scaling", "autocast", "base", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "head_dim", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_theta", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelYuanDenseV1Config", "ModelYuanDenseV1DecoderLayer", "ModelYuanDenseV1Model", "ModelYuanDenseV1PreTrainedModel", "ModelYuanDenseV1RMSNorm", "ModelYuanDenseV1RotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelYuanDenseV1ForCausalLM", "ModelYuanDenseV1Model", "ModelYuanDenseV1PreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "hunyuan_v1_dense/modeling_hunyuan_v1_dense.py:HunYuanDenseV1ForSequenceClassification": ["GenericForSequenceClassification", "ModelYuanDenseV1ForSequenceClassification", "ModelYuanDenseV1PreTrainedModel", "class", "pass"], "siglip2/modeling_siglip2.py:Siglip2VisionOutput": ["FloatTensor", "ModelOutput", "ModelVisionOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "siglip2/modeling_siglip2.py:Siglip2TextOutput": ["FloatTensor", "ModelOutput", "ModelTextOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "siglip2/modeling_siglip2.py:Siglip2Output": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "siglip2/modeling_siglip2.py:Siglip2VisionEmbeddings": ["Embedding", "F", "False", "FloatTensor", "Linear", "LongTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "True", "__init__", "align_corners", "antialias", "batch_size", "bilinear", "class", "config", "cpu", "def", "device", "dtype", "embed_dim", "embeddings", "empty", "float32", "for", "forward", "height", "hidden_size", "i", "if", "in", "in_features", "int", "interpolate", "max_length", "mode", "nn", "num_channels", "num_patches", "out_features", "patch_embedding", "patch_embeds", "patch_size", "permute", "pixel_values", "position_embedding", "position_embedding_size", "positional_embeddings", "range", "reshape", "resize_positional_embeddings", "resized_embeddings", "resized_positional_embeddings", "resulted_positional_embeddings", "return", "self", "shape", "size", "source_dtype", "spatial_shapes", "staticmethod", "super", "target_dtype", "to", "torch", "transpose", "type", "unsqueeze", "weight", "width"], "siglip2/modeling_siglip2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "siglip2/modeling_siglip2.py:Siglip2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "siglip2/modeling_siglip2.py:Siglip2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "siglip2/modeling_siglip2.py:Siglip2EncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelMLP", "ModelTextConfig", "ModelVisionConfig", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "siglip2/modeling_siglip2.py:Siglip2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "siglip2/modeling_siglip2.py:Siglip2VisionTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "FloatTensor", "LayerNorm", "LongTensor", "ModelEncoder", "ModelMultiheadAttentionPoolingHead", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "None", "Optional", "Tensor", "True", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "and", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "dtype", "else", "embed_dim", "embeddings", "encoder", "encoder_attention_mask", "encoder_outputs", "eps", "flash_attention_2", "forward", "hasattr", "head", "hidden_size", "hidden_states", "if", "inputs_embeds", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooler_output", "post_layernorm", "r", "return", "self", "spatial_shapes", "super", "torch", "use_head", "vision_use_head"], "siglip2/modeling_siglip2.py:_trunc_normal_": ["The", "_trunc_normal_", "a", "add_", "b", "be", "clamp_", "def", "distribution", "erf", "erfinv_", "from", "if", "in", "incorrect", "init", "is", "l", "math", "max", "may", "mean", "min", "more", "mul_", "nn", "norm_cdf", "of", "or", "return", "sqrt", "stacklevel", "std", "tensor", "than", "trunc_normal_", "u", "uniform_", "values", "warn", "warnings", "x"], "siglip2/modeling_siglip2.py:trunc_normal_tf_": ["Model_normal_tf_", "Tensor", "_Model_normal_", "a", "add_", "b", "def", "float", "mean", "mul_", "no_grad", "std", "tensor", "torch", "with"], "siglip2/modeling_siglip2.py:variance_scaling_": ["Model", "Model_scaling_", "ValueError", "_calculate_fan_in_and_fan_out", "bound", "def", "denom", "distribution", "elif", "else", "f", "fan_avg", "fan_in", "fan_out", "if", "invalid", "math", "mode", "no_grad", "normal", "normal_", "raise", "scale", "sqrt", "std", "tensor", "torch", "trunc_normal_tf_", "truncated_normal", "uniform", "uniform_", "with"], "siglip2/modeling_siglip2.py:lecun_normal_": ["Model_normal_", "def", "distribution", "fan_in", "mode", "tensor", "truncated_normal", "variance_scaling_"], "siglip2/modeling_siglip2.py:default_flax_embed_init": ["Model_flax_embed_init", "def", "distribution", "fan_in", "mode", "normal", "tensor", "variance_scaling_"], "siglip2/modeling_siglip2.py:Siglip2PreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelForImageClassification", "ModelMLP", "ModelModel", "ModelMultiheadAttentionPoolingHead", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention", "attentions", "base_model_prefix", "bias", "class", "classifier", "config", "data", "def", "default_flax_embed_init", "elif", "else", "fc1", "fc2", "fill_", "hidden_size", "hidden_states", "if", "in_proj_bias", "in_proj_weight", "init", "initializer_factor", "is", "isinstance", "k_proj", "lecun_normal_", "log", "logit_bias", "logit_scale", "logit_scale_init", "module", "nn", "normal_", "not", "np", "out_proj", "position_embedding", "probe", "q_proj", "self", "sqrt", "std", "supports_gradient_checkpointing", "tensor", "torch", "v_proj", "vision_config", "weight", "width", "xavier_uniform_", "zero_", "zeros_"], "siglip2/modeling_siglip2.py:Siglip2TextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "siglip2/modeling_siglip2.py:Siglip2TextTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "Linear", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "and", "attention_mask", "auto_docstring", "can_return_tuple", "class", "config", "def", "dtype", "elif", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "final_layer_norm", "flash", "forward", "have", "head", "hidden_size", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "pooled_output", "pooler_output", "position_ids", "projection_size", "raise", "return", "self", "size", "specify", "super", "to", "torch", "uses_flash_attention", "view"], "siglip2/modeling_siglip2.py:Siglip2TextModel": ["BaseModelOutputWithPooling", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attention_mask", "auto_docstring", "check_model_inputs", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "kwargs", "nn", "position_ids", "post_init", "r", "return", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "value"], "siglip2/modeling_siglip2.py:Siglip2MultiheadAttentionPoolingHead": ["LayerNorm", "ModelMLP", "ModelMultiheadAttentionPoolingHead", "ModelVisionConfig", "Module", "MultiheadAttention", "None", "Optional", "Parameter", "Tensor", "True", "__init__", "_prepare_4d_attention_mask", "attention", "attention_mask", "attn_mask", "batch_first", "batch_size", "class", "config", "def", "dtype", "eps", "forward", "hidden_size", "hidden_state", "if", "is", "layer_norm_eps", "layernorm", "mlp", "nn", "not", "num_attention_heads", "num_heads", "probe", "randn", "repeat", "reshape", "residual", "return", "self", "shape", "source_len", "super", "target_len", "torch"], "siglip2/modeling_siglip2.py:Siglip2VisionModel": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_attention_mask", "pixel_values", "post_init", "r", "return", "self", "spatial_shapes", "super", "torch", "vision_model"], "siglip2/modeling_siglip2.py:Siglip2Model": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelVisionConfig", "ModelVisionModel", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "__init__", "_from_config", "attention_mask", "auto_docstring", "be", "bool", "but", "can_return_tuple", "class", "config", "def", "device", "dim", "else", "exp", "expected", "eye", "f", "filter_out_non_signature_kwargs", "forward", "functional", "get_image_features", "get_text_features", "if", "image_embeds", "input_ids", "is", "isinstance", "keepdim", "logit_bias", "logit_scale", "logits_per_image", "logits_per_text", "loglik", "logsigmoid", "loss", "m1_diag1", "matmul", "mean", "nll", "nn", "norm", "not", "of", "ones_like", "output_attentions", "output_hidden_states", "p", "pixel_attention_mask", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "r", "raise", "randn", "return", "return_loss", "self", "size", "spatial_shapes", "sum", "super", "t", "text_config", "text_embeds", "text_model", "text_model_output", "text_outputs", "to", "torch", "type", "vision_config", "vision_model", "vision_model_output", "vision_outputs"], "siglip2/modeling_siglip2.py:Siglip2ForImageClassification": ["BaseModelOutputWithPooling", "Identity", "ImageClassifierOutput", "Linear", "LongTensor", "ModelConfig", "ModelForImageClassification", "ModelPreTrainedModel", "ModelVisionModel", "None", "Optional", "Tensor", "__init__", "_from_config", "attention_mask", "attentions", "auto_docstring", "bool", "check_model_inputs", "class", "classifier", "config", "def", "device", "dim", "else", "forward", "hidden_size", "hidden_states", "if", "is", "labels", "last_hidden_state", "logits", "loss", "loss_function", "main_input_name", "mean", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "pixel_attention_mask", "pixel_values", "pool_mask", "post_init", "r", "return", "self", "sequence_output", "spatial_shapes", "sum", "super", "to", "torch", "vision_config", "vision_model"], "deberta_v2/modeling_deberta_v2.py:DebertaV2SelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super"], "deberta_v2/modeling_deberta_v2.py:make_log_bucket_position": ["Model_log_bucket_position", "abs", "abs_pos", "bucket_pos", "bucket_size", "ceil", "def", "int", "log", "log_pos", "max_position", "mid", "relative_pos", "return", "sign", "tensor", "torch", "type_as", "where"], "deberta_v2/modeling_deberta_v2.py:build_relative_position": ["Model_relative_position", "None", "and", "arange", "bucket_size", "def", "device", "dtype", "if", "int", "k_ids", "key_layer", "key_size", "long", "make_log_bucket_position", "max_position", "q_ids", "query_layer", "query_size", "rel_pos_ids", "return", "size", "to", "torch", "unsqueeze"], "deberta_v2/modeling_deberta_v2.py:c2p_dynamic_expand": ["Model_dynamic_expand", "Model_pos", "def", "expand", "query_layer", "relative_pos", "return", "size"], "deberta_v2/modeling_deberta_v2.py:p2c_dynamic_expand": ["Model_dynamic_expand", "c2p_pos", "def", "expand", "key_layer", "query_layer", "return", "size"], "deberta_v2/modeling_deberta_v2.py:pos_dynamic_expand": ["Model_dynamic_expand", "Model_index", "def", "expand", "key_layer", "p2c_att", "return", "size"], "deberta_v2/modeling_deberta_v2.py:scaled_size_sqrt": ["Model_size_sqrt", "Tensor", "def", "dtype", "float", "int", "query_layer", "return", "scale_factor", "size", "sqrt", "tensor", "torch"], "deberta_v2/modeling_deberta_v2.py:build_rpos": ["Model_relative_position", "Model_rpos", "bucket_size", "def", "else", "if", "int", "key_layer", "max_position", "max_relative_positions", "position_buckets", "query_layer", "relative_pos", "return", "size"], "deberta_v2/modeling_deberta_v2.py:DisentangledSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Model_attention_bias", "Module", "None", "Relative", "Tensor", "The", "True", "ValueError", "__init__", "_attention_head_size", "a", "all_head_size", "att_span", "attention", "attention_head_size", "attention_heads", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "be", "bias", "bmm", "bool", "bucket_size", "build_relative_position", "build_rpos", "c2p", "c2p_att", "c2p_pos", "clamp", "class", "config", "context_layer", "contiguous", "def", "device", "dim", "dropout", "dtype", "elif", "else", "expand", "f", "finfo", "forward", "functional", "gather", "getattr", "heads", "hidden", "hidden_dropout_prob", "hidden_size", "hidden_states", "ids", "if", "in", "index", "is", "key_layer", "key_proj", "long", "masked_fill", "max_position", "max_position_embeddings", "max_relative_positions", "min", "multiple", "must", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "p2c", "p2c_att", "p2c_pos", "permute", "pos_att_type", "pos_dropout", "pos_ebd_size", "pos_key_layer", "pos_key_proj", "pos_query_layer", "pos_query_proj", "position", "position_buckets", "query_layer", "query_proj", "query_states", "r_pos", "raise", "rel_att", "rel_embeddings", "relative_attention", "relative_pos", "repeat", "return", "scale", "scale_factor", "scaled_size_sqrt", "score", "self", "share_att_key", "size", "softmax", "squeeze", "super", "the", "to", "torch", "transpose", "transpose_for_scores", "unsqueeze", "value_layer", "value_proj", "view", "x"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Attention": ["DisentangledSelfAttention", "False", "ModelAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "att_matrix", "attention_mask", "attention_output", "bool", "class", "config", "def", "else", "forward", "hidden_states", "if", "is", "nn", "output", "output_attentions", "query_states", "rel_embeddings", "relative_pos", "return", "self", "self_output", "super", "torch", "tuple"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Intermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Output": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Layer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "att_matrix", "attention", "attention_mask", "attention_output", "bool", "class", "config", "def", "else", "forward", "hidden_states", "if", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "query_states", "rel_embeddings", "relative_pos", "return", "self", "super", "torch", "tuple"], "deberta_v2/modeling_deberta_v2.py:ConvLayer": ["ACT2FN", "Dropout", "LayerNorm", "Model", "Model1d", "ModelLayer", "Model_act", "Model_groups", "Model_kernel_size", "Module", "None", "__init__", "bool", "class", "config", "contiguous", "def", "dim", "dropout", "dtype", "else", "expand", "forward", "getattr", "groups", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_mask", "is", "kernel_size", "layer_norm_eps", "layer_norm_input", "masked_fill_", "nn", "out", "output", "output_states", "padding", "permute", "residual_states", "return", "rmask", "self", "size", "squeeze", "super", "tanh", "to", "unsqueeze"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Embeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Linear", "ModelEmbeddings", "Module", "None", "True", "__init__", "arange", "bias", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embed_proj", "embedding_size", "embeddings", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_biased_input", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "squeeze", "super", "to", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros", "zeros_like"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Encoder": ["BaseModelOutput", "ConvLayer", "Embedding", "False", "LayerNorm", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Sequence", "Tensor", "True", "_", "__init__", "all_attentions", "all_hidden_states", "and", "attention_mask", "attentions", "attn_weights", "bucket_size", "build_relative_position", "class", "config", "conv", "conv_kernel_size", "def", "dim", "elementwise_affine", "elif", "else", "enumerate", "extended_attention_mask", "for", "forward", "get_attention_mask", "get_rel_embedding", "get_rel_pos", "getattr", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "input_mask", "is", "isinstance", "last_hidden_state", "layer", "layer_module", "layer_norm", "layer_norm_eps", "len", "lower", "max_position", "max_position_embeddings", "max_relative_positions", "next_kv", "nn", "none", "norm_rel_ebd", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "output_states", "pos_ebd_size", "position_buckets", "query_states", "range", "rel_embeddings", "relative_attention", "relative_pos", "return", "return_dict", "self", "split", "squeeze", "strip", "sum", "super", "torch", "tuple", "unsqueeze", "v", "weight", "x"], "deberta_v2/modeling_deberta_v2.py:DebertaV2PreTrainedModel": ["Embedding", "LayerNorm", "LegacyModelLMPredictionHead", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_keys_to_ignore_on_load_unexpected", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "deberta_v2/modeling_deberta_v2.py:DebertaV2Model": ["BaseModelOutput", "False", "Model", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "and", "append", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoded_layers", "encoder", "encoder_outputs", "for", "forward", "function", "get_attention_mask", "get_input_embeddings", "get_rel_embedding", "get_rel_pos", "have", "heads_to_prune", "hidden_states", "if", "implemented", "in", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "layer", "layers", "long", "mask", "model", "new_embeddings", "not", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prune", "query_states", "raise", "range", "rel_embeddings", "rel_pos", "relative_pos", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "word_embeddings", "z_steps", "zeros"], "deberta_v2/modeling_deberta_v2.py:LegacyDebertaV2PredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "embedding_size", "eps", "forward", "getattr", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "deberta_v2/modeling_deberta_v2.py:LegacyDebertaV2LMPredictionHead": ["False", "Linear", "ModelModelLMPredictionHead", "ModelModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "embedding_size", "forward", "getattr", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "deberta_v2/modeling_deberta_v2.py:LegacyDebertaV2OnlyMLMHead": ["ModelModelLMPredictionHead", "ModelModelOnlyMLMHead", "Module", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super"], "deberta_v2/modeling_deberta_v2.py:DebertaV2LMPredictionHead": ["ACT2FN", "LayerNorm", "Linear", "ModelLMPredictionHead", "Module", "Parameter", "True", "__init__", "bias", "class", "config", "def", "dense", "elementwise_affine", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "matmul", "nn", "return", "self", "str", "super", "t", "torch", "transform_act_fn", "vocab_size", "weight", "word_embeddings", "zeros"], "deberta_v2/modeling_deberta_v2.py:DebertaV2OnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "__init__", "class", "config", "def", "forward", "lm_head", "nn", "prediction_scores", "return", "self", "sequence_output", "super", "word_embeddings"], "deberta_v2/modeling_deberta_v2.py:DebertaV2ForMaskedLM": ["CrossEntropyLoss", "LegacyModelOnlyMLMHead", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_keys_to_ignore_on_load_unexpected", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "dense", "else", "embeddings", "forward", "get_output_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "legacy", "lm_head", "lm_predictions", "logits", "loss", "loss_fct", "mask_predictions", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight", "word_embeddings"], "deberta_v2/modeling_deberta_v2.py:ContextPooler": ["ACT2FN", "Dropout", "Linear", "ModelPooler", "Model_token", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_size", "hidden_states", "nn", "output_dim", "pooled_output", "pooler_dropout", "pooler_hidden_act", "pooler_hidden_size", "property", "return", "self", "super"], "deberta_v2/modeling_deberta_v2.py:DebertaV2ForSequenceClassification": ["BCEWithLogitsLoss", "ContextPooler", "CrossEntropyLoss", "Dropout", "Linear", "LogSoftmax", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "cls_dropout", "config", "def", "dim", "drop_out", "dropout", "dtype", "elif", "else", "encoder_layer", "expand", "float", "forward", "gather", "get_input_embeddings", "getattr", "hidden_dropout_prob", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "label_index", "labeled_logits", "labels", "log_softmax", "logits", "long", "loss", "loss_fct", "loss_fn", "mean", "multi_label_classification", "new_embeddings", "nn", "nonzero", "not", "num_labels", "or", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "pooled_output", "pooler", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "set_input_embeddings", "single_label_classification", "size", "squeeze", "sum", "super", "tensor", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deberta_v2/modeling_deberta_v2.py:DebertaV2ForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deberta_v2/modeling_deberta_v2.py:DebertaV2ForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "deberta_v2/modeling_deberta_v2.py:DebertaV2ForMultipleChoice": ["ContextPooler", "CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "cls_dropout", "config", "def", "drop_out", "dropout", "else", "encoder_layer", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "get_input_embeddings", "getattr", "hidden_dropout_prob", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "new_embeddings", "nn", "not", "num_choices", "num_labels", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "pooled_output", "pooler", "position_ids", "r", "reshaped_logits", "return", "return_dict", "self", "set_input_embeddings", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "auto/modeling_auto.py:AutoModelForMaskGeneration": ["MODEL_FOR_MASK_GENERATION_MAPPING", "ModelModelForMaskGeneration", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForKeypointDetection": ["MODEL_FOR_KEYPOINT_DETECTION_MAPPING", "ModelModelForKeypointDetection", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForKeypointMatching": ["MODEL_FOR_KEYPOINT_MATCHING_MAPPING", "ModelModelForKeypointMatching", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForTextEncoding": ["MODEL_FOR_TEXT_ENCODING_MAPPING", "ModelModelForTextEncoding", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForImageToImage": ["MODEL_FOR_IMAGE_TO_IMAGE_MAPPING", "ModelModelForImageToImage", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModel": ["MODEL_MAPPING", "ModelModel", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForPreTraining": ["MODEL_FOR_PRETRAINING_MAPPING", "ModelModelForPreTraining", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:_AutoModelWithLMHead": ["MODEL_WITH_LM_HEAD_MAPPING", "_BaseModelModelClass", "_ModelModelWithLMHead", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForCausalLM": ["MODEL_FOR_CAUSAL_LM_MAPPING", "ModelModelForCausalLM", "PathLike", "Union", "_BaseModelModelClass", "_BaseModelWithGenerate", "_model_mapping", "class", "classmethod", "cls", "def", "from_pretrained", "kwargs", "model_args", "os", "pretrained_model_name_or_path", "return", "str", "super", "type"], "auto/modeling_auto.py:AutoModelForMaskedLM": ["MODEL_FOR_MASKED_LM_MAPPING", "ModelModelForMaskedLM", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForSeq2SeqLM": ["MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING", "ModelModelForSeq2SeqLM", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForSequenceClassification": ["MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING", "ModelModelForSequenceClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForQuestionAnswering": ["MODEL_FOR_QUESTION_ANSWERING_MAPPING", "ModelModelForQuestionAnswering", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForTableQuestionAnswering": ["MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING", "ModelModelForTableQuestionAnswering", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForVisualQuestionAnswering": ["MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING", "ModelModelForVisualQuestionAnswering", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForDocumentQuestionAnswering": ["MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING", "ModelModelForDocumentQuestionAnswering", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForTokenClassification": ["MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING", "ModelModelForTokenClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForMultipleChoice": ["MODEL_FOR_MULTIPLE_CHOICE_MAPPING", "ModelModelForMultipleChoice", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForNextSentencePrediction": ["MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING", "ModelModelForNextSentencePrediction", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForImageClassification": ["MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING", "ModelModelForImageClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForZeroShotImageClassification": ["MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING", "ModelModelForZeroShotImageClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForImageSegmentation": ["MODEL_FOR_IMAGE_SEGMENTATION_MAPPING", "ModelModelForImageSegmentation", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForSemanticSegmentation": ["MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING", "ModelModelForSemanticSegmentation", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForTimeSeriesPrediction": ["MODEL_FOR_TIME_SERIES_PREDICTION_MAPPING", "ModelModelForTimeSeriesPrediction", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForUniversalSegmentation": ["MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPING", "ModelModelForUniversalSegmentation", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForInstanceSegmentation": ["MODEL_FOR_INSTANCE_SEGMENTATION_MAPPING", "ModelModelForInstanceSegmentation", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForObjectDetection": ["MODEL_FOR_OBJECT_DETECTION_MAPPING", "ModelModelForObjectDetection", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForZeroShotObjectDetection": ["MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING", "ModelModelForZeroShotObjectDetection", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForDepthEstimation": ["MODEL_FOR_DEPTH_ESTIMATION_MAPPING", "ModelModelForDepthEstimation", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForVideoClassification": ["MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING", "ModelModelForVideoClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:_AutoModelForVision2Seq": ["MODEL_FOR_VISION_2_SEQ_MAPPING", "_BaseModelModelClass", "_ModelModelForVision2Seq", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForImageTextToText": ["MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING", "ModelModelForImageTextToText", "PathLike", "Union", "_BaseModelModelClass", "_BaseModelWithGenerate", "_model_mapping", "class", "classmethod", "cls", "def", "from_pretrained", "kwargs", "model_args", "os", "pretrained_model_name_or_path", "return", "str", "super", "type"], "auto/modeling_auto.py:AutoModelForAudioClassification": ["MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING", "ModelModelForAudioClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForCTC": ["MODEL_FOR_CTC_MAPPING", "ModelModelForCTC", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForSpeechSeq2Seq": ["MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING", "ModelModelForSpeechSeq2Seq", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForAudioFrameClassification": ["MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING", "ModelModelForAudioFrameClassification", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForAudioXVector": ["MODEL_FOR_AUDIO_XVECTOR_MAPPING", "ModelModelForAudioXVector", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForTextToSpectrogram": ["MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING", "ModelModelForTextToSpectrogram", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForTextToWaveform": ["MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING", "ModelModelForTextToWaveform", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoBackbone": ["MODEL_FOR_BACKBONE_MAPPING", "ModelBackbone", "_BaseModelBackboneClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForMaskedImageModeling": ["MODEL_FOR_MASKED_IMAGE_MODELING_MAPPING", "ModelModelForMaskedImageModeling", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelForAudioTokenization": ["MODEL_FOR_AUDIO_TOKENIZATION_MAPPING", "ModelModelForAudioTokenization", "_BaseModelModelClass", "_model_mapping", "class"], "auto/modeling_auto.py:AutoModelWithLMHead": ["FutureWarning", "ModelModelForCausalLM", "ModelModelForMaskedLM", "ModelModelForSeq2SeqLM", "ModelModelWithLMHead", "Please", "The", "_ModelModelWithLMHead", "a", "and", "be", "causal", "class", "classmethod", "cls", "config", "decoder", "def", "deprecated", "encoder", "for", "from_config", "from_pretrained", "future", "in", "is", "kwargs", "language", "masked", "model_args", "models", "pretrained_model_name_or_path", "removed", "return", "super", "use", "version", "warn", "warnings", "will"], "auto/modeling_auto.py:AutoModelForVision2Seq": ["FutureWarning", "ModelModelForImageTextToText", "ModelModelForVision2Seq", "Please", "The", "_ModelModelForVision2Seq", "and", "be", "class", "classmethod", "cls", "config", "def", "deprecated", "from_config", "from_pretrained", "in", "instead", "is", "kwargs", "model_args", "pretrained_model_name_or_path", "removed", "return", "super", "use", "v5", "warn", "warnings", "will"], "arcee/modeling_arcee.py:ArceeMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "arcee/modeling_arcee.py:ArceeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "arcee/modeling_arcee.py:ArceeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "arcee/modeling_arcee.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "arcee/modeling_arcee.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "arcee/modeling_arcee.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "arcee/modeling_arcee.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "arcee/modeling_arcee.py:ArceeAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "arcee/modeling_arcee.py:ArceeDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "arcee/modeling_arcee.py:ArceePreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "arcee/modeling_arcee.py:ArceeModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "arcee/modeling_arcee.py:ArceeForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "arcee/modeling_arcee.py:ArceeForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "arcee/modeling_arcee.py:ArceeForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "arcee/modeling_arcee.py:ArceeForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "poolformer/modeling_poolformer.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "poolformer/modeling_poolformer.py:PoolFormerDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "poolformer/modeling_poolformer.py:PoolFormerEmbeddings": ["Conv2d", "Identity", "Iterable", "ModelEmbeddings", "Module", "None", "__init__", "abc", "class", "collections", "def", "else", "embeddings", "forward", "hidden_size", "if", "isinstance", "kernel_size", "nn", "norm", "norm_layer", "num_channels", "padding", "patch_size", "pixel_values", "projection", "return", "self", "stride", "super"], "poolformer/modeling_poolformer.py:PoolFormerGroupNorm": ["GroupNorm", "ModelGroupNorm", "__init__", "class", "def", "kwargs", "nn", "num_channels", "self", "super"], "poolformer/modeling_poolformer.py:PoolFormerPooling": ["AvgModel2d", "False", "Model", "ModelModeling", "Model_size", "Module", "__init__", "class", "count_include_pad", "def", "forward", "hidden_states", "nn", "padding", "return", "self", "stride", "super"], "poolformer/modeling_poolformer.py:PoolFormerOutput": ["ACT2FN", "Conv2d", "ModelDropPath", "ModelOutput", "Module", "__init__", "act_fn", "class", "config", "conv1", "conv2", "def", "drop", "dropout_prob", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super"], "poolformer/modeling_poolformer.py:PoolFormerLayer": ["Identity", "ModelDropPath", "ModelGroupNorm", "ModelLayer", "ModelModeling", "ModelOutput", "Model_size", "Modeling", "Modeling_output", "Module", "Parameter", "True", "__init__", "after_norm", "before_norm", "class", "config", "def", "drop_path", "else", "forward", "hidden_size", "hidden_states", "if", "intermediate_size", "layer_output", "layer_scale_1", "layer_scale_2", "layer_scale_init_value", "nn", "num_channels", "ones", "output", "outputs", "requires_grad", "return", "scaled_op", "self", "super", "torch", "unsqueeze", "use_layer_scale"], "poolformer/modeling_poolformer.py:PoolFormerEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "Model_size", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "append", "blk", "block", "block_layer", "blocks", "class", "config", "cpu", "cur", "def", "depths", "device", "dpr", "drop_path", "drop_path_rate", "else", "embedding_layer", "embeddings", "enumerate", "for", "forward", "hidden_size", "hidden_sizes", "hidden_states", "i", "idx", "if", "in", "int", "intermediate_size", "is", "item", "j", "last_hidden_state", "layer_outputs", "layers", "linspace", "mlp_ratio", "nn", "not", "num_channels", "num_encoder_blocks", "output_hidden_states", "padding", "patch_embeddings", "patch_size", "patch_sizes", "pixel_values", "range", "return", "return_dict", "self", "stride", "strides", "sum", "super", "torch", "tuple", "v", "x", "zip"], "poolformer/modeling_poolformer.py:PoolFormerPreTrainedModel": ["Conv2d", "GroupNorm", "Linear", "Model", "ModelConfig", "ModelLayer", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "layer_scale_1", "layer_scale_2", "layer_scale_init_value", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "poolformer/modeling_poolformer.py:PoolFormerModel": ["BaseModelOutputWithNoAttention", "FloatTensor", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "ValueError", "You", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embeddings", "encoder", "encoder_outputs", "forward", "get_input_embeddings", "have", "hidden_states", "if", "is", "last_hidden_state", "not", "output_hidden_states", "patch_embeddings", "pixel_values", "post_init", "raise", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "poolformer/modeling_poolformer.py:PoolFormerFinalPooler": ["Linear", "ModelFinalModeler", "Module", "__init__", "class", "config", "def", "dense", "forward", "hidden_size", "hidden_states", "nn", "output", "return", "self", "super"], "poolformer/modeling_poolformer.py:PoolFormerForImageClassification": ["FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelGroupNorm", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "mean", "nn", "norm", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "longformer/modeling_longformer.py:LongformerBaseModelOutput": ["FloatTensor", "ModelBaseModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "longformer/modeling_longformer.py:LongformerBaseModelOutputWithPooling": ["FloatTensor", "ModelBaseModelOutputWithPooling", "ModelOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "last_hidden_state", "pooler_output", "r", "torch", "tuple"], "longformer/modeling_longformer.py:LongformerMaskedLMOutput": ["FloatTensor", "ModelMaskedLMOutput", "ModelOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "longformer/modeling_longformer.py:LongformerQuestionAnsweringModelOutput": ["FloatTensor", "ModelOutput", "ModelQuestionAnsweringModelOutput", "None", "Optional", "attentions", "class", "end_logits", "global_attentions", "hidden_states", "loss", "r", "start_logits", "torch", "tuple"], "longformer/modeling_longformer.py:LongformerSequenceClassifierOutput": ["FloatTensor", "ModelOutput", "ModelSequenceClassifierOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "longformer/modeling_longformer.py:LongformerMultipleChoiceModelOutput": ["FloatTensor", "ModelMultipleChoiceModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "longformer/modeling_longformer.py:LongformerTokenClassifierOutput": ["FloatTensor", "ModelOutput", "ModelTokenClassifierOutput", "None", "Optional", "attentions", "class", "global_attentions", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "longformer/modeling_longformer.py:_get_question_end_index": ["There", "You", "_get_question_end_index", "also", "answering", "assert", "avoid", "batch_size", "be", "consider", "def", "dimensions", "error", "every", "exactly", "f", "for", "forward", "function", "global_attention_mask", "have", "in", "input_ids", "manually", "might", "nonzero", "questions", "return", "sample", "sep_token_id", "sep_token_indices", "separator", "set", "shape", "should", "the", "this", "three", "to", "tokens", "two", "view"], "longformer/modeling_longformer.py:_compute_global_attention_mask": ["True", "_compute_global_attention_mask", "_get_question_end_index", "arange", "attention_mask", "before_sep_token", "bool", "def", "device", "dim", "else", "expand_as", "if", "input_ids", "is", "question_end_index", "return", "sep_token_id", "shape", "to", "torch", "unsqueeze"], "longformer/modeling_longformer.py:create_position_ids_from_input_ids": ["Model_position_ids_from_input_ids", "cumsum", "def", "dim", "incremental_indices", "input_ids", "int", "long", "mask", "ne", "padding_idx", "return", "torch", "type_as"], "longformer/modeling_longformer.py:LongformerEmbeddings": ["Dropout", "Embedding", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "arange", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "position_embeddings", "position_ids", "return", "self", "sequence_length", "size", "super", "to", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "longformer/modeling_longformer.py:LongformerSelfAttention": ["False", "Given", "Head", "Linear", "ModelSelfAttention", "Module", "None", "Sequence", "Size", "Tensor", "The", "True", "Unexpected", "ValueError", "__init__", "_chunk", "_compute_attn_output_with_global_indices", "_compute_global_attn_output_from_hidden", "_concat_with_global_key_attn_probs", "_get_global_attn_indices", "_mask_invalid_locations", "_pad_and_diagonalize", "_pad_and_transpose_last_two_dims", "_sliding_chunks_matmul_attn_probs_value", "_sliding_chunks_query_key_matmul", "a", "affected_seq_len", "an", "and", "arange", "as_strided", "as_tuple", "assert", "attention", "attention_mask", "attention_probs_dropout_prob", "attention_window", "attn_output", "attn_output_only_global", "attn_output_without_global", "attn_probs", "attn_probs_from_global_key", "attn_probs_only_global", "attn_probs_without_global", "attn_scores", "batch_size", "bcdh", "bcwd", "bcwh", "bcxd", "bcxy", "bcyd", "be", "beginning_input", "beginning_mask", "beginning_mask_2d", "blhd", "blhs", "bmm", "bool", "bshd", "but", "cat", "chunk", "chunk_size", "chunk_stride", "chunked_attn_probs", "chunked_hidden_states", "chunked_value", "chunked_value_size", "chunked_value_stride", "chunks_count", "class", "clone", "config", "context", "contiguous", "def", "del", "device", "diagonal_attention_scores", "diagonal_chunked_attention_scores", "diagonal_mask", "dim", "dims", "div", "dropout", "dtype", "einsum", "else", "embed_dim", "empty", "ending_input", "ending_mask", "even", "expand", "f", "finfo", "flip", "float", "float32", "float_mask", "for", "forward", "full_like", "functional", "getattr", "global_attn_hidden_states", "global_attn_output", "global_attn_probs", "global_attn_probs_float", "global_attn_scores", "global_key_attn_scores", "global_key_vectors", "global_query_vectors_only_global", "global_value_vectors", "has", "have", "head_dim", "heads", "hidden", "hidden_dim", "hidden_size", "hidden_states", "hidden_states_padded", "if", "in", "inf", "input_tensor", "int", "is", "is_global_attn", "is_index_global_attn", "is_index_global_attn_nonzero", "is_index_masked", "is_local_index_global_attn", "is_local_index_global_attn_nonzero", "is_local_index_no_global_attn_nonzero", "key", "key_global", "key_vectors", "key_vectors_only_global", "layer", "layer_head_mask", "layer_id", "len", "length", "list", "local_attn_probs", "long", "mask", "masked_fill", "math", "matmul", "max", "max_num_global_attn_indices", "min", "multiple", "narrow", "new_ones", "new_zeros", "nn", "nonzero", "nonzero_global_attn_output", "not", "num_attention_heads", "num_chunks", "num_global_attn_indices", "num_heads", "number", "of", "one_sided_attn_window_size", "onnx_export", "output_attentions", "outputs", "overlapping_chunks", "p", "pad", "padded_value", "padding", "positive", "query", "query_global", "query_vectors", "raise", "range", "remove_from_windowed_attention_mask", "reshape", "return", "rounding_mode", "self", "seq_len", "shape", "should", "single", "size", "softmax", "sqrt", "staticmethod", "stride", "sum", "super", "tensor", "the", "to", "torch", "total_num_heads", "training", "transpose", "tril", "trunc", "type_as", "unsqueeze", "value", "value_global", "value_vectors", "value_vectors_only_global", "view", "where", "window_overlap", "wrong"], "longformer/modeling_longformer.py:LongformerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "longformer/modeling_longformer.py:LongformerAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attn_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_states", "if", "index", "is_global_attn", "is_index_global_attn", "is_index_masked", "key", "layer_head_mask", "layer_id", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "longformer/modeling_longformer.py:LongformerIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "longformer/modeling_longformer.py:LongformerOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "longformer/modeling_longformer.py:LongformerLayer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attn_output", "chunk_size_feed_forward", "class", "config", "def", "ff_chunk", "forward", "hidden_states", "intermediate", "intermediate_output", "is_global_attn", "is_index_global_attn", "is_index_masked", "layer_head_mask", "layer_id", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attn_outputs", "seq_len_dim", "super"], "longformer/modeling_longformer.py:LongformerEncoder": ["False", "ModelBaseModelOutput", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "The", "True", "__init__", "all_attentions", "all_global_attentions", "all_hidden_states", "and", "any", "assert", "attention_mask", "attentions", "be", "but", "class", "config", "def", "else", "enumerate", "f", "flatten", "for", "forward", "global_attentions", "gradient_checkpointing", "head_mask", "hidden_states", "i", "idx", "if", "in", "is", "is_global_attn", "is_index_global_attn", "is_index_masked", "it", "item", "last_hidden_state", "layer", "layer_head_mask", "layer_id", "layer_module", "layer_outputs", "layers", "len", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "padding_len", "range", "return", "return_dict", "self", "shape", "should", "size", "specified", "state", "super", "transpose", "tuple", "v"], "longformer/modeling_longformer.py:LongformerPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "longformer/modeling_longformer.py:LongformerLMHead": ["LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "longformer/modeling_longformer.py:LongformerPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "longformer/modeling_longformer.py:LongformerModel": ["Expected", "Given", "Input", "ModelBaseModelOutputWithPooling", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_merge_to_attention_mask", "_pad_to_window_size", "_prune_heads", "a", "add_pooling_layer", "an", "and", "are", "assert", "at", "attention", "attention_mask", "attention_window", "attentions", "auto_docstring", "automatically", "batch_size", "be", "bool", "both", "cannot", "cat", "class", "config", "def", "device", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "equal", "even", "extended_attention_mask", "f", "for", "forward", "functional", "get_extended_attention_mask", "get_input_embeddings", "given", "global_attention_mask", "global_attentions", "has", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "ids", "if", "in", "input_ids", "input_ids_padding", "input_shape", "inputs_embeds", "inputs_embeds_padding", "int", "is", "isinstance", "items", "last_hidden_state", "layer", "len", "logger", "long", "max", "multiple", "new_full", "nn", "not", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "pad", "pad_token_id", "padded", "padding_len", "pooled_output", "pooler", "pooler_output", "position_ids", "positive", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_len", "sequence_output", "set_input_embeddings", "shape", "should", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "warning_once", "word_embeddings", "zeros"], "longformer/modeling_longformer.py:LongformerForMaskedLM": ["CrossEntropyLoss", "False", "Model", "ModelForMaskedLM", "ModelLMHead", "ModelMaskedLMOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "decoder", "def", "device", "else", "forward", "get_output_embeddings", "global_attention_mask", "global_attentions", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_head", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size"], "longformer/modeling_longformer.py:LongformerForSequenceClassification": ["BCEWithLogitsLoss", "CLS", "CrossEntropyLoss", "False", "Initializing", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "ModelSequenceClassifierOutput", "None", "Optional", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "global", "global_attention_mask", "global_attentions", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logger", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "on", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "warning_once", "zeros_like"], "longformer/modeling_longformer.py:LongformerClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "kwargs", "nn", "num_labels", "out_proj", "output", "return", "self", "super", "tanh", "torch"], "longformer/modeling_longformer.py:LongformerForQuestionAnswering": ["CrossEntropyLoss", "False", "It", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "ModelQuestionAnsweringModelOutput", "None", "Optional", "Please", "Tensor", "Union", "__init__", "_compute_global_attention_mask", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "automatically", "because", "bool", "clamp", "class", "config", "contiguous", "correctly", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "generate", "global_attention_mask", "global_attentions", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "it", "len", "logger", "logits", "loss", "loss_fct", "make", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "possible", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sep_token_id", "sequence_output", "set", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "sure", "that", "the", "to", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "warning"], "longformer/modeling_longformer.py:LongformerForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "ModelTokenClassifierOutput", "None", "Optional", "Tensor", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "device", "dropout", "else", "forward", "global_attention_mask", "global_attentions", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "longformer/modeling_longformer.py:LongformerForMultipleChoice": ["CrossEntropyLoss", "Dropout", "False", "Initializing", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelMultipleChoiceModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_compute_global_attention_mask", "and", "attention", "attention_mask", "attentions", "auto_docstring", "before_sep_token", "bool", "choice", "class", "classifier", "config", "def", "device", "dim", "dropout", "else", "flat_attention_mask", "flat_global_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "for", "forward", "global", "global_attention_mask", "global_attentions", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "i", "if", "in", "input_ids", "inputs_embeds", "is", "labels", "logger", "logits", "loss", "loss_fct", "multiple", "nn", "not", "num_choices", "on", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "range", "reshaped_logits", "return", "return_dict", "self", "sep_token_id", "shape", "size", "stack", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "warning_once"], "esm/modeling_esmfold.py:EsmForProteinFoldingOutput": ["FloatTensor", "ModelForProteinFoldingOutput", "ModelOutput", "None", "Optional", "aatype", "aligned_confidence_probs", "angles", "atom14_atom_exists", "atom37_atom_exists", "class", "distogram_logits", "frames", "lddt_head", "lm_logits", "max_predicted_aligned_error", "plddt", "positions", "predicted_aligned_error", "ptm", "ptm_logits", "r", "residue_index", "residx_atom14_to_atom37", "residx_atom37_to_atom14", "s_s", "s_z", "sidechain_frames", "states", "torch", "unnormalized_angles"], "esm/modeling_esmfold.py:is_fp16_enabled": ["and", "autocast_dtype", "def", "device_type", "else", "float16", "fp16_enabled", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "if", "is_autocast_enabled", "is_fp16_enabled", "return", "torch"], "esm/modeling_esmfold.py:is_deepspeed_initialized": ["Exception", "False", "deepspeed", "def", "else", "except", "if", "is_deepspeed_available", "is_deepspeed_initialized", "is_initialized", "return", "try", "utils"], "esm/modeling_esmfold.py:collate_dense_tensors": ["Model_dense_tensors", "RuntimeError", "Samples", "Tensor", "def", "device", "dim", "dimensions", "dtype", "empty", "f", "fill_", "float", "for", "has", "i", "if", "in", "k", "len", "list", "lst", "max", "max_shape", "pad_v", "raise", "range", "result", "result_i", "return", "samples", "shape", "slice", "t", "torch", "tuple", "varying", "x", "zip"], "esm/modeling_esmfold.py:flatten_final_dims": ["Model_final_dims", "Tensor", "def", "int", "no_dims", "reshape", "return", "shape", "t", "torch"], "esm/modeling_esmfold.py:permute_final_dims": ["Model", "Model_final_dims", "Tensor", "def", "first_inds", "for", "i", "in", "inds", "int", "len", "list", "range", "return", "shape", "tensor", "torch", "zero_index"], "esm/modeling_esmfold.py:dict_multimap": ["Model", "Model_multimap", "Models", "all_v", "d", "def", "else", "first", "fn", "for", "if", "in", "isinstance", "items", "k", "new_Model", "return", "v"], "esm/modeling_esmfold.py:trunc_normal_init_": ["Model_normal_init_", "Modelnorm", "This", "a", "an", "approximation", "b", "be", "but", "clamp", "copy_", "def", "default", "device", "else", "equivalent", "fan", "fan_in", "found", "if", "init", "is_scipy_available", "loc", "logger", "math", "max", "might", "min", "nn", "normal_", "not", "np", "numel", "requires", "reshape", "rvs", "samples", "scale", "scipy", "shape", "size", "sqrt", "std", "tensor", "that", "to", "torch", "warning", "was", "weights"], "esm/modeling_esmfold.py:ipa_point_weights_init_": ["Model_point_weights_init_", "def", "fill_", "no_grad", "softplus_inverse_1", "torch", "weights", "with"], "esm/modeling_esmfold.py:EsmFoldLinear": ["Callable", "Invalid", "Linear", "ModelFoldLinear", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "bias", "bool", "class", "def", "default", "fill_", "final", "gating", "glorot", "if", "in", "in_dim", "init", "init_fn", "int", "nn", "no_grad", "normal", "not", "out_dim", "raise", "relu", "self", "str", "string", "super", "torch", "with"], "esm/modeling_esmfold.py:EsmFoldLayerNorm": ["False", "ModelFoldLayerNorm", "Module", "Parameter", "__init__", "and", "autocast", "bfloat16", "bias", "c_in", "class", "cuda", "d", "def", "device_type", "dtype", "else", "enabled", "eps", "forward", "functional", "if", "is", "is_deepspeed_initialized", "layer_norm", "nn", "not", "ones", "out", "return", "self", "super", "to", "torch", "weight", "with", "x", "zeros"], "esm/modeling_esmfold.py:softmax_no_cast": ["False", "Model", "Model_no_cast", "Tensor", "and", "autocast", "bfloat16", "cuda", "d", "def", "device_type", "dim", "dtype", "else", "enabled", "functional", "if", "int", "is", "is_deepspeed_initialized", "nn", "not", "return", "s", "t", "torch", "with"], "esm/modeling_esmfold.py:EsmFoldAttention": ["Choose", "False", "For", "If", "ModelFoldAttention", "ModelFoldLinear", "Module", "None", "Optional", "Sigmoid", "Tensor", "True", "ValueError", "__init__", "_prep_qkv", "_wrap_up", "algorithm", "alternative", "and", "at", "attention", "attn_options", "b", "be", "bias", "biases", "bool", "c_hidden", "c_k", "c_q", "c_v", "class", "def", "final", "flash_mask", "flatten_final_dims", "for", "forward", "g", "gating", "glorot", "if", "in", "incompatible", "init", "instead", "int", "is", "k", "key", "kv_x", "linear_g", "linear_k", "linear_o", "linear_q", "linear_v", "list", "lma_kv_chunk_size", "lma_q_chunk_size", "masking", "math", "matmul", "most", "must", "nn", "no_heads", "not", "o", "one", "option", "or", "output", "permute_final_dims", "provided", "q", "q_x", "query", "raise", "return", "self", "shape", "sigmoid", "softmax_no_cast", "specified", "sqrt", "sum", "super", "the", "torch", "transpose", "tuple", "use", "use_flash", "use_lma", "use_memory_efficient_kernel", "v", "value", "view", "with"], "esm/modeling_esmfold.py:EsmFoldTriangleAttention": ["False", "LayerNorm", "ModelFoldAttention", "ModelFoldLinear", "ModelFoldTriangleAttention", "Module", "None", "Optional", "Tensor", "True", "__init__", "_chunk", "_out", "bias", "biases", "bool", "c_hidden", "c_in", "chunk_layer", "chunk_size", "class", "def", "else", "forward", "if", "ignore", "inf", "init", "inplace_safe", "int", "is", "jit", "kv_x", "layer_norm", "len", "linear", "list", "mask", "mask_bias", "mha", "mha_inputs", "new_ones", "nn", "no_batch_dims", "no_heads", "normal", "not", "partial", "permute_final_dims", "q_x", "return", "self", "shape", "starting", "super", "torch", "transpose", "triangle", "triangle_bias", "unsqueeze", "use_lma", "use_memory_efficient_kernel", "x"], "esm/modeling_esmfold.py:EsmFoldTriangleMultiplicativeUpdate": ["False", "LayerNorm", "ModelFoldLinear", "ModelFoldTriangleMultiplicativeUpdate", "Module", "None", "Optional", "Sigmoid", "Tensor", "True", "_", "__init__", "_add_with_inplace", "_combine_projections", "_inference_forward", "_inplace_chunk_size", "_outgoing", "a", "a_chunk", "after_half", "after_half_offsets", "and", "autocast", "b", "b_chunk", "b_chunk_dim", "bias", "bool", "c", "c_hidden", "chunked", "class", "clone", "col_dim", "combined_range_with_offsets", "compute_projection", "compute_projection_helper", "config", "copy_", "cpu", "def", "del", "device", "device_type", "dim", "else", "empty_slicer", "enabled", "end", "final", "first_half_slicer", "flip_z_cache_", "float", "for", "forward", "g", "g_chunk", "gating", "half_n", "i", "i_1", "i_2", "i_range", "if", "in", "init", "initial_offsets", "inplace_chunk_size", "inplace_safe", "int", "is", "is_fp16_enabled", "layer_norm_in", "layer_norm_out", "linear_a_g", "linear_a_p", "linear_b_g", "linear_b_p", "linear_g", "linear_p", "linear_z", "list", "mask", "mask_chunk", "matmul", "mps", "n", "need_transpose", "new_ones", "new_zeros", "nn", "not", "offset", "out_shape", "p", "pair", "pair_chunk", "pairwise_state_dim", "permute_final_dims", "quadrant_3", "quadrant_3_slicer", "quadrant_4", "range", "return", "row_dim", "s", "self", "shape", "sigmoid", "sigmoid_", "slice", "slice_tensor", "start", "super", "t", "torch", "transpose", "type", "unsqueeze", "with", "with_add", "x", "x_chunk", "z", "z_cache", "z_cache_offset", "z_cache_rotated", "z_cache_shape", "z_cache_slicer", "z_chunk_b", "z_chunk_g", "z_chunk_slicer", "z_slicer", "zip"], "esm/modeling_esmfold.py:EsmFoldPreTrainedModel": ["ModelFoldInvariantPointAttention", "ModelFoldLinear", "ModelFoldPreTrainedModel", "ModelFoldTriangularSelfAttentionBlock", "ModelPreTrainedModel", "None", "_init_weights", "bias", "class", "def", "default", "elif", "else", "fill_", "final", "gain", "gating", "glorot", "head_weights", "if", "init", "init_fn", "ipa_point_weights_init_", "is", "isinstance", "kaiming_normal_", "linear", "linear_o", "linear_z", "mha", "mlp", "mlp_pair", "mlp_seq", "module", "nn", "no_grad", "nonlinearity", "normal", "not", "o_proj", "pair_to_sequence", "relu", "scale", "self", "seq_attention", "sequence_to_pair", "super", "torch", "tri_att_end", "tri_att_start", "tri_mul_in", "tri_mul_out", "trunc_normal_init_", "weight", "with", "xavier_uniform_", "zeros_"], "esm/modeling_esmfold.py:EsmFoldSelfAttention": ["False", "Linear", "ModelFoldSelfAttention", "Module", "None", "True", "__init__", "a", "assert", "bias", "chunk", "class", "def", "dim", "einsum", "embed_dim", "forward", "functional", "g_proj", "gated", "head_width", "hkc", "hqk", "if", "indices", "inf", "init", "is", "k", "kc", "mask", "masked_fill", "nn", "not", "np", "num_heads", "o_proj", "ones_", "permute", "proj", "q", "qc", "qhc", "qk", "rescale_factor", "reshape", "return", "self", "shape", "sigmoid", "softmax", "super", "t", "torch", "v", "view", "weight", "x", "y", "zeros_"], "esm/modeling_esmfold.py:EsmFoldDropout": ["Dropout", "ModelFoldDropout", "Module", "None", "Tensor", "Union", "__init__", "batch_dim", "bd", "class", "def", "dropout", "float", "for", "forward", "if", "in", "int", "is", "isinstance", "list", "new_ones", "nn", "not", "r", "return", "self", "shape", "super", "torch", "x"], "esm/modeling_esmfold.py:EsmFoldSequenceToPair": ["LayerNorm", "Linear", "ModelFoldSequenceToPair", "Module", "None", "True", "__init__", "assert", "bias", "cat", "chunk", "class", "def", "diff", "dim", "forward", "init", "inner_dim", "k", "layernorm", "len", "nn", "o_proj", "pairwise_state_dim", "prod", "proj", "q", "return", "s", "self", "sequence_state", "sequence_state_dim", "shape", "super", "torch", "x", "zeros_"], "esm/modeling_esmfold.py:EsmFoldPairToSequence": ["False", "LayerNorm", "Linear", "ModelFoldPairToSequence", "Module", "__init__", "assert", "bias", "class", "def", "forward", "layernorm", "len", "linear", "nn", "num_heads", "pairwise_bias", "pairwise_state", "pairwise_state_dim", "return", "self", "shape", "super", "z"], "esm/modeling_esmfold.py:EsmFoldResidueMLP": ["Dropout", "LayerNorm", "Linear", "ModelFoldResidueMLP", "Module", "ReLU", "Sequential", "__init__", "class", "def", "dropout", "embed_dim", "forward", "inner_dim", "mlp", "nn", "return", "self", "super", "x"], "esm/modeling_esmfold.py:EsmFoldTriangularSelfAttentionBlock": ["Dropout", "False", "Got", "LayerNorm", "ModelFoldDropout", "ModelFoldPairToSequence", "ModelFoldResidueMLP", "ModelFoldSelfAttention", "ModelFoldSequenceToPair", "ModelFoldTriangleAttention", "ModelFoldTriangleMultiplicativeUpdate", "ModelFoldTriangularSelfAttentionBlock", "Module", "None", "True", "ValueError", "_", "__init__", "__kwargs", "_outgoing", "a", "and", "batch", "batch_dim", "be", "bias", "chunk_size", "class", "col_drop", "config", "def", "dimension", "dims", "drop", "dropout", "else", "equal", "f", "forward", "gated", "got", "have", "if", "inconsistent", "inf", "is", "last", "layernorm_1", "len", "length", "mask", "mlp_pair", "mlp_seq", "nn", "not", "or", "pair_to_sequence", "pairwise_head_width", "pairwise_num_heads", "pairwise_state", "pairwise_state_dim", "raise", "return", "row_drop", "self", "seq_attention", "seq_dim", "sequence", "sequence_head_width", "sequence_num_heads", "sequence_state", "sequence_state_dim", "sequence_to_pair", "shape", "should", "size", "starting", "super", "tensor", "to", "tri_att_end", "tri_att_start", "tri_mask", "tri_mul_in", "tri_mul_out", "unsqueeze", "y"], "esm/modeling_esmfold.py:EsmCategoricalMixture": ["ModelCategoricalMixture", "None", "__init__", "abs", "argmin", "bins", "class", "def", "device", "dim", "dtype", "end", "linspace", "log_prob", "log_softmax", "logits", "mean", "ndim", "nll", "param", "return", "self", "softmax", "squeeze", "start", "take_along_dim", "torch", "true", "true_index", "unsqueeze", "v_bins"], "esm/modeling_esmfold.py:categorical_lddt": ["ModelModelMixture", "Model_lddt", "bins", "def", "logits", "mean", "return"], "esm/modeling_esmfold.py:get_axial_mask": ["Model_axial_mask", "None", "ValueError", "a", "batch_dim", "be", "def", "dims", "expand", "f", "got", "if", "is", "len", "m", "mask", "raise", "reshape", "return", "seq_dim", "shape", "should", "tensor", "unsqueeze"], "esm/modeling_esmfold.py:EsmFoldRelativePosition": ["Embedding", "False", "ModelFoldRelativePosition", "Module", "None", "ValueError", "__init__", "and", "be", "bins", "clamp", "class", "config", "def", "diff", "dtype", "embedding", "f", "forward", "has", "have", "if", "inconsistent", "is", "it", "long", "mask", "nn", "not", "output", "pairwise_state_dim", "position_bins", "raise", "residue_index", "return", "self", "shape", "shapes", "should", "super", "torch"], "esm/modeling_esmfold.py:EsmFoldAngleResnetBlock": ["ModelFoldAngleResnetBlock", "ModelFoldLinear", "Module", "ReLU", "Tensor", "__init__", "a", "class", "config", "def", "final", "forward", "init", "linear_1", "linear_2", "nn", "relu", "resnet_dim", "return", "s_initial", "self", "super", "torch"], "esm/modeling_esmfold.py:EsmFoldAngleResnet": ["ModelFoldAngleResnet", "ModelFoldAngleResnetBlock", "ModelFoldLinear", "Module", "ModuleList", "ReLU", "Tensor", "True", "_", "__init__", "append", "clamp", "class", "config", "def", "dim", "epsilon", "for", "forward", "in", "keepdim", "l", "layer", "layers", "linear_in", "linear_initial", "linear_out", "min", "nn", "norm_denom", "num_angles", "num_resnet_blocks", "range", "relu", "resnet_dim", "return", "s", "s_initial", "self", "sequence_dim", "shape", "sqrt", "sum", "super", "torch", "tuple", "unnormalized_s", "view"], "esm/modeling_esmfold.py:EsmFoldInvariantPointAttention": ["False", "ModelFoldInvariantPointAttention", "ModelFoldLinear", "Module", "None", "Optional", "Parameter", "Rigid", "Sequence", "Softmax", "Softplus", "Tensor", "__init__", "_offload_inference", "_z_reference_list", "a", "apply", "assert", "autocast", "b", "bool", "c_s", "c_z", "cat", "class", "concat_out_dim", "config", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "enabled", "epsilon", "final", "flatten_final_dims", "float", "forward", "getrefcount", "hc", "head_weights", "hidden_dim", "hpkv", "hpq", "if", "inf", "init", "invert_apply", "ipa_dim", "is_fp16_enabled", "k", "k_pts", "kv", "kv_pts", "len", "linear_b", "linear_kv", "linear_kv_points", "linear_out", "linear_q", "linear_q_points", "mask", "math", "matmul", "mps", "nn", "num_heads", "num_heads_ipa", "num_qk_points", "num_v_points", "o", "o_pair", "o_pt", "o_pt_norm", "pairwise_dim", "permute_final_dims", "pt_att", "q", "q_pts", "r", "reshape", "return", "s", "self", "sequence_dim", "shape", "softmax", "softplus", "split", "sqrt", "square_mask", "stack", "sum", "super", "sys", "to", "torch", "transpose", "type", "unbind", "unsqueeze", "v", "v_pts", "view", "with", "z", "zeros"], "esm/modeling_esmfold.py:EsmFoldBackboneUpdate": ["ModelFoldBackboneUpdate", "ModelFoldLinear", "Module", "Tensor", "__init__", "class", "config", "def", "final", "forward", "init", "linear", "nn", "return", "s", "self", "sequence_dim", "super", "torch", "tuple", "update"], "esm/modeling_esmfold.py:EsmFoldStructureModuleTransitionLayer": ["ModelFoldLinear", "ModelFoldStructureModuleTransitionLayer", "Module", "ReLU", "__init__", "class", "config", "def", "final", "forward", "init", "linear_1", "linear_2", "linear_3", "nn", "relu", "return", "s", "s_initial", "self", "sequence_dim", "super"], "esm/modeling_esmfold.py:EsmFoldStructureModuleTransition": ["Dropout", "LayerNorm", "ModelFoldStructureModuleTransition", "ModelFoldStructureModuleTransitionLayer", "Module", "ModuleList", "_", "__init__", "append", "class", "config", "def", "dropout", "dropout_rate", "for", "forward", "in", "l", "layer_norm", "layers", "nn", "num_transition_layers", "range", "return", "s", "self", "sequence_dim", "super"], "esm/modeling_esmfold.py:EsmFoldStructureModule": ["Dropout", "False", "LayerNorm", "ModelFoldAngleResnet", "ModelFoldBackboneUpdate", "ModelFoldInvariantPointAttention", "ModelFoldLinear", "ModelFoldStructureModule", "ModelFoldStructureModuleTransition", "Module", "None", "Rigid", "Rotation", "__init__", "_init_residue_constants", "_offload_inference", "_z_reference_list", "aatype", "all_frames_to_global", "alpha", "angle_resnet", "angles", "append", "assert", "atom_mask", "backb_to_global", "bb_update", "class", "compose_q_update_vec", "config", "cpu", "def", "default_frames", "del", "device", "dict_multimap", "dropout_rate", "dtype", "evoformer_output_dict", "f", "float_dtype", "fmt", "for", "forward", "frames", "frames_and_literature_positions_to_atom14_pos", "get_rot_mats", "get_rots", "get_trans", "getrefcount", "group_idx", "hasattr", "i", "identity", "if", "in", "ipa", "ipa_dropout", "is", "layer_norm_ipa", "layer_norm_s", "layer_norm_z", "linear_in", "lit_positions", "mask", "new_ones", "nn", "not", "num_blocks", "outputs", "pair", "pairwise_dim", "persistent", "positions", "pred_xyz", "preds", "quat", "quats", "r", "range", "register_buffer", "requires_grad", "residue_constants", "restype_atom14_mask", "restype_atom14_rigid_group_positions", "restype_atom14_to_rigid_group", "restype_rigid_group_default_frame", "return", "rigids", "rot_mats", "s", "s_initial", "scale_translation", "scaled_rigids", "self", "sequence_dim", "shape", "sidechain_frames", "single", "stack", "states", "stop_rot_gradient", "super", "sys", "tensor", "to", "to_tensor_4x4", "to_tensor_7", "torch", "torsion_angles_to_frames", "training", "trans_scale_factor", "transition", "unnormalized_angles", "z", "z_reference_list"], "esm/modeling_esmfold.py:EsmFoldingTrunk": ["C", "CA", "CB", "ContextManagers", "Embedding", "LayerNorm", "Linear", "ModelFoldRelativePosition", "ModelFoldStructureModule", "ModelFoldTriangularSelfAttentionBlock", "ModelFoldingTrunk", "Module", "ModuleList", "N", "None", "Number", "True", "ValueError", "_", "__init__", "a", "b", "be", "bins", "block", "blocks", "boundaries", "c", "c_s", "c_z", "chunk", "chunk_size", "class", "config", "coords", "cross", "def", "detach", "device", "dim", "distogram", "dists", "dtype", "else", "float", "for", "forward", "if", "in", "int64", "is", "keepdims", "linspace", "mask", "max_bin", "max_recycles", "min_bin", "must", "negative", "nn", "no_grad", "no_recycles", "not", "num_bins", "num_blocks", "of", "pair", "pair_feats", "pairwise_dim", "pairwise_positional_embedding", "pairwise_state_dim", "positions", "pow", "raise", "range", "recycle_bins", "recycle_disto", "recycle_idx", "recycle_s", "recycle_s_norm", "recycle_z", "recycle_z_norm", "recycles", "residue_index", "residx", "return", "s", "s_s", "s_s_0", "s_z", "s_z_0", "self", "seq_feats", "sequence_dim", "sequence_state_dim", "set_chunk_size", "shape", "single", "squeeze", "staticmethod", "structure", "structure_module", "sum", "super", "to", "torch", "true_aa", "trunk2sm_s", "trunk2sm_z", "trunk_iter", "weight", "with", "x", "z", "zero_", "zeros", "zeros_like"], "esm/modeling_esmfold.py:EsmForProteinFolding": ["B", "Embedding", "False", "L", "LayerNorm", "Linear", "Model", "ModelFoldStructureModule", "ModelFoldTriangularSelfAttentionBlock", "ModelFoldingTrunk", "ModelForProteinFolding", "ModelForProteinFoldingOutput", "ModelModel", "ModelPreTrainedModel", "Model_ablate_sequence", "Model_attns", "Model_dict_cls_idx", "Model_dict_eos_idx", "Model_dict_mask_idx", "Model_dict_padding_idx", "Model_feats", "Model_hidden_states", "Model_layers", "Model_reorder", "Model_s", "Model_s_combine", "Model_s_mlp", "Modelaa", "Modelfold_config", "None", "OFProtein", "Optional", "Parameter", "ReLU", "Sequential", "Tensor", "True", "Union", "__init__", "_af2_to_Model_from_vocab_list", "_can_record_outputs", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "aa", "aatype", "add_pooling_layer", "af2_idx_to_Model_idx", "af2_to_Model", "angles", "append", "arange", "argmax", "args", "assert", "atom14_atom_exists", "atom14_to_atom37", "atom37_atom_exists", "atom_mask", "atom_positions", "attention_mask", "auto_docstring", "b_factors", "bert_mask", "bins", "bool", "bos", "bosi", "bypass_lm", "c_s", "c_z", "cat", "categorical_lddt", "cfg", "class", "clone", "cls", "collate_dense_tensors", "compute_language_model_representations", "compute_predicted_aligned_error", "compute_tm", "config", "cpu", "def", "detach", "device", "dict", "dim", "disto_logits", "distogram_bins", "distogram_head", "distogram_logits", "dtype", "else", "embed_aa", "embedding", "eos", "eosi", "expand", "expand_as", "final_atom_mask", "final_atom_positions", "for", "forward", "fp16_Model", "frames", "from_numpy", "half", "hidden_size", "hidden_states", "i", "if", "in", "index", "infer", "infer_pdb", "infer_pdbs", "input_ids", "int", "is", "isinstance", "items", "k", "kwargs", "lddt_bins", "lddt_head", "lddt_head_hid_dim", "len", "list", "lm_head", "lm_logits", "lst", "make_atom14_masks", "map_unknown_to_x", "mapping", "mask", "mask_idx", "masked_aa", "masked_fill", "masking_pattern", "max_bin", "mlm_targets", "n_tokens_embed", "ndim", "new_Modelaa", "new_aa", "new_full", "new_ones", "new_zeros", "next", "nn", "no_bins", "no_grad", "no_recycles", "not", "num_attention_heads", "num_hidden_layers", "num_recycles", "numpy", "ones_like", "output", "output_hidden_states", "output_to_pdb", "pad", "pad_idx", "padding_idx", "pairwise_state_dim", "parameters", "pattern", "pdbs", "plddt", "position_ids", "positions", "pred", "pred_pos", "ptm", "ptm_head", "ptm_logits", "r", "range", "register_buffer", "requires_grad_", "reshape", "resid", "residue_constants", "residue_index", "restype_num", "restype_order_with_x", "restypes_with_x", "return", "s_s", "s_s_0", "s_z", "s_z_0", "self", "seq", "seqs", "sequence", "sequence_dim", "sequence_state_dim", "sequence_to_onehot", "shape", "sidechain_frames", "size", "softmax", "squeeze", "stack", "states", "staticmethod", "str", "structure", "structure_module", "structure_module_config", "sum", "super", "target", "tensor", "to", "to_pdb", "torch", "transpose", "trunk", "trunk_config", "unk_idx", "unnormalized_angles", "unsqueeze", "update", "v", "vocab_list", "zeros"], "esm/modeling_esm.py:rotate_half": ["Model_half", "cat", "chunk", "def", "dim", "return", "torch", "x", "x1", "x2"], "esm/modeling_esm.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "cos", "def", "return", "rotate_half", "shape", "sin", "x"], "esm/modeling_esm.py:gelu": ["Model", "def", "erf", "math", "return", "sqrt", "torch", "x"], "esm/modeling_esm.py:symmetrize": ["Make", "Model", "contact", "def", "dimensions", "final", "for", "in", "layer", "prediction", "return", "symmetric", "transpose", "two", "used", "x"], "esm/modeling_esm.py:average_product_correct": ["Model", "Model_product_correct", "Perform", "True", "a1", "a12", "a2", "avg", "contact", "correct", "def", "div_", "for", "keepdims", "normalized", "prediction", "product", "return", "sum", "used", "x"], "esm/modeling_esm.py:RotaryEmbedding": ["ModelEmbedding", "Module", "None", "Tensor", "__init__", "_cos_cached", "_seq_len_cached", "_sin_cached", "_update_cos_sin_tables", "apply_Model_pos_emb", "arange", "cat", "class", "cos", "def", "device", "dim", "dtype", "emb", "float", "forward", "freqs", "if", "int", "int64", "inv_freq", "k", "nn", "or", "outer", "q", "register_buffer", "return", "self", "seq_dimension", "seq_len", "shape", "sin", "super", "t", "to", "torch", "tuple", "type_as", "x"], "esm/modeling_esm.py:EsmContactPredictionHead": ["Linear", "ModelContactPredictionHead", "Module", "None", "Sigmoid", "True", "_", "__init__", "activation", "attentions", "average_product_correct", "batch_size", "bias", "class", "def", "device", "eos_idx", "eos_mask", "forward", "heads", "in_features", "int", "layers", "ne", "nn", "permute", "regression", "return", "self", "seqlen", "size", "squeeze", "super", "symmetrize", "to", "tokens", "unsqueeze", "view", "weight"], "esm/modeling_esm.py:EsmEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "and", "arange", "attention_mask", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "def", "device", "dropout", "dtype", "else", "emb_layer_norm_before", "embeddings", "eps", "expand", "float", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm", "layer_norm_eps", "long", "mask_ratio_observed", "mask_ratio_train", "mask_token_id", "masked_fill", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "sequence_length", "shape", "size", "src_lengths", "sum", "super", "to", "token_dropout", "torch", "unsqueeze", "vocab_size", "word_embeddings"], "esm/modeling_esm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "causal_mask", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "float", "float32", "functional", "hasattr", "head_mask", "if", "in", "is", "key", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "nn", "not", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "seq_length", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "esm/modeling_esm.py:EsmSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Embedding", "False", "FloatTensor", "Linear", "Model", "ModelSelfAttention", "Module", "None", "Optional", "RotaryEmbedding", "Set", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "batch_size", "class", "config", "contiguous", "current_states", "def", "dim", "distance_embedding", "does", "dropout", "eager", "eager_attention_forward", "elif", "else", "embedding_size", "embeddings", "encoder_attention_mask", "encoder_hidden_states", "explicitly", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "in", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "key", "key_layer", "kwargs", "layer_idx", "max_position_embeddings", "model", "multiple", "nn", "not", "num_attention_heads", "number", "of", "or", "position_embedding_type", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "rotary", "rotary_embeddings", "scaling", "self", "seq_length", "set_attn_implementation", "shape", "size", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "value", "value_layer", "view", "with"], "esm/modeling_esm.py:EsmSelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super"], "esm/modeling_esm.py:EsmAttention": ["False", "LayerNorm", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "TransformersKwargs", "Unpack", "_", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attn_output", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "eps", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "hidden_states_ln", "if", "index", "is_cross_attention", "key", "kwargs", "layer_idx", "layer_norm_eps", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "union", "value"], "esm/modeling_esm.py:EsmIntermediate": ["Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "forward", "gelu", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "esm/modeling_esm.py:EsmOutput": ["Dropout", "Linear", "ModelOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super"], "esm/modeling_esm.py:EsmLayer": ["AttributeError", "GradientCheckpointingLayer", "If", "LayerNorm", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "RuntimeError", "TransformersKwargs", "True", "Unpack", "__init__", "a", "add_cross_attention", "added", "and", "are", "as", "attention", "attention_mask", "attention_output", "attention_output_ln", "be", "by", "chunk_size_feed_forward", "class", "config", "cross", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "eps", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_cross_attention", "is_decoder", "kwargs", "layer_norm_eps", "layer_output", "layers", "model", "nn", "not", "output", "passed", "raise", "return", "self", "seq_len_dim", "setting", "should", "super", "to", "used", "with"], "esm/modeling_esm.py:EsmEncoder": ["BaseModelOutputWithCrossAttentions", "False", "LayerNorm", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "else", "emb_layer_norm_after", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "eps", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_norm_eps", "nn", "not", "num_hidden_layers", "range", "return", "self", "super"], "esm/modeling_esm.py:EsmPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "esm/modeling_esm.py:EsmPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelFoldTriangularSelfAttentionBlock", "ModelLMHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "OutputRecorder", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention", "attentions", "base_model_prefix", "bias", "class", "config", "cross_attentions", "crossattention", "data", "def", "elif", "fill_", "get_output_embeddings", "hidden_states", "if", "index", "initializer_range", "is", "isinstance", "layer_name", "mean", "module", "nn", "normal_", "not", "padding_idx", "position_embeddings", "return", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "esm/modeling_esm.py:EsmModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "ModelContactPredictionHead", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_attn_implementation", "_prune_heads", "add_pooling_layer", "and", "attention", "attention_mask", "attentions", "attns", "auto_docstring", "batch_size", "bias", "check_model_inputs", "class", "config", "contact_head", "def", "device", "dim", "else", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "exactly", "flash_attention_2", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "in_features", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "items", "kwargs", "last_hidden_state", "layer", "must", "not", "num_attention_heads", "num_hidden_layers", "of", "one", "ones", "or", "output_attentions", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "predict_contacts", "prune_heads", "r", "raise", "return", "return_dict", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "stack", "super", "tokens", "torch", "tuple", "unsqueeze", "value", "word_embeddings"], "esm/modeling_esm.py:EsmForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "can_return_tuple", "class", "config", "decoder", "def", "device", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "predict_contacts", "prediction_scores", "r", "return", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "tokens", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "esm/modeling_esm.py:EsmLMHead": ["False", "LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "bias", "class", "config", "decoder", "def", "dense", "eps", "features", "forward", "gelu", "hidden_size", "kwargs", "layer_norm", "layer_norm_eps", "nn", "return", "self", "super", "torch", "vocab_size", "x", "zeros"], "esm/modeling_esm.py:EsmForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "torch", "tuple", "view"], "esm/modeling_esm.py:EsmForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dropout", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "self", "sequence_output", "super", "to", "torch", "tuple", "view"], "esm/modeling_esm.py:EsmClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_dropout_prob", "hidden_size", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "esm/modeling_esm.py:create_position_ids_from_input_ids": ["Model_position_ids_from_input_ids", "cumsum", "def", "dim", "incremental_indices", "input_ids", "int", "long", "mask", "ne", "padding_idx", "return", "torch", "type_as"], "vilt/modeling_vilt.py:ViltForImagesAndTextClassificationOutput": ["FloatTensor", "ModelForImagesAndTextClassificationOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "r", "torch", "tuple"], "vilt/modeling_vilt.py:ViltEmbeddings": ["Dropout", "Embedding", "False", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Parameter", "TextEmbeddings", "True", "_", "__init__", "align_corners", "append", "arange", "as_tuple", "attention_mask", "batch_size", "bilinear", "cat", "class", "cls_token", "cls_tokens", "config", "def", "device", "dim", "dropout", "dtype", "effective_resolution", "else", "embeddings", "enumerate", "expand", "flatten", "float", "for", "forward", "full_like", "functional", "h", "height", "hidden_dropout_prob", "hidden_size", "i", "if", "ij", "image_embeds", "image_masks", "image_size", "image_token_type_idx", "in", "indexing", "input_ids", "inputs_embeds", "int", "interpolate", "is", "isinstance", "long", "masks", "max", "max_image_length", "meshgrid", "min", "modality_type_vocab_size", "mode", "multinomial", "nn", "non_valid_idx", "non_valid_nums", "non_valid_row_idx", "nonzero", "not", "num_channels", "num_patches", "nv", "ones", "or", "p", "pad", "pad_choice", "pad_nums", "patch_dim", "patch_embeddings", "patch_index", "patch_size", "ph", "pixel_mask", "pixel_values", "pos_embed", "position_embeddings", "projection", "pw", "replacement", "return", "select", "self", "shape", "size", "spatial_pos", "stack", "sum", "super", "text_embeddings", "text_embeds", "to", "token_type_embeddings", "token_type_ids", "torch", "transpose", "u", "unique", "unique_rows", "v", "valid_choice", "valid_idx", "valid_nums", "valid_row_idx", "view", "visual_embed", "w", "weight", "width", "x", "x_h", "x_mask", "x_w", "zeros", "zeros_like", "zip"], "vilt/modeling_vilt.py:TextEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "vilt/modeling_vilt.py:ViltPatchEmbeddings": ["Conv2d", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "dtype", "else", "forward", "height", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "target_dtype", "that", "the", "to", "values", "weight", "width", "with", "x"], "vilt/modeling_vilt.py:ViltSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "Softmax", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "sqrt", "super", "the", "torch", "transpose", "value", "value_layer", "view"], "vilt/modeling_vilt.py:ViltSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "vilt/modeling_vilt.py:ViltAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "vilt/modeling_vilt.py:ViltIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "vilt/modeling_vilt.py:ViltOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "vilt/modeling_vilt.py:ViltLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "device", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "to"], "vilt/modeling_vilt.py:ViltEncoder": ["BaseModelOutput", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "vilt/modeling_vilt.py:ViltPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "vilt/modeling_vilt.py:ViltModel": ["BaseModelOutputWithPooling", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch", "bool", "both", "cannot", "class", "config", "def", "device", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "image", "image_batch_size", "image_embeds", "image_size", "image_token_type_idx", "in", "input_ids", "input_shape", "inputs", "inputs_embeds", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "need", "nn", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pixel_mask", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "text", "text_batch_size", "text_embeddings", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings"], "vilt/modeling_vilt.py:ViltPooler": ["Linear", "ModelPooler", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super"], "vilt/modeling_vilt.py:ViltForMaskedLM": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelMLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "_", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "decoder", "def", "device", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "image_embeds", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "mlm_logits", "mlm_score", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "pooled_output", "post_init", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "shape", "super", "text_features", "text_seq_len", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "vilt/modeling_vilt.py:ViltPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "vilt/modeling_vilt.py:ViltMLMHead": ["False", "Linear", "ModelMLMHead", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "if", "is", "nn", "not", "return", "self", "super", "torch", "transform", "vocab_size", "weight", "x", "zeros"], "vilt/modeling_vilt.py:ViltForQuestionAnswering": ["FloatTensor", "GELU", "LayerNorm", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Sequential", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "binary_cross_entropy_with_logits", "bool", "class", "classifier", "config", "def", "device", "else", "forward", "functional", "head_mask", "hidden_size", "hidden_states", "if", "image_embeds", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "pooler_output", "post_init", "r", "return", "return_dict", "self", "shape", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict"], "vilt/modeling_vilt.py:ViltForImageAndTextRetrieval": ["FloatTensor", "Linear", "LongTensor", "Model", "ModelForImageAndTextRetrieval", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "SequenceClassifierOutput", "Training", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "image_embeds", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "pooler_output", "post_init", "r", "raise", "rank_output", "return", "return_dict", "self", "super", "supported", "token_type_ids", "torch", "tuple", "use_return_dict", "yet"], "vilt/modeling_vilt.py:ViltForImagesAndTextClassification": ["CrossEntropyLoss", "FloatTensor", "GELU", "LayerNorm", "Linear", "LongTensor", "Make", "Model", "ModelForImagesAndTextClassification", "ModelForImagesAndTextClassificationOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Sequential", "Union", "ValueError", "__init__", "and", "append", "attention_mask", "attentions", "auto_docstring", "bool", "cat", "class", "classifier", "config", "def", "device", "dim", "else", "for", "forward", "head_mask", "hidden_size", "hidden_states", "i", "if", "image_embeds", "image_token_type_idx", "images", "in", "input", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "match", "model", "ndim", "nn", "not", "num_images", "num_labels", "number", "of", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "pooled_output", "pooler_output", "pooler_outputs", "post_init", "r", "raise", "range", "return", "return_dict", "self", "shape", "super", "sure", "the", "to", "token_type_ids", "torch", "tuple", "unsqueeze", "use_return_dict", "view", "with"], "vilt/modeling_vilt.py:ViltForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image_embeds", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "shape", "super", "text_input_size", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaCache": ["LongTensor", "ModelCache", "None", "PretrainedConfig", "Tensor", "True", "Union", "_", "__init__", "_dtype", "_dynamo", "append", "cache_position", "clamp", "class", "config", "conv_kernel", "conv_kernel_size", "conv_state", "conv_states", "def", "device", "dims", "dtype", "else", "float16", "for", "if", "in", "int", "intermediate_size", "is", "is_compileable", "layer_idx", "len", "list", "mark_static_address", "max_batch_size", "new_conv_state", "new_ssm_state", "not", "num_hidden_layers", "range", "reset", "return", "roll", "self", "shifts", "ssm_state", "ssm_state_size", "ssm_states", "state_size", "str", "to", "torch", "update_conv_state", "update_ssm_state", "zero_", "zeros"], "falcon_mamba/modeling_falcon_mamba.py:_lazy_load_causal_conv1d": ["None", "_causal_conv1d_cache", "_causal_conv1d_kernel", "_lazy_load_causal_conv1d", "causal", "causal_conv1d_fn", "causal_conv1d_update", "community", "conv1d", "def", "elif", "else", "get_kernel", "global", "if", "is", "is_causal_conv1d_available", "is_kernels_available", "kernels", "not", "return"], "falcon_mamba/modeling_falcon_mamba.py:rms_forward": ["Model_forward", "True", "def", "dtype", "float32", "hidden_states", "input_dtype", "keepdim", "mean", "pow", "return", "rsqrt", "to", "torch", "variance", "variance_epsilon"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaMixer": ["A", "ACT2FN", "AILab", "A_log", "B", "C", "Conv1d", "D", "Dao", "Falling", "False", "For", "ImportError", "Linear", "LongTensor", "Mamba", "ModelCache", "ModelConfig", "ModelMixer", "Module", "None", "Optional", "Parameter", "Tensor", "The", "To", "True", "_", "__init__", "_dynamo", "_lazy_load_causal_conv1d", "_pre_quantization_dtype", "act", "activation", "all", "alxndrTL", "and", "append", "arange", "as", "attention_mask", "available", "b_c_dt_rms_eps", "b_c_rms", "b_rms_weight", "back", "backend", "batch_size", "because", "bias", "but", "c_rms_weight", "cache_params", "cache_position", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk", "class", "clone", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_kernel", "conv_kernel_size", "conv_state", "conv_states", "conv_weights", "cuda", "cuda_kernels_forward", "def", "deltaB_u", "delta_bias", "delta_softplus", "device", "dim", "discrete_A", "discrete_B", "discrete_time_step", "dt_proj", "dt_rms", "dt_rms_weight", "dt_softplus", "dtype", "else", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "hasattr", "hidden_act", "hidden_size", "hidden_states", "hs", "https", "i", "if", "implementation", "in", "in_channels", "in_proj", "input_states", "install", "installed", "int", "intermediate_size", "is", "is_compiling", "is_fast_path_available", "is_mambapy_available", "it", "kernel_size", "kernels", "layer_idx", "log", "logger", "mamba", "mamba_inner_fn", "mambapy", "matmul", "mixer_rms_eps", "nn", "not", "of", "one", "ones", "or", "out_channels", "out_proj", "package", "pad", "padding", "path", "persistent", "pip", "projected_states", "pscan", "py", "raise", "range", "register_buffer", "requires_grad", "return", "return_last_state", "rms_eps", "rms_forward", "scan_output", "scan_outputs", "selective_scan_fn", "selective_state_update", "self", "seq_len", "sequential", "set", "shape", "size", "slow_forward", "softplus", "spaces", "split", "squeeze", "ssm_parameters", "ssm_state", "ssm_state_size", "ssm_states", "stack", "state", "state_size", "sum", "super", "the", "time_proj_bias", "time_step", "time_step_rank", "to", "torch", "training", "transpose", "type", "unsqueeze", "update_conv_state", "update_ssm_state", "use_Modelpy", "use_bias", "use_conv_bias", "use_mambapy", "variance_epsilon", "view", "warn_slow_implementation", "warning_once", "weight", "x_proj", "zeros"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "__init__", "class", "def", "device", "eps", "extra_repr", "f", "forward", "hidden_size", "hidden_states", "nn", "ones", "return", "rms_forward", "self", "shape", "super", "to", "torch", "variance_epsilon", "weight"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaBlock": ["GradientCheckpointingLayer", "LongTensor", "ModelBlock", "ModelCache", "ModelMixer", "ModelRMSNorm", "None", "Optional", "__init__", "attention_mask", "cache_params", "cache_position", "class", "config", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "if", "layer_idx", "layer_norm_epsilon", "mixer", "norm", "residual", "residual_in_fp32", "return", "self", "super", "to", "torch", "weight"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaPreTrainedModel": ["A", "A_log", "D", "Embedding", "False", "Linear", "ModelBlock", "ModelConfig", "ModelMixer", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_reinit", "_no_split_modules", "a", "arange", "backbone", "base_model_prefix", "bias", "clamp", "class", "config", "constant", "constant_", "contiguous", "conv1d", "copy_", "data", "def", "dt", "dt_init_std", "dt_proj", "dtype", "elif", "exp", "expand", "expm1", "fill_", "float32", "getattr", "if", "init", "initializer_range", "intermediate_size", "inv_dt", "is", "isinstance", "kaiming_uniform_", "log", "math", "min", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "p", "rand", "random", "rescale_prenorm_residual", "self", "sqrt", "ssm_state_size", "std", "supports_gradient_checkpointing", "time_step_floor", "time_step_init_scheme", "time_step_max", "time_step_min", "time_step_rank", "time_step_scale", "torch", "uniform_", "weight", "zeros_"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaOutput": ["FloatTensor", "ModelCache", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "last_hidden_state", "r", "torch", "tuple"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaCausalLMOutput": ["FloatTensor", "ModelCache", "ModelCausalLMOutput", "ModelOutput", "None", "Optional", "cache_params", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaModel": ["Embedding", "False", "LongTensor", "ModelBlock", "ModelCache", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "a", "all_hidden_states", "and", "arange", "are", "attention_mask", "auto_docstring", "automatically", "be", "because", "bool", "cache_params", "cache_position", "case", "class", "config", "conv_kernel", "def", "device", "don", "dtype", "elif", "else", "embeddings", "eps", "exactly", "for", "forward", "get_input_embeddings", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "initialized", "input_ids", "inputs_embeds", "is", "it", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layers", "manually", "mixer_block", "must", "new_embeddings", "nn", "norm_f", "not", "num_hidden_layers", "of", "one", "or", "output_hidden_states", "pass", "passed", "post_init", "prefilling", "r", "raise", "range", "return", "return_dict", "self", "set_input_embeddings", "size", "specify", "stage", "super", "t", "that", "the", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "vocab_size", "when", "will", "you"], "falcon_mamba/modeling_falcon_mamba.py:FalconMambaForCausalLM": ["Any", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCache", "ModelCausalLMOutput", "ModelForCausalLM", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "_update_model_kwargs_for_generation", "and", "arange", "attention_mask", "auto_docstring", "backbone", "bias", "bool", "cache_params", "cache_position", "cat", "class", "config", "contiguous", "conv_kernel", "def", "device", "dict", "dim", "dtype", "else", "float", "for", "forward", "get", "get_input_embeddings", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "items", "key", "kwargs", "labels", "lm_head", "logits", "loss", "loss_fct", "max_batch_size", "model_inputs", "model_kwargs", "new_embeddings", "new_ones", "nn", "not", "num_new_tokens", "output", "output_hidden_states", "outputs", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "set_input_embeddings", "shape", "shift_labels", "shift_logits", "size", "str", "super", "to", "torch", "tuple", "unsqueeze", "update", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "switch_transformers/modeling_switch_transformers.py:router_z_loss_func": ["Model_logits", "Model_z_loss_func", "Tensor", "_", "def", "dim", "float", "log_z", "logsumexp", "num_groups", "r", "return", "shape", "sum", "tokens_per_group", "torch", "z_loss"], "switch_transformers/modeling_switch_transformers.py:load_balancing_loss_func": ["Model_balancing_loss_func", "Tensor", "axis", "def", "dtype", "expert_indices", "expert_mask", "float", "float32", "functional", "if", "int64", "len", "max", "mean", "nn", "num_experts", "one_hot", "r", "return", "router_prob_per_group_and_expert", "router_probs", "shape", "to", "tokens_per_group_and_expert", "torch", "unsqueeze", "values"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersTop1Router": ["CB", "Linear", "ModelConfig", "ModelTop1Router", "Module", "SCB", "Tensor", "__init__", "_cast_classifier", "_compute_router_probabilities", "and", "argmax", "bias", "class", "classifier", "config", "cumsum", "def", "dim", "dtype", "empty_like", "expert_capacity", "expert_capacity_mask", "expert_index", "forward", "functional", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "ignore_padding_tokens", "input_dtype", "jitter_noise", "max", "nn", "not", "num_classes", "num_experts", "one_hot", "or", "r", "return", "router_bias", "router_dtype", "router_ignore_padding_tokens", "router_jitter_noise", "router_logits", "router_probabilities", "router_probs", "self", "softmax", "super", "to", "token_priority", "torch", "training", "tuple", "uniform_", "unsqueeze", "values"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersLayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bfloat16", "class", "def", "dtype", "eps", "float16", "float32", "forward", "hidden_size", "hidden_states", "if", "in", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersDenseActDense": ["ACT2FN", "Dropout", "False", "Linear", "ModelConfig", "ModelDenseActDense", "Module", "Tensor", "__init__", "act", "and", "bias", "class", "config", "d_ff", "d_model", "def", "dense_act_fn", "dropout", "dropout_rate", "dtype", "forward", "hidden_states", "if", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "weight", "wi", "wo"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersSparseMLP": ["ModelConfig", "ModelDenseActDense", "ModelSparseMLP", "ModelTop1Router", "Module", "ModuleDict", "True", "__init__", "argmax", "as_tuple", "batch_size", "bool", "class", "config", "def", "device", "dim", "dtype", "expert_", "expert_class", "expert_index", "experts", "f", "for", "forward", "getattr", "hidden_states", "idx", "idx_mask", "in", "next_states", "nn", "nonzero", "num_experts", "r", "range", "reshape", "return", "router", "router_logits", "router_mask", "router_probs", "self", "seq_len", "shape", "sum", "super", "tolist", "torch", "zeros"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersLayerFF": ["Dropout", "False", "ModelConfig", "ModelDenseActDense", "ModelLayerFF", "ModelLayerNorm", "ModelSparseMLP", "Module", "None", "__init__", "and", "class", "config", "d_model", "def", "dropout", "dropout_rate", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is", "is_sparse", "isinstance", "layer_norm", "layer_norm_epsilon", "mlp", "nn", "not", "output", "output_router_logits", "r", "return", "router_tuple", "self", "super", "tuple"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersAttention": ["Embedding", "EncoderDecoderCache", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "True", "__class__", "__init__", "__name__", "_relative_position_bucket", "a", "abs", "and", "arange", "attn_output", "attn_weights", "batch_size", "bias", "bidirectional", "bool", "cache_position", "caching", "call", "causal_mask", "class", "compute_bias", "config", "context_position", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "d_kv", "d_model", "decoder", "def", "deprecate_kwarg", "device", "dim", "dropout", "dropout_rate", "dtype", "during", "else", "errors", "f", "find_pruneable_heads_and_indices", "float", "forward", "full_like", "functional", "get", "gradient_checkpointing", "has_relative_attention_bias", "heads", "hidden_states", "if", "index", "inner_dim", "int", "is", "is_cross_attention", "is_decoder", "is_small", "is_updated", "isinstance", "k", "key_length", "key_states", "key_value_proj_dim", "key_value_states", "keys", "layer_head_mask", "layer_idx", "layers", "len", "list", "log", "logger", "long", "make", "mask", "math", "matmul", "max_distance", "max_exact", "memory_position", "min", "n_heads", "new_name", "nn", "not", "num_buckets", "num_heads", "o", "ones", "output_attentions", "outputs", "p", "passing", "past_key_value", "past_key_values", "permute", "position_bias", "position_bias_masked", "provide", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "query_length", "query_states", "real_seq_length", "recommended", "relative_attention_bias", "relative_attention_max_distance", "relative_attention_num_buckets", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "requires_grad", "return", "scores", "self", "self_attention_cache", "seq_length", "set", "shape", "softmax", "staticmethod", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "type_as", "union", "unsqueeze", "update", "use_cache", "used", "v", "value_states", "values", "version", "view", "warning_once", "weight", "when", "where", "will", "without", "zeros", "zeros_like"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersLayerSelfAttention": ["Dropout", "False", "ModelAttention", "ModelLayerNorm", "ModelLayerSelfAttention", "Module", "None", "Optional", "SelfAttention", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "return", "self", "super", "use_cache", "version"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersLayerCrossAttention": ["Dropout", "EncDecAttention", "False", "ModelAttention", "ModelLayerCrossAttention", "ModelLayerNorm", "Module", "None", "Optional", "__init__", "attention_mask", "attention_output", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dropout_rate", "eps", "forward", "has_relative_attention_bias", "hidden_states", "int", "key_value_states", "layer_head_mask", "layer_idx", "layer_norm", "layer_norm_epsilon", "layer_output", "mask", "new_name", "nn", "normed_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_bias", "query_length", "return", "self", "super", "use_cache", "version"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersBlock": ["False", "GradientCheckpointingLayer", "ModelBlock", "ModelLayerCrossAttention", "ModelLayerFF", "ModelLayerSelfAttention", "ModuleList", "None", "Optional", "True", "__init__", "and", "any", "append", "attention_mask", "attention_outputs", "cache_position", "clamp", "clamp_value", "class", "config", "cross_attention_outputs", "cross_attn_layer_head_mask", "def", "device", "do_cross_attention", "dtype", "else", "encoder_attention_mask", "encoder_decoder_position_bias", "encoder_hidden_states", "finfo", "float16", "forward", "has_relative_attention_bias", "hidden_states", "if", "int", "int64", "is", "is_decoder", "is_sparse", "isinf", "isinstance", "key_value_states", "layer", "layer_head_mask", "layer_idx", "max", "min", "nn", "not", "output_attentions", "output_router_logits", "outputs", "past_key_values", "position_bias", "query_length", "return", "return_dict", "router_tuple", "self", "self_attention_outputs", "super", "torch", "tuple", "use_cache", "zeros"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersPreTrainedModel": ["DUMMY_INPUTS", "DUMMY_MASK", "False", "In", "Model", "ModelAttention", "ModelBlock", "ModelConfig", "ModelDenseActDense", "ModelEncoderModel", "ModelForConditionalGeneration", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "ModelSparseMLP", "None", "PreTrainedModel", "See", "True", "ValueError", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_shift_right", "and", "base_model_prefix", "be", "bias", "cat", "class", "classifier", "clone", "config", "d_ff", "d_kv", "d_model", "data", "decoder_attention_mask", "decoder_input_ids", "decoder_start_token_id", "def", "defined", "dim", "docs", "dummy_inputs", "elif", "else", "expert_", "experts", "f", "factor", "fill_", "for", "full", "has", "has_relative_attention_bias", "hasattr", "idx", "if", "in", "information", "initializer_factor", "input_ids", "input_mask", "is", "is_torch_fx_proxy", "isinstance", "it", "k", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "normal_", "not", "num_experts", "num_heads", "o", "pad_token_id", "property", "q", "raise", "range", "relative_attention_bias", "return", "router", "self", "set", "shape", "shared", "shifted_input_ids", "std", "supports_gradient_checkpointing", "tensor", "the", "tie_word_embeddings", "to", "torch", "usually", "v", "weight", "wi", "wo", "zero_"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersStack": ["AttentionMaskConverter", "BlockMask", "Cache", "Dropout", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "MoEModelOutputWithPastAndCrossAttentions", "ModelBlock", "ModelLayerNorm", "ModelPreTrainedModel", "ModelStack", "ModuleList", "None", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "a", "all_attentions", "all_cross_attentions", "all_hidden_states", "all_router_probs", "and", "any", "append", "arange", "as", "at", "attention_mask", "attentions", "batch_size", "be", "block", "bool", "both", "cache_position", "can", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "cuda", "d_model", "decoder", "decoder_", "decoder_sparse_step", "def", "device", "device_map", "diagonal", "dim", "dropout", "dropout_rate", "dtype", "either", "elif", "else", "embed_tokens", "embeddings", "encoder_attention_mask", "encoder_batch_size", "encoder_decoder_position_bias", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_sequence_length", "encoder_sparse_step", "enumerate", "eps", "err_msg_prefix", "expand", "f", "fill_value", "final_layer_norm", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_relative_attention_bias", "have", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "initialize", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "invert_attention_mask", "is", "is_compileable", "is_decoder", "is_encoder_decoder", "is_sparse", "is_torchdynamo_compiling", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_module", "layer_norm_epsilon", "layer_outputs", "logger", "make_flex_block_causal_mask", "mask_length", "mask_seq_length", "masked_fill", "min", "min_dtype", "model", "new_embeddings", "nn", "not", "npu", "num_decoder_layers", "num_layers", "ones", "only", "or", "output_attentions", "output_hidden_states", "output_router_logits", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_bias", "post_init", "raise", "range", "reshape", "return", "return_dict", "router_probs", "same", "sdpa", "self", "self_attention_cache", "seq_length", "sequence_length", "set", "set_input_embeddings", "shape", "size", "sparse_step", "specify", "staticmethod", "super", "target_length", "the", "time", "to", "token", "torch", "training", "triu", "tuple", "type", "use_cache", "use_return_dict", "used", "using_compilable_cache", "v", "valid", "view", "vocab_size", "warning_once", "weight", "with", "xpu"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersModel": ["BoolTensor", "Cache", "Embedding", "False", "FloatTensor", "FutureWarning", "Layers", "LongTensor", "MLP", "MoEModelOutput", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Seq2SeqMoEModelOutput", "Set", "Tensor", "True", "Union", "ValueError", "You", "__HEAD_MASK_WARNING_MSG", "__init__", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "any", "asked", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "but", "cache_position", "class", "config", "contain", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_router_logits", "deepcopy", "def", "dense", "device_map", "does", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "encoder_router_logits", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "items", "last_hidden_state", "layer", "len", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "num_sparse_encoder_layers", "output_attentions", "output_hidden_states", "output_router_logits", "past_key_values", "post_init", "prune_heads", "r", "raise", "restart", "return", "return_dict", "router_probs", "self", "set_input_embeddings", "shared", "sparse", "super", "the", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "transformer", "tuple", "use_cache", "use_return_dict", "vocab_size", "warn", "warnings", "weight"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersForConditionalGeneration": ["BoolTensor", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "FutureWarning", "GenerationMixin", "Linear", "LongTensor", "MoEModelOutput", "ModelConfig", "ModelForConditionalGeneration", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "Seq2SeqMoEOutput", "Softmax", "Tensor", "True", "Union", "__HEAD_MASK_WARNING_MSG", "__init__", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "_unpack_router_logits", "and", "append", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "cat", "class", "config", "copy", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_aux_loss", "decoder_config", "decoder_expert_indexes", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_router_logits", "decoder_router_probs", "decoder_sparse_step", "decoder_z_loss", "deepcopy", "def", "device", "device_map", "dim", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_aux_loss", "encoder_config", "encoder_expert_indexes", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "encoder_router_logits", "encoder_router_probs", "encoder_sparse_step", "encoder_z_loss", "expert_indexes", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "is", "is_decoder", "isinstance", "labels", "last_hidden_state", "len", "lm_head", "lm_logits", "load_balancing_loss_func", "logits", "loss", "loss_fct", "model_dim", "new_embeddings", "nn", "not", "num_decoder_layers", "num_layers", "output", "output_attentions", "output_hidden_states", "output_router_logits", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "router_output", "router_outputs", "router_probs", "router_z_loss_coef", "router_z_loss_func", "self", "sequence_output", "set_input_embeddings", "shape", "shared", "size", "super", "tie_encoder_decoder", "tie_word_embeddings", "to", "torch", "total_expert_indexes", "total_router_logits", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warn", "warnings", "weight", "z_loss"], "switch_transformers/modeling_switch_transformers.py:SwitchTransformersEncoderModel": ["Embedding", "False", "FloatTensor", "LongTensor", "MoEModelOutput", "ModelConfig", "ModelEncoderModel", "ModelPreTrainedModel", "ModelStack", "None", "Optional", "SelfAttention", "True", "Union", "__init__", "_prune_heads", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "attention_mask", "auto_docstring", "block", "bool", "class", "config", "copy", "d_model", "deepcopy", "def", "device_map", "else", "embed_tokens", "encoder", "encoder_config", "encoder_outputs", "for", "forward", "get_encoder", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "if", "in", "input_ids", "inputs_embeds", "is", "is_encoder_decoder", "items", "layer", "new_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "output_router_logits", "post_init", "prune_heads", "r", "return", "return_dict", "self", "set_input_embeddings", "shared", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "dpr/modeling_dpr.py:DPRContextEncoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "pooler_output", "r", "torch", "tuple"], "dpr/modeling_dpr.py:DPRQuestionEncoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "pooler_output", "r", "torch", "tuple"], "dpr/modeling_dpr.py:DPRReaderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "end_logits", "hidden_states", "r", "relevance_logits", "start_logits", "torch", "tuple"], "dpr/modeling_dpr.py:DPRPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "None", "PreTrainedModel", "True", "_init_weights", "_supports_sdpa", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "dpr/modeling_dpr.py:DPREncoder": ["BaseModelOutputWithPooling", "BertModel", "Encoder", "False", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "__init__", "add_pooling_layer", "attention_mask", "attentions", "base_model_prefix", "be", "bert_model", "bool", "can", "class", "config", "def", "embeddings_size", "encode_proj", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "last_hidden_state", "nn", "not", "out_features", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "pooler_output", "post_init", "projection_dim", "property", "raise", "return", "return_dict", "self", "sequence_output", "super", "t", "token_type_ids", "tuple", "zero"], "dpr/modeling_dpr.py:DPRSpanPredictor": ["False", "Linear", "Model", "ModelConfig", "ModelEncoder", "ModelPreTrainedModel", "ModelReaderOutput", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "base_model_prefix", "bool", "class", "config", "contiguous", "def", "dim", "else", "embeddings_size", "encoder", "end_logits", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "logits", "n_passages", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "post_init", "qa_classifier", "qa_outputs", "relevance_logits", "return", "return_dict", "self", "sequence_length", "sequence_output", "size", "split", "squeeze", "start_logits", "super", "tuple", "view"], "dpr/modeling_dpr.py:DPRPretrainedContextEncoder": ["Model", "ModelConfig", "ModelPreTrainedModel", "base_model_prefix", "class", "config", "ctx_encoder"], "dpr/modeling_dpr.py:DPRPretrainedQuestionEncoder": ["Model", "ModelConfig", "ModelPreTrainedModel", "base_model_prefix", "class", "config", "question_encoder"], "dpr/modeling_dpr.py:DPRPretrainedReader": ["Model", "ModelConfig", "ModelPreTrainedModel", "base_model_prefix", "class", "config", "span_predictor"], "dpr/modeling_dpr.py:DPRContextEncoder": ["Model", "ModelConfig", "ModelEncoder", "ModelOutput", "ModelPretrainedContextEncoder", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "and", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "ctx_encoder", "def", "device", "dtype", "either", "elif", "else", "forward", "have", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "long", "not", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "pooler_output", "post_init", "r", "raise", "return", "return_dict", "same", "self", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "zeros"], "dpr/modeling_dpr.py:DPRQuestionEncoder": ["Model", "ModelConfig", "ModelEncoder", "ModelOutput", "ModelPretrainedQuestionEncoder", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "and", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "forward", "have", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "long", "not", "ones", "or", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "pooler_output", "post_init", "question_encoder", "r", "raise", "return", "return_dict", "same", "self", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "zeros"], "dpr/modeling_dpr.py:DPRReader": ["Model", "ModelConfig", "ModelOutput", "ModelPretrainedReader", "ModelSpanPredictor", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "and", "at", "attention_mask", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "either", "elif", "else", "forward", "have", "if", "input_ids", "input_shape", "inputs_embeds", "is", "not", "ones", "or", "output_attentions", "output_hidden_states", "post_init", "r", "raise", "return", "return_dict", "same", "self", "size", "span_predictor", "specify", "super", "the", "time", "to", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2MoEGate": ["F", "False", "ModelConfig", "ModelMoEGate", "Module", "None", "Parameter", "Tensor", "__init__", "alpha", "aux_loss_alpha", "batch_size", "bool", "class", "config", "def", "dim", "dtype", "elif", "empty", "expand", "float32", "forward", "gating_dim", "greedy", "group_idx", "group_limited_greedy", "group_mask", "group_scores", "hidden_dim", "hidden_size", "hidden_states", "if", "k", "linear", "logits", "masked_fill", "max", "n_group", "n_routed_experts", "nn", "norm_topk_prob", "num_experts", "num_experts_per_tok", "num_group", "reshape", "return", "routed_scaling_factor", "scatter_", "score_mask", "scores", "self", "seq_aux", "seq_len", "shape", "softmax", "sorted", "super", "tmp_scores", "top_k", "topk", "topk_group", "topk_idx", "topk_method", "topk_weight", "torch", "type", "unsqueeze", "values", "view", "weight", "zeros_like"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2MoE": ["ModelConfig", "ModelMLP", "ModelMoE", "ModelMoEGate", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "append", "argsort", "cat", "class", "cnts", "config", "continue", "def", "dim", "dtype", "else", "empty_like", "end_idx", "enumerate", "ep_rank", "expert", "expert_out", "experts", "experts_per_rank", "for", "forward", "gate", "hidden_states", "i", "if", "in", "indices", "intermediate_size", "is", "len", "moe", "moe_intermediate_size", "mul_", "n_routed_experts", "n_shared_experts", "new_empty", "new_x", "new_zeros", "nn", "not", "num_experts_per_tok", "num_tokens", "orig_shape", "outputs", "outs", "range", "residuals", "return", "scatter_", "self", "shape", "shared_experts", "sorted_tokens", "start_idx", "sum", "super", "tokens_for_this_expert", "tokens_per_expert", "topk_ids", "topk_indices", "topk_weight", "topk_weights", "torch", "type", "unsqueeze", "view"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2MLP": ["ACT2FN", "Linear", "ModelConfig", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "class", "config", "cpu", "def", "default", "device", "device_type", "dynamic_rope_update", "else", "enabled", "expand", "float", "forward", "freqs", "freqs_cis", "get", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "ones_like", "original_inv_freq", "original_max_seq_len", "persistent", "polar", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "deepseek_v2/modeling_deepseek_v2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "deepseek_v2/modeling_deepseek_v2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "deepseek_v2/modeling_deepseek_v2.py:apply_rotary_emb": ["Model_rotary_emb", "Tensor", "def", "device", "flatten", "float", "freqs_cis", "reshape", "return", "shape", "to", "torch", "tuple", "type_as", "unsqueeze", "view_as_complex", "view_as_real", "xk", "xk_", "xk_out", "xq", "xq_", "xq_out"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "F", "False", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Passing", "Please", "Tensor", "True", "__init__", "_attn_implementation", "and", "apply_rotary_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "cache_kwargs", "cache_position", "cat", "class", "compressed_kv", "config", "contiguous", "def", "deprecate_kwarg", "deprecated", "device", "dim", "dropout", "eager", "eager_attention_forward", "else", "expand", "flash_attention_2", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "instead", "int", "is", "is_causal", "k_nope", "k_pe", "key_shape", "key_states", "kv_a_layernorm", "kv_a_proj_with_mqa", "kv_b_proj", "kv_lora_rank", "kwargs", "layer_idx", "make", "max_position_embeddings", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "pad", "padding_mask", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q", "q_a_layernorm", "q_a_proj", "q_b_proj", "q_lora_rank", "q_nope", "q_pe", "q_proj", "qk_head_dim", "qk_nope_head_dim", "qk_rope_head_dim", "query_shape", "query_states", "removed", "reshape", "return", "rope_theta", "scaling", "self", "seq_length", "shape", "split", "super", "sure", "to", "torch", "training", "transpose", "tuple", "update", "use", "v4", "v_head_dim", "value_states", "version", "view", "warn", "warnings", "will"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "first_k_dense_replace", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2PreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMoEGate", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "deepseek_v2/modeling_deepseek_v2.py:DeepseekV2ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "informer/modeling_informer.py:InformerFeatureEmbedder": ["Embedding", "ModelFeatureEmbedder", "Module", "ModuleList", "None", "Tensor", "__init__", "c", "cardinalities", "cat", "cat_feature_slice", "cat_feature_slices", "chunk", "class", "d", "def", "dim", "else", "embed", "embedders", "embedding_dims", "features", "for", "forward", "if", "in", "int", "len", "list", "nn", "num_features", "return", "self", "squeeze", "super", "torch", "zip"], "informer/modeling_informer.py:InformerStdScaler": ["ModelConfig", "ModelStdScaler", "Module", "Tensor", "True", "__init__", "clamp_min", "class", "config", "data", "def", "denominator", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "minimum_scale", "nn", "observed_indicator", "return", "scale", "scaling_dim", "self", "sqrt", "sum", "super", "torch", "tuple", "variance"], "informer/modeling_informer.py:InformerMeanScaler": ["ModelConfig", "ModelMeanScaler", "Module", "None", "Tensor", "True", "__init__", "abs", "batch_observations", "batch_sum", "clamp", "class", "config", "data", "def", "default_scale", "dim", "else", "forward", "hasattr", "if", "is", "keepdim", "min", "minimum_scale", "nn", "not", "num_observed", "observed_indicator", "ones_like", "return", "scale", "scaled_data", "scaling_dim", "self", "squeeze", "sum", "super", "torch", "ts_sum", "tuple", "where", "zeros_like"], "informer/modeling_informer.py:InformerNOPScaler": ["False", "ModelConfig", "ModelNOPScaler", "Module", "None", "Optional", "Tensor", "True", "__init__", "class", "config", "data", "def", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "mean", "nn", "observed_indicator", "ones_like", "requires_grad", "return", "scale", "scaling_dim", "self", "super", "torch", "tuple", "zeros_like"], "informer/modeling_informer.py:InformerSinusoidalPositionalEmbedding": ["Embedding", "False", "FloatTensor", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_init_weight", "arange", "array", "bsz", "class", "cos", "def", "device", "dim", "dtype", "else", "embedding_dim", "empty", "for", "forward", "if", "in", "input_ids_shape", "int", "is", "j", "long", "n_pos", "nn", "no_grad", "np", "num_positions", "out", "padding_idx", "past_key_values_length", "pos", "position_enc", "position_ids", "power", "range", "requires_grad", "return", "self", "sentinel", "seq_len", "shape", "sin", "super", "torch", "weight"], "informer/modeling_informer.py:InformerValueEmbedding": ["False", "Linear", "ModelValueEmbedding", "Module", "__init__", "bias", "class", "d_model", "def", "feature_size", "forward", "in_features", "nn", "out_features", "return", "self", "super", "value_projection", "x"], "informer/modeling_informer.py:InformerPreTrainedModel": ["False", "ModelConfig", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "Module", "None", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_init_weight", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask", "_prepare_4d_causal_attention_mask_for_sdpa", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "attention_mask", "base_model_prefix", "class", "config", "def", "device", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "flash", "flash_attention_2", "flex_attention", "if", "in", "input_shape", "inputs_embeds", "int", "is", "is_causal", "isinstance", "main_input_name", "make_flex_block_causal_mask", "model", "module", "nn", "not", "ones", "past_key_values_length", "past_values", "query_length", "return", "sdpa", "self", "size", "super", "supports_gradient_checkpointing", "tgt_len", "torch"], "informer/modeling_informer.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "informer/modeling_informer.py:InformerAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "informer/modeling_informer.py:InformerProbSparseAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelProbSparseAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_shape", "a", "and", "arange", "astype", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "ceil", "class", "clone", "context", "contiguous", "cross_attention_cache", "cumsum", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "dim_for_slice", "div", "divisible", "dropout", "dtype", "else", "embed_dim", "expand", "f", "factor", "float", "float32", "for", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "index_sample", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "item", "k_proj", "k_sample", "key_states", "key_states_time_length", "key_value_states", "keys", "kv_input_shape", "layer", "layer_head_mask", "layer_idx", "layers", "log1p", "log_key_states_time_length", "log_query_states_time_length", "mask", "max", "mean", "min", "must", "new_name", "nn", "not", "np", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "prob_mask", "proj_shape", "q_proj", "q_reduce", "queries_keys_sample", "query_states", "query_states_time_length", "raise", "randint", "reshape", "return", "sampling_factor", "scaling", "self", "self_attention_cache", "seq_len", "shape", "should", "single", "size", "softmax", "sorted", "sparsity_measurement", "src_len", "sum", "super", "tensor", "tgt_len", "to", "top_u_sparsity_measurement", "topk", "torch", "training", "transpose", "tuple", "u", "u_part", "unsqueeze", "update", "v_mean_dim_time", "v_proj", "value_states", "values", "version", "view", "weights"], "informer/modeling_informer.py:InformerConvLayer": ["BatchNorm1d", "Conv1d", "ELU", "GradientCheckpointingLayer", "MaxPool1d", "ModelConvLayer", "__init__", "activation", "c_in", "circular", "class", "def", "downConv", "forward", "in_channels", "kernel_size", "maxPool", "nn", "norm", "out_channels", "padding", "padding_mode", "permute", "return", "self", "stride", "super", "transpose", "x"], "informer/modeling_informer.py:InformerEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelProbSparseAttention", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attention_type", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "else", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "layer_head_mask", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "prob", "residual", "return", "sampling_factor", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "informer/modeling_informer.py:InformerDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelProbSparseAttention", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "prob", "residual", "return", "sampling_factor", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "informer/modeling_informer.py:InformerEncoder": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelConfig", "ModelConvLayer", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModelValueEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "append", "attention_mask", "attentions", "be", "bool", "but", "class", "config", "context_length", "conv_layer", "conv_layers", "d_model", "def", "distil", "dropout", "dropout_probability", "dtype", "else", "embed_pos", "embed_positions", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "feature_size", "for", "forward", "functional", "gradient_checkpointing", "head_mask", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "needs", "nn", "not", "output", "output_attentions", "output_hidden_states", "p", "post_init", "prediction_length", "r", "raise", "rand", "range", "return", "return_dict", "self", "should", "size", "specified", "super", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "value_embedding", "zip"], "informer/modeling_informer.py:InformerDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModelValueEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "attention_mask", "attentions", "attn_mask", "be", "bool", "but", "cache_position", "checkpointing", "class", "config", "context_length", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "else", "embed_pos", "embed_positions", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "feature_size", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "needs", "nn", "not", "of", "output_attentions", "output_hidden_states", "p", "pass", "past_key_values", "past_key_values_length", "post_init", "prediction_length", "r", "raise", "rand", "range", "removed", "return", "return_dict", "self", "should", "size", "specified", "super", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "value_embedding", "warning_once", "will", "with", "zip"], "informer/modeling_informer.py:InformerModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelFeatureEmbedder", "ModelMeanScaler", "ModelModel", "ModelNOPScaler", "ModelPreTrainedModel", "ModelStdScaler", "None", "Optional", "Seq2SeqTSModelOutput", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_past_length", "abs", "and", "append", "attention_mask", "attentions", "auto_docstring", "begin_index", "bool", "bsz", "cache_position", "cannot", "cardinalities", "cardinality", "cat", "class", "config", "context", "context_length", "create_network_inputs", "cross_attentions", "cross_attn_head_mask", "dec_input", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_outputs", "def", "device", "dim", "does", "dtype", "elif", "else", "embedded_cat", "embedder", "embedding_dimension", "embedding_dims", "enc_input", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_index", "expand", "expanded_static_feat", "f", "feature", "features", "for", "forward", "found", "further", "future_time_features", "future_values", "get_encoder", "get_lagged_subsequences", "go", "head_mask", "hidden_states", "history", "if", "in", "indices", "input", "inputs", "inputs_embeds", "int", "is", "isinstance", "lag", "lag_index", "lagged_sequence", "lagged_values", "lags", "lags_sequence", "lags_shape", "last_hidden_state", "len", "length", "lengths", "list", "loc", "log", "log1p", "log_abs_loc", "log_scale", "match", "max", "mean", "ndim", "not", "num_static_categorical_features", "observed_context", "ones_like", "only", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_observed_mask", "past_time_features", "past_values", "post_init", "prediction_length", "property", "r", "raise", "reshape", "reshaped_lagged_sequence", "return", "return_dict", "scale", "scaler", "scaling", "self", "sequence", "sequence_length", "shape", "shift", "size", "squeeze", "squeezed_loc", "squeezed_scale", "stack", "static_categorical_features", "static_feat", "static_features", "static_real_features", "std", "subsequences_length", "super", "than", "time", "time_feat", "torch", "transformer_inputs", "tuple", "unsqueeze", "use_cache", "use_return_dict", "while", "zeros"], "informer/modeling_informer.py:weighted_average": ["Model_average", "Model_tensor", "None", "Optional", "Tensor", "clamp", "def", "dim", "else", "if", "input_tensor", "is", "mean", "min", "not", "return", "sum", "sum_weights", "torch", "weights", "where", "zeros_like"], "informer/modeling_informer.py:nll": ["Distribution", "Model", "Tensor", "def", "distributions", "input", "log_prob", "return", "target", "torch"], "informer/modeling_informer.py:InformerForPrediction": ["Cache", "Distribution", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelForPrediction", "ModelModel", "ModelPreTrainedModel", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleTSPredictionOutput", "Seq2SeqTSModelOutput", "Seq2SeqTSPredictionOutput", "StudentTOutput", "Tensor", "True", "Union", "Unknown", "ValueError", "_", "__init__", "append", "auto_docstring", "bool", "cache_position", "cat", "class", "concat_future_samples", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "dec_last_hidden", "dec_output", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input", "def", "dim", "distr", "distribution", "distribution_output", "distributions", "elif", "else", "enc_last_hidden", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "event_shape", "expand", "expanded_static_feat", "f", "features", "for", "forward", "function", "future_observed_mask", "future_samples", "future_time_features", "future_values", "generate", "get_decoder", "get_encoder", "get_lagged_subsequences", "get_parameter_projection", "head_mask", "if", "ignore", "in", "input_size", "inputs_embeds", "is", "jit", "k", "keepdim", "lagged_sequence", "lags_shape", "last_hidden_state", "len", "list", "loc", "loss", "loss_weights", "min", "model", "negative_binomial", "next_sample", "nll", "no_grad", "normal", "not", "num_parallel_samples", "ones_like", "output", "output_attentions", "output_distribution", "output_hidden_states", "output_params", "outputs", "p", "parameter_projection", "params", "past_key_values", "past_observed_mask", "past_time_features", "past_values", "post_init", "prediction_length", "prediction_loss", "r", "raise", "range", "repeat_interleave", "repeated_enc_last_hidden", "repeated_features", "repeated_loc", "repeated_past_values", "repeated_scale", "repeats", "reshape", "reshaped_lagged_sequence", "return", "return_dict", "sample", "scale", "self", "sequence", "sequences", "shape", "shift", "sliced_params", "static_categorical_features", "static_feat", "static_features", "static_real_features", "student_t", "subsequences_length", "super", "target_shape", "torch", "trailing_n", "tuple", "unsqueeze", "use_cache", "use_return_dict", "weighted_average", "weights"], "camembert/modeling_camembert.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "camembert/modeling_camembert.py:CamembertSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "camembert/modeling_camembert.py:CamembertCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "camembert/modeling_camembert.py:CamembertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "camembert/modeling_camembert.py:CamembertAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "camembert/modeling_camembert.py:CamembertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "camembert/modeling_camembert.py:CamembertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "camembert/modeling_camembert.py:CamembertLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "camembert/modeling_camembert.py:CamembertLMHead": ["LayerNorm", "Linear", "ModelLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "eps", "features", "forward", "gelu", "hidden_size", "if", "kwargs", "layer_norm", "layer_norm_eps", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "x", "zeros"], "camembert/modeling_camembert.py:CamembertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelCrossAttention", "ModelLMHead", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "roberta", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "camembert/modeling_camembert.py:CamembertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "camembert/modeling_camembert.py:CamembertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "camembert/modeling_camembert.py:CamembertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "camembert/modeling_camembert.py:CamembertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "camembert/modeling_camembert.py:CamembertForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "ModelForMaskedLM", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention", "attention_mask", "attentions", "auto_docstring", "bi", "bias", "can_return_tuple", "class", "config", "decoder", "def", "device", "directional", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "logger", "logits", "loss", "loss_fct", "make", "masked_lm_loss", "new_embeddings", "not", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "super", "sure", "to", "token_type_ids", "torch", "tuple", "use", "view", "vocab_size", "want", "warning", "weight", "you"], "camembert/modeling_camembert.py:CamembertClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "camembert/modeling_camembert.py:CamembertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "roberta", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "view"], "camembert/modeling_camembert.py:CamembertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "config", "def", "device", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "flat_token_type_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "roberta", "self", "shape", "size", "super", "to", "token_type_ids", "torch", "tuple", "view"], "camembert/modeling_camembert.py:CamembertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "roberta", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "view"], "camembert/modeling_camembert.py:CamembertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "roberta", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "camembert/modeling_camembert.py:CamembertForCausalLM": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "LongTensor", "ModelForCausalLM", "ModelLMHead", "ModelLMHeadModel", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "add_pooling_layer", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "device", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "roberta", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "mobilevit/modeling_mobilevit.py:make_divisible": ["Model_divisible", "None", "Optional", "def", "divisor", "if", "int", "is", "max", "min_value", "new_value", "return", "value"], "mobilevit/modeling_mobilevit.py:MobileViTConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Input", "ModelConfig", "ModelConvLayer", "Module", "None", "Output", "Tensor", "True", "Union", "ValueError", "__init__", "activation", "affine", "are", "bias", "bool", "by", "channels", "class", "config", "convolution", "def", "dilation", "divisible", "elif", "else", "eps", "f", "features", "forward", "groups", "hidden_act", "if", "in_channels", "int", "is", "isinstance", "kernel_size", "momentum", "nn", "normalization", "not", "num_features", "out_channels", "padding", "padding_mode", "raise", "return", "self", "str", "stride", "super", "torch", "track_running_stats", "use_activation", "use_normalization", "zeros"], "mobilevit/modeling_mobilevit.py:MobileViTInvertedResidual": ["False", "Invalid", "ModelConfig", "ModelConvLayer", "ModelInvertedResidual", "Module", "None", "Tensor", "ValueError", "__init__", "and", "class", "config", "conv_3x3", "def", "dilation", "else", "expand_1x1", "expand_ratio", "expanded_channels", "f", "features", "forward", "groups", "if", "in", "in_channels", "int", "kernel_size", "make_divisible", "nn", "not", "out_channels", "raise", "reduce_1x1", "residual", "return", "round", "self", "stride", "super", "torch", "use_activation", "use_residual"], "mobilevit/modeling_mobilevit.py:MobileViTMobileNetLayer": ["ModelConfig", "ModelInvertedResidual", "ModelModelNetLayer", "Module", "ModuleList", "None", "Tensor", "__init__", "append", "class", "config", "def", "else", "features", "for", "forward", "i", "if", "in", "in_channels", "int", "layer", "layer_module", "nn", "num_stages", "out_channels", "range", "return", "self", "stride", "super", "torch"], "mobilevit/modeling_mobilevit.py:MobileViTSelfAttention": ["Dropout", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Tensor", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "f", "forward", "functional", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "value", "value_layer", "view"], "mobilevit/modeling_mobilevit.py:MobileViTSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "int", "nn", "return", "self", "super", "torch"], "mobilevit/modeling_mobilevit.py:MobileViTAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Tensor", "__init__", "all_head_size", "attention", "attention_head_size", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_size", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "union", "value"], "mobilevit/modeling_mobilevit.py:MobileViTIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "int", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "mobilevit/modeling_mobilevit.py:MobileViTOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "int", "intermediate_size", "nn", "return", "self", "super", "torch"], "mobilevit/modeling_mobilevit.py:MobileViTTransformerLayer": ["LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelOutput", "ModelTransformerLayer", "Module", "None", "Tensor", "__init__", "attention", "attention_output", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "int", "intermediate", "intermediate_size", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "super", "torch"], "mobilevit/modeling_mobilevit.py:MobileViTTransformer": ["ModelConfig", "ModelTransformer", "ModelTransformerLayer", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "append", "class", "config", "def", "for", "forward", "hidden_size", "hidden_states", "in", "int", "intermediate_size", "layer", "layer_module", "mlp_ratio", "nn", "num_stages", "range", "return", "self", "super", "torch", "transformer_layer"], "mobilevit/modeling_mobilevit.py:MobileViTLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelConfig", "ModelConvLayer", "ModelInvertedResidual", "ModelLayer", "ModelTransformer", "None", "Tensor", "True", "__init__", "align_corners", "batch_size", "bilinear", "cat", "ceil", "channels", "class", "config", "contiguous", "conv_1x1", "conv_kernel_size", "conv_kxk", "conv_projection", "def", "dict", "dilation", "dim", "downsampling_layer", "else", "eps", "features", "folding", "forward", "functional", "fusion", "hidden_size", "if", "in_channels", "info_dict", "int", "interpolate", "is_tracing", "jit", "kernel_size", "layer_norm_eps", "layernorm", "math", "mode", "new_height", "new_width", "nn", "num_patch_height", "num_patch_width", "num_patches", "num_patches_height", "num_patches_width", "num_stages", "or", "orig_height", "orig_size", "orig_width", "out_channels", "patch_area", "patch_height", "patch_size", "patch_width", "patches", "reshape", "residual", "return", "self", "shape", "size", "stride", "super", "torch", "torch_int", "transformer", "transpose", "tuple", "unfolding", "use_activation", "use_normalization", "view"], "mobilevit/modeling_mobilevit.py:MobileViTEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "ModelModelNetLayer", "Module", "ModuleList", "None", "Tensor", "True", "Union", "__init__", "all_hidden_states", "append", "bool", "class", "config", "def", "dilate_layer_4", "dilate_layer_5", "dilation", "elif", "else", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_sizes", "hidden_states", "i", "if", "in", "in_channels", "is", "last_hidden_state", "layer", "layer_1", "layer_2", "layer_3", "layer_4", "layer_5", "layer_module", "neck_hidden_sizes", "nn", "not", "num_stages", "out_channels", "output_hidden_states", "output_stride", "return", "return_dict", "self", "stride", "super", "torch", "tuple", "v"], "mobilevit/modeling_mobilevit.py:MobileViTPreTrainedModel": ["BatchNorm2d", "Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLayer", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "mobilevit/modeling_mobilevit.py:MobileViTModel": ["BaseModelOutputWithPoolingAndNoAttention", "False", "ModelConfig", "ModelConvLayer", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPreTrainedModel", "Model_layer", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "bool", "class", "config", "conv_1x1_exp", "conv_stem", "def", "dim", "else", "embedding_output", "encoder", "encoder_outputs", "expand_output", "for", "forward", "have", "heads", "heads_to_prune", "hidden_states", "if", "in", "in_channels", "is", "isinstance", "items", "keepdim", "kernel_size", "last_hidden_state", "layer", "layer_index", "mean", "neck_hidden_sizes", "not", "num_channels", "out_channels", "output", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "self", "specify", "stride", "super", "to", "torch", "transformer", "transformer_layer", "tuple", "use_return_dict"], "mobilevit/modeling_mobilevit.py:MobileViTForImageClassification": ["Dropout", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "else", "forward", "hidden_states", "if", "inplace", "is", "labels", "logits", "loss", "loss_function", "neck_hidden_sizes", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "mobilevit/modeling_mobilevit.py:MobileViTASPPPooling": ["AdaptiveAvgPool2d", "False", "ModelASPPPooling", "ModelConfig", "ModelConvLayer", "Module", "None", "Tensor", "True", "__init__", "align_corners", "bilinear", "class", "config", "conv_1x1", "def", "features", "forward", "functional", "global_pool", "in_channels", "int", "interpolate", "kernel_size", "mode", "nn", "out_channels", "output_size", "relu", "return", "self", "shape", "size", "spatial_size", "stride", "super", "torch", "use_activation", "use_normalization"], "mobilevit/modeling_mobilevit.py:MobileViTASPP": ["Dropout", "Expected", "ModelASPP", "ModelASPPPooling", "ModelConfig", "ModelConvLayer", "Module", "ModuleList", "None", "Tensor", "ValueError", "__init__", "append", "aspp_dropout_prob", "aspp_out_channels", "atrous_rates", "cat", "class", "config", "conv", "convs", "def", "dilation", "dim", "dropout", "extend", "features", "for", "forward", "if", "in", "in_channels", "in_projection", "kernel_size", "len", "neck_hidden_sizes", "nn", "out_channels", "p", "pool_layer", "pooled_features", "project", "pyramid", "raise", "rate", "relu", "return", "self", "super", "torch", "use_activation", "values"], "mobilevit/modeling_mobilevit.py:MobileViTDeepLabV3": ["Dropout2d", "False", "ModelASPP", "ModelConfig", "ModelConvLayer", "ModelDeepLabV3", "Module", "None", "Tensor", "True", "__init__", "aspp", "aspp_out_channels", "bias", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "features", "forward", "hidden_states", "in_channels", "kernel_size", "nn", "num_labels", "out_channels", "return", "self", "super", "torch", "use_activation", "use_normalization"], "mobilevit/modeling_mobilevit.py:MobileViTForSemanticSegmentation": ["CrossEntropyLoss", "False", "Model", "ModelConfig", "ModelDeepLabV3", "ModelForSemanticSegmentation", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SemanticSegmenterOutput", "Tensor", "The", "True", "Union", "ValueError", "__init__", "align_corners", "and", "attentions", "auto_docstring", "be", "bilinear", "bool", "class", "config", "def", "else", "encoder_hidden_states", "expand_output", "forward", "functional", "greater", "hidden_states", "if", "ignore_index", "interpolate", "is", "labels", "logits", "loss", "loss_fct", "mode", "nn", "not", "num_labels", "number", "of", "one", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "segmentation_head", "self", "semantic_loss_ignore_index", "shape", "should", "size", "super", "than", "torch", "tuple", "upsampled_logits", "use_return_dict"], "albert/modeling_albert.py:AlbertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embedding_size", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "if", "index", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "shape", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "albert/modeling_albert.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "albert/modeling_albert.py:AlbertAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "False", "FloatTensor", "However", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "can", "class", "config", "contiguous", "def", "dense", "dim", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "eps", "f", "find_pruneable_heads_and_indices", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_dropout_prob", "hidden_shape", "hidden_size", "hidden_states", "if", "index", "input_shape", "int", "is", "is_causal", "key", "key_layer", "kwargs", "layer_norm_eps", "len", "list", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "output_dropout", "p", "position_embedding_type", "positional", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "set", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "union", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "albert/modeling_albert.py:AlbertLayer": ["ACT2FN", "Dropout", "FloatTensor", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelLayer", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "activation", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "dropout", "eps", "ff_chunk", "ffn", "ffn_output", "forward", "full_layer_layer_norm", "head_mask", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "intermediate_size", "kwargs", "layer_norm_eps", "nn", "return", "self", "seq_len_dim", "super", "torch", "tuple"], "albert/modeling_albert.py:AlbertLayerGroup": ["FloatTensor", "ModelConfig", "ModelLayer", "ModelLayerGroup", "Model_layer", "Model_layers", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "class", "config", "def", "enumerate", "for", "forward", "head_mask", "hidden_states", "in", "inner_group_num", "kwargs", "layer_index", "nn", "range", "return", "self", "super", "torch", "tuple"], "albert/modeling_albert.py:AlbertTransformer": ["BaseModelOutput", "FloatTensor", "Linear", "ModelConfig", "ModelLayerGroup", "ModelTransformer", "Model_layer_groups", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "class", "config", "def", "else", "embedding_hidden_mapping_in", "embedding_size", "for", "forward", "group_idx", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "int", "is", "kwargs", "last_hidden_state", "layers_per_group", "nn", "num_hidden_groups", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "albert/modeling_albert.py:AlbertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelLayer", "ModelMLMHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "albert/modeling_albert.py:AlbertForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "sop_logits", "torch", "tuple"], "albert/modeling_albert.py:AlbertModel": ["BaseModelOutputWithPooling", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelEmbeddings", "ModelModel", "ModelPreTrainedModel", "ModelTransformer", "Model_layer_groups", "Model_layers", "None", "Optional", "Tanh", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_full_mask", "add_pooling_layer", "attention", "attention_mask", "attn_implementation", "auto_docstring", "base_model_prefix", "bool", "check_model_inputs", "class", "config", "config_class", "def", "dict", "dtype", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "exactly", "flash", "flex_attention", "for", "forward", "get_head_mask", "get_input_embeddings", "group_idx", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "inner_group_idx", "inner_group_num", "input_ids", "inputs_embeds", "int", "is", "is_causal", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "make_flex_block_causal_mask", "must", "nn", "not", "num_hidden_layers", "of", "one", "or", "pooled_output", "pooler", "pooler_activation", "pooler_output", "position_embedding_type", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "sdpa", "self", "sequence_output", "set_input_embeddings", "specify", "super", "token_type_ids", "torch", "tuple", "value", "word_embeddings"], "albert/modeling_albert.py:AlbertForPreTraining": ["CrossEntropyLoss", "Embedding", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelMLMHead", "ModelModel", "ModelPreTrainedModel", "ModelSOPHead", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "can_return_tuple", "class", "config", "decoder", "def", "embeddings", "forward", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "nn", "not", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sentence_order_label", "sentence_order_loss", "sequence_output", "set_output_embeddings", "sop_classifier", "sop_logits", "sop_scores", "super", "token_type_ids", "torch", "total_loss", "tuple", "view", "vocab_size", "weight", "word_embeddings"], "albert/modeling_albert.py:AlbertMLMHead": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelMLMHead", "Module", "None", "Parameter", "Tensor", "__init__", "_tie_weights", "activation", "bias", "class", "config", "decoder", "def", "dense", "device", "else", "embedding_size", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "layer_norm_eps", "meta", "nn", "prediction_scores", "return", "self", "super", "torch", "type", "vocab_size", "zeros"], "albert/modeling_albert.py:AlbertSOPHead": ["Dropout", "Linear", "ModelConfig", "ModelSOPHead", "Module", "Tensor", "__init__", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "dropout_pooled_output", "forward", "hidden_size", "logits", "nn", "num_labels", "pooled_output", "return", "self", "super", "torch"], "albert/modeling_albert.py:AlbertForMaskedLM": ["CrossEntropyLoss", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelMLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bias", "can_return_tuple", "class", "config", "decoder", "def", "embeddings", "forward", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "nn", "not", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_outputs", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "view", "vocab_size", "weight", "word_embeddings"], "albert/modeling_albert.py:AlbertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "view"], "albert/modeling_albert.py:AlbertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "view"], "albert/modeling_albert.py:AlbertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForPreTrainingOutput", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "kwargs", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple"], "albert/modeling_albert.py:AlbertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForMultipleChoice", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "classifier", "classifier_dropout_prob", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "view"], "bert_generation/modeling_bert_generation.py:BertGenerationSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bert_generation/modeling_bert_generation.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "bert_generation/modeling_bert_generation.py:BertGenerationSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "bert_generation/modeling_bert_generation.py:BertGenerationCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "bert_generation/modeling_bert_generation.py:BertGenerationAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "bert_generation/modeling_bert_generation.py:BertGenerationIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "bert_generation/modeling_bert_generation.py:BertGenerationOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bert_generation/modeling_bert_generation.py:BertGenerationLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "_", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "bert_generation/modeling_bert_generation.py:BertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "bert_generation/modeling_bert_generation.py:BertGenerationEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelEmbeddings", "Module", "None", "__init__", "arange", "class", "config", "def", "dropout", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "torch", "vocab_size", "word_embeddings"], "bert_generation/modeling_bert_generation.py:BertGenerationPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelCrossAttention", "ModelLayer", "ModelOnlyLMHead", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "config_class", "cross_attentions", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "bert_generation/modeling_bert_generation.py:BertGenerationEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Passing", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "an", "and", "arange", "attention", "attention_mask", "auto_docstring", "be", "bool", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "pass", "past_key_values", "past_key_values_length", "please", "position_ids", "post_init", "prune_heads", "query_length", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "torch", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "bert_generation/modeling_bert_generation.py:BertGenerationOnlyLMHead": ["Linear", "ModelOnlyLMHead", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "logits", "meta", "nn", "return", "self", "super", "torch", "type", "vocab_size", "zeros"], "bert_generation/modeling_bert_generation.py:BertGenerationDecoder": ["CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "If", "Model", "ModelDecoder", "ModelEncoder", "ModelOnlyLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "a", "add", "as", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "cross_attentions", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "kwargs", "labels", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_function", "new_embeddings", "not", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "standalone", "super", "to", "torch", "tuple", "use", "use_cache", "vocab_size", "want", "warning", "weight", "you"], "swiftformer/modeling_swiftformer.py:SwiftFormerPatchEmbedding": ["BatchNorm2d", "Conv2d", "ModelConfig", "ModelPatchEmbedding", "Module", "ReLU", "Sequential", "__init__", "batch_norm_eps", "class", "config", "def", "embed_dims", "eps", "forward", "in_chs", "kernel_size", "nn", "num_channels", "out_chs", "padding", "patch_embedding", "return", "self", "stride", "super", "x"], "swiftformer/modeling_swiftformer.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "swiftformer/modeling_swiftformer.py:SwiftFormerDropPath": ["ModelConfig", "ModelDropPath", "Module", "None", "Tensor", "__init__", "class", "config", "def", "drop_path", "drop_path_rate", "drop_prob", "extra_repr", "f", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "swiftformer/modeling_swiftformer.py:SwiftFormerEmbeddings": ["BatchNorm2d", "Conv2d", "Iterable", "ModelConfig", "ModelEmbeddings", "Module", "__init__", "abc", "batch_norm_eps", "class", "collections", "config", "def", "down_pad", "down_patch_size", "down_stride", "else", "embed_dim", "embed_dims", "eps", "forward", "if", "in_chans", "index", "int", "isinstance", "kernel_size", "nn", "norm", "padding", "patch_size", "proj", "return", "self", "stride", "super", "x"], "swiftformer/modeling_swiftformer.py:SwiftFormerConvEncoder": ["BatchNorm2d", "Conv2d", "Dropout", "GELU", "ModelConfig", "ModelConvEncoder", "Module", "Parameter", "True", "__init__", "act", "batch_norm_eps", "class", "config", "def", "depth_wise_conv", "dim", "drop_conv_encoder_rate", "drop_path", "eps", "forward", "groups", "hidden_dim", "input", "int", "kernel_size", "layer_scale", "mlp_ratio", "nn", "norm", "ones", "p", "padding", "point_wise_conv1", "point_wise_conv2", "requires_grad", "return", "self", "super", "torch", "unsqueeze", "x"], "swiftformer/modeling_swiftformer.py:SwiftFormerMlp": ["ACT2CLS", "BatchNorm2d", "Conv2d", "Dropout", "ModelConfig", "ModelMlp", "Module", "__init__", "act", "act_layer", "batch_norm_eps", "class", "config", "def", "drop", "drop_mlp_rate", "eps", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "in_features", "int", "mlp_ratio", "nn", "norm1", "p", "return", "self", "super", "x"], "swiftformer/modeling_swiftformer.py:SwiftFormerEfficientAdditiveAttention": ["Linear", "ModelConfig", "ModelEfficientAdditiveAttention", "Module", "Parameter", "__init__", "class", "config", "def", "dim", "final", "forward", "functional", "global_queries", "int", "key", "nn", "normalize", "out", "proj", "query", "query_weight", "randn", "repeat", "return", "scale_factor", "scaled_query_weight", "self", "shape", "softmax", "sum", "super", "to_key", "to_query", "torch", "unsqueeze", "w_g", "x"], "swiftformer/modeling_swiftformer.py:SwiftFormerLocalRepresentation": ["BatchNorm2d", "Conv2d", "GELU", "Identity", "ModelConfig", "ModelLocalRepresentation", "Module", "Parameter", "True", "__init__", "act", "batch_norm_eps", "class", "config", "def", "depth_wise_conv", "dim", "drop_path", "eps", "forward", "groups", "input", "int", "kernel_size", "layer_scale", "nn", "norm", "ones", "padding", "point_wise_conv1", "point_wise_conv2", "requires_grad", "return", "self", "super", "torch", "unsqueeze", "x"], "swiftformer/modeling_swiftformer.py:SwiftFormerEncoderBlock": ["Identity", "ModelConfig", "ModelDropPath", "ModelEfficientAdditiveAttention", "ModelEncoderBlock", "ModelLocalRepresentation", "ModelMlp", "Module", "None", "Parameter", "True", "__init__", "attn", "batch_size", "channels", "class", "config", "def", "dim", "drop_path", "else", "float", "forward", "height", "if", "in_features", "int", "layer_scale_1", "layer_scale_2", "layer_scale_init_value", "linear", "local_representation", "nn", "ones", "permute", "requires_grad", "res", "reshape", "return", "self", "shape", "super", "torch", "unsqueeze", "use_layer_scale", "width", "x"], "swiftformer/modeling_swiftformer.py:SwiftFormerStage": ["ModelConfig", "ModelConvEncoder", "ModelEncoderBlock", "ModelStage", "Module", "ModuleList", "None", "__init__", "append", "block", "block_dpr", "block_idx", "blocks", "class", "config", "def", "depth", "depths", "dim", "drop_path", "drop_path_rate", "else", "embed_dims", "for", "forward", "if", "in", "index", "input", "int", "layer_depths", "nn", "range", "return", "self", "sum", "super"], "swiftformer/modeling_swiftformer.py:SwiftFormerEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "__init__", "all_hidden_states", "append", "block", "bool", "break", "class", "config", "def", "depths", "downsamples", "else", "embed_dims", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "if", "in", "index", "is", "last_hidden_state", "layer_depths", "len", "network", "nn", "not", "or", "output_hidden_states", "range", "return", "return_dict", "self", "stage", "super", "torch", "tuple", "use_return_dict", "v"], "swiftformer/modeling_swiftformer.py:SwiftFormerPreTrainedModel": ["BatchNorm2d", "Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelConvEncoder", "ModelEfficientAdditiveAttention", "ModelEncoderBlock", "ModelLocalRepresentation", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "constant_", "data", "def", "elif", "fill_", "if", "init", "is", "isinstance", "layer_scale", "layer_scale_1", "layer_scale_2", "layer_scale_init_value", "main_input_name", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "use_layer_scale", "w_g", "weight"], "swiftformer/modeling_swiftformer.py:SwiftFormerModel": ["BaseModelOutputWithNoAttention", "ModelConfig", "ModelEncoder", "ModelModel", "ModelPatchEmbedding", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embedding_output", "encoder", "encoder_outputs", "for", "forward", "have", "hidden_states", "if", "in", "is", "last_hidden_state", "not", "output_hidden_states", "patch_embed", "pixel_values", "post_init", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict", "v"], "swiftformer/modeling_swiftformer.py:SwiftFormerForImageClassification": ["BatchNorm2d", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "auto_docstring", "batch_norm_eps", "bool", "class", "cls_out", "config", "def", "dist_head", "distillation_out", "else", "embed_dims", "eps", "flatten", "forward", "head", "hidden_states", "if", "is", "labels", "last_hidden_state", "logits", "loss", "loss_function", "mean", "nn", "norm", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesFeatureEmbedder": ["Embedding", "ModelSeriesFeatureEmbedder", "Module", "ModuleList", "None", "Tensor", "__init__", "c", "cardinalities", "cat", "cat_feature_slice", "cat_feature_slices", "chunk", "class", "d", "def", "dim", "else", "embed", "embedders", "embedding_dims", "features", "for", "forward", "if", "in", "int", "len", "list", "nn", "num_features", "return", "self", "squeeze", "super", "torch", "zip"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesStdScaler": ["ModelConfig", "ModelSeriesStdScaler", "Module", "Tensor", "True", "__init__", "clamp_min", "class", "config", "data", "def", "denominator", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "minimum_scale", "nn", "observed_indicator", "return", "scale", "scaling_dim", "self", "sqrt", "sum", "super", "torch", "tuple", "variance"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesMeanScaler": ["ModelConfig", "ModelSeriesMeanScaler", "Module", "None", "Tensor", "True", "__init__", "abs", "batch_observations", "batch_sum", "clamp", "class", "config", "data", "def", "default_scale", "dim", "else", "forward", "hasattr", "if", "is", "keepdim", "min", "minimum_scale", "nn", "not", "num_observed", "observed_indicator", "ones_like", "return", "scale", "scaled_data", "scaling_dim", "self", "squeeze", "sum", "super", "torch", "ts_sum", "tuple", "where", "zeros_like"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesNOPScaler": ["False", "ModelConfig", "ModelSeriesNOPScaler", "Module", "None", "Optional", "Tensor", "True", "__init__", "class", "config", "data", "def", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "mean", "nn", "observed_indicator", "ones_like", "requires_grad", "return", "scale", "scaling_dim", "self", "super", "torch", "tuple", "zeros_like"], "time_series_transformer/modeling_time_series_transformer.py:nll": ["Distribution", "Model", "Tensor", "def", "distributions", "input", "log_prob", "return", "target", "torch"], "time_series_transformer/modeling_time_series_transformer.py:weighted_average": ["Model_average", "Model_tensor", "None", "Optional", "Tensor", "clamp", "def", "dim", "else", "if", "input_tensor", "is", "mean", "min", "not", "return", "sum", "sum_weights", "torch", "weights", "where", "zeros_like"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesSinusoidalPositionalEmbedding": ["Embedding", "False", "FloatTensor", "ModelSeriesSinusoidalPositionalEmbedding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_init_weight", "arange", "array", "bsz", "class", "cos", "def", "device", "dim", "dtype", "else", "embedding_dim", "empty", "for", "forward", "if", "in", "input_ids_shape", "int", "is", "j", "long", "n_pos", "nn", "no_grad", "np", "num_positions", "out", "padding_idx", "past_key_values_length", "pos", "position_enc", "position_ids", "power", "range", "requires_grad", "return", "self", "sentinel", "seq_len", "shape", "sin", "super", "torch", "weight"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesValueEmbedding": ["False", "Linear", "ModelSeriesValueEmbedding", "Module", "__init__", "bias", "class", "d_model", "def", "feature_size", "forward", "in_features", "nn", "out_features", "return", "self", "super", "value_projection", "x"], "time_series_transformer/modeling_time_series_transformer.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "None", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "int", "isinf", "isnan", "layer_head_mask", "layer_idx", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerPreTrainedModel": ["Embedding", "False", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelSeriesSinusoidalPositionalEmbedding", "None", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_init_weight", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask", "_prepare_4d_causal_attention_mask_for_sdpa", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "attention_mask", "base_model_prefix", "bias", "class", "config", "data", "def", "device", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "flash", "flash_attention_2", "flex_attention", "if", "in", "init_std", "input_shape", "inputs_embeds", "int", "is", "is_causal", "isinstance", "main_input_name", "make_flex_block_causal_mask", "mean", "model", "module", "nn", "normal_", "not", "ones", "padding_idx", "past_key_values_length", "past_values", "query_length", "return", "sdpa", "self", "size", "std", "supports_gradient_checkpointing", "tgt_len", "torch", "weight", "zero_"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerEncoder": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelSeriesSinusoidalPositionalEmbedding", "ModelSeriesValueEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "_", "__init__", "_update_full_mask", "all_attentions", "attention_mask", "attentions", "be", "bool", "but", "class", "config", "context_length", "d_model", "def", "dropout", "dropout_probability", "else", "embed_pos", "embed_positions", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "feature_size", "for", "forward", "functional", "gradient_checkpointing", "head_mask", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "needs", "nn", "not", "output_attentions", "output_hidden_states", "p", "post_init", "prediction_length", "r", "raise", "rand", "range", "return", "return_dict", "self", "should", "size", "specified", "super", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "value_embedding"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSeriesSinusoidalPositionalEmbedding", "ModelSeriesValueEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "attention_mask", "attentions", "attn_mask", "be", "bool", "but", "cache_position", "checkpointing", "class", "config", "context_length", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "else", "embed_pos", "embed_positions", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "feature_size", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "needs", "nn", "not", "of", "output_attentions", "output_hidden_states", "p", "pass", "past_key_values", "past_key_values_length", "post_init", "prediction_length", "r", "raise", "rand", "range", "removed", "return", "return_dict", "self", "should", "size", "specified", "super", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "value_embedding", "warning_once", "will", "with", "zip"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelSeriesFeatureEmbedder", "ModelSeriesMeanScaler", "ModelSeriesNOPScaler", "ModelSeriesStdScaler", "Model_feat", "None", "Optional", "Seq2SeqTSModelOutput", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_past_length", "abs", "and", "append", "attention_mask", "attentions", "auto_docstring", "begin_index", "bool", "bsz", "cache_position", "cannot", "cardinalities", "cardinality", "cat", "class", "config", "context", "context_length", "create_network_inputs", "cross_attentions", "cross_attn_head_mask", "dec_input", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_outputs", "def", "device", "dim", "does", "dtype", "elif", "else", "embedded_cat", "embedder", "embedding_dimension", "embedding_dims", "enc_input", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_index", "expand", "expanded_static_feat", "f", "feature", "features", "for", "forward", "found", "further", "future_Model_features", "future_values", "get_encoder", "get_lagged_subsequences", "go", "head_mask", "hidden_states", "history", "if", "in", "indices", "input", "inputs", "inputs_embeds", "int", "is", "isinstance", "lag", "lag_index", "lagged_sequence", "lagged_values", "lags", "lags_sequence", "lags_shape", "last_hidden_state", "len", "length", "lengths", "list", "loc", "log", "log1p", "log_abs_loc", "log_scale", "match", "max", "mean", "ndim", "not", "num_static_categorical_features", "observed_context", "ones_like", "only", "or", "output_attentions", "output_hidden_states", "past_Model_features", "past_key_values", "past_observed_mask", "past_values", "post_init", "prediction_length", "property", "r", "raise", "reshape", "reshaped_lagged_sequence", "return", "return_dict", "scale", "scaler", "scaling", "self", "sequence", "sequence_length", "shape", "shift", "size", "squeeze", "squeezed_loc", "squeezed_scale", "stack", "static_categorical_features", "static_feat", "static_features", "static_real_features", "std", "subsequences_length", "super", "than", "torch", "transformer_inputs", "tuple", "unsqueeze", "use_cache", "use_return_dict", "while", "zeros"], "time_series_transformer/modeling_time_series_transformer.py:TimeSeriesTransformerForPrediction": ["Cache", "Distribution", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelForPrediction", "ModelModel", "ModelPreTrainedModel", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleTSPredictionOutput", "Seq2SeqTSModelOutput", "Seq2SeqTSPredictionOutput", "StudentTOutput", "Tensor", "True", "Union", "Unknown", "ValueError", "_", "__init__", "append", "auto_docstring", "bool", "cache_position", "cat", "class", "concat_future_samples", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "dec_last_hidden", "dec_output", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input", "def", "dim", "distr", "distribution", "distribution_output", "distributions", "elif", "else", "enc_last_hidden", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "event_shape", "expand", "expanded_static_feat", "f", "features", "for", "forward", "function", "future_Model_features", "future_observed_mask", "future_samples", "future_values", "generate", "get_decoder", "get_encoder", "get_lagged_subsequences", "get_parameter_projection", "head_mask", "if", "ignore", "in", "input_size", "inputs_embeds", "is", "jit", "k", "keepdim", "lagged_sequence", "lags_shape", "last_hidden_state", "len", "list", "loc", "loss", "loss_weights", "min", "model", "negative_binomial", "next_sample", "nll", "no_grad", "normal", "not", "num_parallel_samples", "ones_like", "output", "output_attentions", "output_distribution", "output_hidden_states", "output_params", "outputs", "p", "parameter_projection", "params", "past_Model_features", "past_key_values", "past_observed_mask", "past_values", "post_init", "prediction_length", "prediction_loss", "r", "raise", "range", "repeat_interleave", "repeated_enc_last_hidden", "repeated_features", "repeated_loc", "repeated_past_values", "repeated_scale", "repeats", "reshape", "reshaped_lagged_sequence", "return", "return_dict", "sample", "scale", "self", "sequence", "sequences", "shape", "shift", "sliced_params", "static_categorical_features", "static_feat", "static_features", "static_real_features", "student_t", "subsequences_length", "super", "target_shape", "torch", "trailing_n", "tuple", "unsqueeze", "use_cache", "use_return_dict", "weighted_average", "weights"], "bart/modeling_bart.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "bart/modeling_bart.py:BartLearnedPositionalEmbedding": ["Embedding", "ModelLearnedPositionalEmbedding", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "else", "embedding_dim", "expand", "forward", "if", "input_ids", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "seq_len", "shape", "super", "torch", "unsqueeze", "weight"], "bart/modeling_bart.py:BartScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "bart/modeling_bart.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "bart/modeling_bart.py:BartAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "bart/modeling_bart.py:BartEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "None", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "int", "isinf", "isnan", "layer_head_mask", "layer_idx", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "bart/modeling_bart.py:BartDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "bart/modeling_bart.py:BartClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "Tensor", "__init__", "class", "def", "dense", "dropout", "float", "forward", "hidden_states", "inner_dim", "input_dim", "int", "nn", "num_classes", "out_proj", "p", "pooler_dropout", "return", "self", "super", "tanh", "torch"], "bart/modeling_bart.py:BartPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "decoder", "def", "device", "diagonal", "dim", "dtype", "dummy_inputs", "elif", "else", "encoder", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "ne", "nn", "normal_", "not", "npu", "ones", "pad_token", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "property", "query_length", "r", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tensor", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "version", "weight", "xpu", "zero_"], "bart/modeling_bart.py:PretrainedBartModel": ["FutureWarning", "ModelModelModel", "The", "__init_subclass__", "been", "class", "def", "depreciated", "has", "instead", "please", "self", "use", "warn", "warnings"], "bart/modeling_bart.py:BartPretrainedModel": ["FutureWarning", "ModelPreTrainedModel", "ModelPretrainedModel", "PretrainedModelModel", "The", "__init_subclass__", "been", "class", "def", "depreciated", "has", "instead", "please", "self", "use", "warn", "warnings"], "bart/modeling_bart.py:BartEncoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "__init__", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "input", "input_ids", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "weight"], "bart/modeling_bart.py:BartDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with", "zip"], "bart/modeling_bart.py:BartModel": ["BaseModelOutput", "Cache", "FloatTensor", "If", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Please", "Seq2SeqModelOutput", "Tensor", "Union", "ValueError", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "are", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache_position", "cannot", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "device", "either", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "list", "math", "meta", "no", "not", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "pass", "passed", "past_key_values", "post_init", "r", "raise", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "shift_tokens_right", "sqrt", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "bart/modeling_bart.py:BartForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "list", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "super", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "bart/modeling_bart.py:BartForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "ModelClassificationHead", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "classification_head", "classifier_dropout", "config", "cross_attentions", "cross_attn_head_mask", "currently", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_states", "if", "input", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "len", "list", "logits", "long", "loss", "loss_fct", "model", "multi_label_classification", "must", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "same", "self", "sentence_representation", "single_label_classification", "size", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "bart/modeling_bart.py:BartForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqQuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "clamp", "class", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "list", "logits", "loss", "loss_fct", "model", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "weight"], "bart/modeling_bart.py:BartDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "bart/modeling_bart.py:BartForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "tvp/modeling_tvp.py:TvpVideoGroundingOutput": ["FloatTensor", "ModelOutput", "ModelVideoGroundingOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "torch", "tuple"], "tvp/modeling_tvp.py:TvpLoss": ["Loss", "ModelLoss", "Module", "ValueError", "__init__", "add", "candidates", "candidates_end_time", "candidates_start_time", "clamp", "class", "def", "distance", "distance_diff", "div", "duration", "duration_candidates", "duration_diff", "duration_groundtruth", "end_time", "f", "float", "for", "forward", "if", "in", "inter", "iou", "labels", "logits", "loss", "loss_distance", "loss_duration", "loss_iou", "loss_map", "losses", "losses_dict", "max", "mid_candidates", "mid_groundtruth", "min", "mul", "nn", "not", "raise", "return", "self", "square", "start_time", "sub", "super", "supported", "torch", "union", "update"], "tvp/modeling_tvp.py:TvpVisionModel": ["Backbone", "Conv2d", "False", "ModelVisionModel", "Module", "None", "True", "ValueError", "__init__", "and", "backbone", "backbone_config", "batch_size", "bias", "class", "config", "def", "elif", "else", "feature_maps", "forward", "found", "functional", "grid", "grid_encoder_conv", "grid_feat_outputs", "groups", "hasattr", "height", "hidden_size", "hidden_sizes", "if", "in_channels", "inplace", "is", "kernel_size", "load_backbone", "max_pool2d", "new_channel", "new_height", "new_width", "nn", "not", "num_channels", "num_frames", "padding", "permute", "pixel_values", "raise", "relu", "return", "self", "shape", "stride", "super", "view", "width"], "tvp/modeling_tvp.py:TvpVisualInputEmbedding": ["Dropout", "Embedding", "False", "LayerNorm", "ModelVisualInputEmbedding", "Module", "Tensor", "__init__", "add_2d_positional_embeddings", "align_corners", "and", "arange", "batch_size", "bicubic", "bool", "class", "col_position_embeddings", "col_position_ids", "col_shape", "config", "def", "device", "dropout", "dtype", "else", "embedding", "embeddings", "eps", "forward", "functional", "grid", "h0", "height", "hidden_dim", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "layer_norm", "layer_norm_eps", "len", "long", "max_grid_col_position_embeddings", "max_grid_row_position_embeddings", "max_position_embeddings", "mean", "min", "mode", "nn", "num_channels", "num_frames", "or", "permute", "position_embeddings", "positional_embeddings", "return", "row_height", "row_position_embeddings", "row_position_ids", "row_shape", "row_width", "scale_factor", "self", "shape", "super", "token_type_embeddings", "token_type_ids", "torch", "view", "visual_tokens", "visual_tokens_shape", "w0", "width", "zeros"], "tvp/modeling_tvp.py:TvpTextInputEmbeddings": ["Dropout", "Embedding", "LayerNorm", "ModelTextInputEmbeddings", "Module", "None", "__init__", "arange", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "position_embeddings", "position_ids", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "tvp/modeling_tvp.py:TvpAttention": ["Dropout", "LayerNorm", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_reshape", "a", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "attn_dropout", "attn_output", "batch_size", "bool", "class", "config", "contiguous", "def", "dense", "dim", "dropout", "else", "embedding_size", "eps", "eq", "f", "for", "forward", "functional", "h", "hasattr", "head", "head_mask", "heads", "hidden", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "index", "int", "is", "key", "key_layer", "layer_norm", "layer_norm_eps", "len", "long", "mask", "math", "matmul", "mixed_key_layer", "mixed_query_layer", "mixed_value_layer", "multiple", "nn", "not", "num_attention_heads", "number", "of", "ones", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "query_layer", "raise", "reshape", "return", "self", "sequence_length", "set", "shape", "size", "softmax", "sqrt", "sum", "super", "tensor", "the", "torch", "transpose", "union", "value", "value_layer", "view"], "tvp/modeling_tvp.py:TvpIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "tvp/modeling_tvp.py:TvpOutputLayer": ["Dropout", "LayerNorm", "Linear", "ModelOutputLayer", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "tvp/modeling_tvp.py:TvpEncodeLayer": ["GradientCheckpointingLayer", "ModelAttention", "ModelEncodeLayer", "ModelIntermediate", "ModelOutputLayer", "None", "Optional", "__init__", "attention", "attention_mask", "attention_output", "bool", "class", "config", "def", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "super"], "tvp/modeling_tvp.py:TvpEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelEncodeLayer", "ModelEncoder", "Module", "ModuleList", "None", "Optional", "_", "__init__", "all_attentions", "all_hidden_states", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "outputs", "range", "return", "return_dict", "self", "super", "torch"], "tvp/modeling_tvp.py:TvpPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "tvp/modeling_tvp.py:TvpPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "and", "base_model_prefix", "bias", "class", "config", "constant_", "data", "def", "elif", "fan_out", "fill_", "hasattr", "if", "init", "initializer_range", "is", "isinstance", "kaiming_normal_", "mean", "mode", "model", "module", "nn", "nonlinearity", "normal_", "not", "pad_down", "pad_left", "pad_right", "pad_up", "relu", "self", "std", "supports_gradient_checkpointing", "text_prompt", "weight", "zero_"], "tvp/modeling_tvp.py:TvpFrameDownPadPrompter": ["ModelFrameDownPadPrompter", "Module", "Parameter", "ValueError", "__init__", "add", "be", "class", "config", "def", "device", "dtype", "forward", "frame_num", "if", "in", "max_img_size", "must", "nn", "not", "ones", "pad_down", "pixel_values", "prompt", "raise", "randn", "remove", "replace", "return", "self", "shape", "start_point", "super", "to", "torch", "visual_prompt_mask", "visual_prompt_size", "visual_prompter_apply", "zeros"], "tvp/modeling_tvp.py:TvpFramePadPrompter": ["False", "Invalid", "ModelFramePadPrompter", "Module", "Parameter", "Tensor", "ValueError", "__init__", "add", "align_corners", "base", "base_size", "batch", "be", "bicubic", "bool", "cat", "channels", "class", "config", "def", "device", "dim", "dtype", "else", "f", "forward", "functional", "h0", "height", "if", "in", "int", "interpolate", "interpolate_pad_encoding", "max_img_size", "mode", "must", "nn", "not", "num_frames", "ones", "pad_down", "pad_left", "pad_right", "pad_up", "pixel_values", "prompt", "prompt_height", "prompt_width", "raise", "randn", "remove", "replace", "reshape", "return", "scale_factor", "self", "shape", "size", "super", "to", "torch", "value", "visual_prompt_mask", "visual_prompt_size", "visual_prompter_apply", "w0", "width", "zeros"], "tvp/modeling_tvp.py:TvpModel": ["BaseModelOutputWithPooling", "Dropout", "False", "FloatTensor", "LongTensor", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "ModelTextInputEmbeddings", "ModelVisionModel", "ModelVisualInputEmbedding", "Model_PROMPTER_CLASSES_MAPPING", "None", "Optional", "Parameter", "ValueError", "__init__", "_prune_heads", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cat", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "for", "forward", "framedownpad", "framepad", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "head_mask", "heads", "heads_to_prune", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "input_ids", "interpolate_pad_encoding", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "must", "new_ones", "nn", "not", "num_hidden_layers", "ones", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "pt_mask", "r", "raise", "randn", "return", "return_dict", "self", "set_input_embeddings", "shape", "size", "super", "text_embedding_output", "text_prompt", "to", "torch", "value", "vision_model", "visual_attention_mask", "visual_embedding_output", "visual_embeddings", "visual_prompter", "visual_prompter_type", "word_embeddings"], "tvp/modeling_tvp.py:TvpVideoGroundingHead": ["Linear", "ModelVideoGroundingHead", "Module", "ReLU", "Sigmoid", "__init__", "activation_0", "activation_1", "class", "config", "def", "forward", "hidden_size", "layer_0", "layer_1", "logits", "nn", "pooler_output", "return", "self", "super"], "tvp/modeling_tvp.py:TvpForVideoGrounding": ["False", "FloatTensor", "LongTensor", "ModelForVideoGrounding", "ModelLoss", "ModelModel", "ModelPreTrainedModel", "ModelVideoGroundingHead", "ModelVideoGroundingOutput", "None", "Optional", "Tensor", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "criterion", "def", "device", "distance", "distance_loss_weight", "duration", "duration_loss_weight", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "interpolate_pos_encoding", "iou", "is", "labels", "logits", "loss", "loss_dict", "model", "not", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "to", "torch", "tuple", "video_grounding_head"], "colqwen2/modeling_colqwen2.py:ColQwen2PreTrainedModel": ["Conv2d", "Embedding", "Linear", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "text_config", "vlm_config", "weight", "zero_"], "colqwen2/modeling_colqwen2.py:ColQwen2ForRetrievalOutput": ["Cache", "FloatTensor", "ModelForRetrievalOutput", "ModelOutput", "None", "Optional", "Tensor", "attentions", "class", "embeddings", "hidden_states", "loss", "past_key_values", "r", "torch", "tuple"], "colqwen2/modeling_colqwen2.py:ColQwen2ForRetrieval": ["AutoModelForImageTextToText", "Cache", "Embedding", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForRetrieval", "ModelForRetrievalOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "def", "device", "dim", "dtype", "else", "embed_tokens", "embedding_dim", "embedding_proj_layer", "embeddings", "expand_as", "f", "for", "forward", "from_config", "get_dtype", "get_input_embeddings", "get_output_embeddings", "get_rope_index", "grid_thw", "hidden_size", "hidden_states", "if", "image_embeds", "image_grid_thw", "image_mask", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "k", "keepdim", "labels", "language_model", "last_hidden_states", "masked_scatter", "mean_resizing", "model", "model_embeds", "new_embeddings", "new_num_tokens", "nn", "norm", "not", "num_embeddings", "offset", "offsets", "or", "output_attentions", "output_hidden_states", "pad_to_multiple_of", "past_key_values", "pixel_sequence", "pixel_values", "position_ids", "post_init", "proj_dtype", "r", "resize_token_embeddings", "return", "return_dict", "rope_deltas", "self", "set_input_embeddings", "set_output_embeddings", "super", "text_config", "tie_weights", "to", "torch", "type", "unsqueeze", "use_cache", "use_return_dict", "value", "video_grid_thw", "visual", "vlm", "vlm_config", "vlm_hidden_states", "vlm_output", "vocab_size", "weight", "zip"], "bridgetower/modeling_bridgetower.py:BridgeTowerModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_features", "pooler_output", "r", "text_features", "torch", "tuple"], "bridgetower/modeling_bridgetower.py:BridgeTowerContrastiveOutput": ["FloatTensor", "ModelContrastiveOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_embeds", "hidden_states", "image_embeds", "logits", "loss", "r", "text_embeds", "torch", "tuple"], "bridgetower/modeling_bridgetower.py:BridgeTowerResidualAttention": ["False", "LayerNorm", "Linear", "ModelResidualAttention", "Module", "ModuleDict", "MultiheadAttention", "None", "Optional", "OrderedDict", "QuickGELUActivation", "Tensor", "__init__", "attention", "attention_mask", "attn", "attn_mask", "bool", "c_fc", "c_proj", "class", "config", "def", "device", "dtype", "else", "eps", "for", "forward", "gelu", "hidden_size", "hidden_state", "if", "in", "is", "key_padding_mask", "layer", "layer_norm_eps", "ln_1", "ln_2", "mlp", "need_weights", "nn", "not", "residual_state", "return", "self", "super", "to", "torch", "values"], "bridgetower/modeling_bridgetower.py:BridgeTowerTransformer": ["ModelResidualAttention", "ModelTransformer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "append", "attention_mask", "block", "class", "config", "def", "detach", "else", "for", "forward", "hidden_size", "hidden_state", "hidden_states", "if", "in", "nn", "num_hidden_layers", "range", "remove_last_layer", "resblocks", "return", "self", "stop_gradient", "super", "torch"], "bridgetower/modeling_bridgetower.py:BridgeTowerVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "bridgetower/modeling_bridgetower.py:BridgeTowerVisionTransformer": ["False", "LayerNorm", "ModelTransformer", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "ModuleList", "Tensor", "_", "__init__", "append", "attention_mask", "bool", "class", "config", "def", "dim", "else", "embeddings", "eps", "for", "forward", "forward_post", "forward_pre", "hidden_size", "hidden_state", "hidden_states", "hidden_states_stack", "if", "in", "interpolate_pos_encoding", "layer_norm_eps", "ln", "ln_post", "ln_pre", "ln_separate", "nn", "not", "num_hidden_layers", "permute", "pixel_values", "range", "return", "self", "share_layernorm", "stack", "super", "torch", "transformer", "visual_output_post", "zip"], "bridgetower/modeling_bridgetower.py:BridgeTowerLinkTower": ["LayerNorm", "ModelLinkTower", "Module", "NotImplementedError", "Parameter", "__init__", "add", "attention_mask", "beta", "class", "config", "cross_modal_hidden_states", "def", "elif", "else", "eps", "f", "forward", "hidden_size", "hidden_states", "if", "implemented", "in", "interpolate", "is", "layer_norm_eps", "link_tower_type", "nn", "not", "raise", "return", "scaled_add", "scaled_factor", "self", "super", "tensor", "torch"], "bridgetower/modeling_bridgetower.py:BridgeTowerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bridgetower/modeling_bridgetower.py:BridgeTowerIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "bridgetower/modeling_bridgetower.py:BridgeTowerOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "bridgetower/modeling_bridgetower.py:BridgeTowerPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "bridgetower/modeling_bridgetower.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "and", "arange", "attention_mask", "attn_output", "attn_weights", "bhld", "bhlr", "bhrd", "bool", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "float", "functional", "head_mask", "if", "is", "key", "key_length", "kwargs", "long", "lrd", "matmul", "max_position_embeddings", "module", "ndim", "nn", "not", "or", "p", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_length", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "scaling", "shape", "size", "softmax", "tensor", "to", "torch", "training", "transpose", "use_cache", "value", "view"], "bridgetower/modeling_bridgetower.py:BridgeTowerSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "cache_position", "can", "class", "config", "contiguous", "current_past_key_value", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_decoder", "isinstance", "key", "key_layer", "kwargs", "layer_idx", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "size", "super", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "view", "with", "work"], "bridgetower/modeling_bridgetower.py:BridgeTowerCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "However", "Linear", "ModelCrossAttention", "Module", "None", "Optional", "Please", "Tensor", "The", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "a", "absolute", "all_head_size", "and", "are", "as", "attention", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_implementation", "attn_output", "attn_weights", "bsz", "can", "class", "config", "contiguous", "cross_attention_cache", "def", "distance_embedding", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "embeddings", "encoder_hidden_states", "f", "forward", "get", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "is_updated", "key", "key_layer", "keys", "kv_input_shape", "kwargs", "layer_idx", "layers", "load", "max_position_embeddings", "model", "multiple", "nn", "non", "not", "num_attention_heads", "number", "of", "or", "p", "past_key_value", "position_embedding_type", "positional", "q_input_shape", "query", "query_layer", "raise", "relative_key", "relative_key_query", "reshape", "return", "scaling", "self", "shape", "size", "src_len", "super", "tgt_len", "the", "them", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "value", "value_layer", "values", "view", "with", "work"], "bridgetower/modeling_bridgetower.py:BridgeTowerAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelCrossAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "all_head_size", "attention_class", "attention_head_size", "attention_mask", "attention_output", "attn_weights", "cache_position", "class", "config", "def", "dense", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_causal", "is_cross_attention", "key", "kwargs", "layer_idx", "len", "nn", "not", "num_attention_heads", "output", "past_key_value", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "tuple", "union", "value"], "bridgetower/modeling_bridgetower.py:BridgeTowerBertCrossLayer": ["False", "ModelAttention", "ModelBertCrossLayer", "ModelIntermediate", "ModelOutput", "Module", "None", "TransformersKwargs", "True", "Unpack", "__init__", "absolute", "add_cross_attention", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "cross_attention_output", "cross_attn_weights", "crossattention", "def", "encoder_attention_mask", "encoder_hidden_states", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "nn", "output", "past_key_value", "position_embedding_type", "return", "self", "self_attention_output", "self_attn_weights", "seq_len_dim", "super"], "bridgetower/modeling_bridgetower.py:BridgeTowerTextLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelOutput", "ModelTextLayer", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_output", "cross_attn_weights", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_causal", "is_cross_attention", "is_decoder", "kwargs", "layer_idx", "layer_output", "layers", "model", "not", "output", "outputs", "passed", "past_key_value", "position_embedding_type", "raise", "return", "self", "self_attention_output", "self_attn_weights", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "bridgetower/modeling_bridgetower.py:BridgeTowerTextEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "False", "FloatTensor", "ModelTextEncoder", "ModelTextLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "cache_position", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "past_key_value", "past_key_values", "range", "return", "self", "super", "torch", "tuple", "use_cache"], "bridgetower/modeling_bridgetower.py:BridgeTowerTextEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "batch_size", "buffered_token_type_ids", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "gather", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "index", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "shape", "size", "staticmethod", "super", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "bridgetower/modeling_bridgetower.py:BridgeTowerPreTrainedModel": ["Conv2d", "Embedding", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelForContrastiveLearning", "ModelMLMHead", "ModelPreTrainedModel", "ModelResidualAttention", "ModelSelfAttention", "ModelVisionTransformer", "Module", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "and", "attn", "attn_std", "base_model_prefix", "bias", "block", "c_fc", "c_proj", "class", "class_embedding", "config", "data", "def", "elif", "embeddings", "fc_std", "fill_", "for", "hidden_size", "if", "in", "in_proj_bias", "in_proj_weight", "init", "initializer_factor", "is", "isinstance", "logit_scale", "logit_scale_init_value", "mean", "mlp", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "past_key_values", "position_embedding", "proj_std", "resblocks", "self", "std", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "bridgetower/modeling_bridgetower.py:BridgeTowerVisionModel": ["False", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "None", "__init__", "class", "config", "def", "dtype", "embeddings", "forward", "image", "image_mask", "interpolate_pos_encoding", "patch_embedding", "property", "return", "self", "super", "type", "visual", "weight"], "bridgetower/modeling_bridgetower.py:BridgeTowerTextModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "EncoderDecoderCache", "False", "FloatTensor", "ModelPooler", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextModel", "None", "Optional", "Passing", "Setting", "Size", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_create_attention_masks", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prune_heads", "_update_cross_attn_mask", "_update_full_mask", "a", "add_pooling_layer", "an", "and", "arange", "attention", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache_position", "can_return_tuple", "checkpointing", "class", "config", "create_causal_mask", "cross_attentions", "def", "deprecated", "device", "dim", "does", "dtype", "e", "eager", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_hidden_states", "encoder_outputs", "exactly", "f", "flash", "flex_attention", "for", "forward", "from_legacy_cache", "g", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "invert_attention_mask", "is", "is_causal", "is_decoder", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "list", "logger", "make_flex_block_causal_mask", "mask", "must", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pass", "past_key_values", "past_key_values_length", "please", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "query_length", "r", "raise", "removed", "return", "return_legacy_cache", "sdpa", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "should", "specify", "super", "tgt_len", "to_legacy_cache", "token_type_ids", "torch", "training", "tuple", "type", "use", "use_cache", "v4", "value", "warning_once", "will", "with", "word_embeddings", "work"], "bridgetower/modeling_bridgetower.py:BridgeTowerModel": ["Embedding", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Make", "ModelBertCrossLayer", "ModelLinkTower", "ModelModel", "ModelModelOutput", "ModelPooler", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionModel", "ModuleList", "None", "NotImplementedError", "Optional", "Tensor", "Union", "_", "__init__", "all_hidden_states", "all_hidden_states_cross", "all_hidden_states_image", "all_hidden_states_text", "all_self_attentions", "and", "attention_mask", "attentions", "auto_docstring", "bias", "block", "bool", "cat", "class", "cls_features", "cls_features_image", "cls_features_text", "config", "cross_image_features", "cross_image_features_", "cross_modal_image", "cross_modal_image_layernorm", "cross_modal_image_layers", "cross_modal_image_link_tower", "cross_modal_image_pooler", "cross_modal_image_transform", "cross_modal_ln_separate", "cross_modal_text", "cross_modal_text_layernorm", "cross_modal_text_layers", "cross_modal_text_link_tower", "cross_modal_text_pooler", "cross_modal_text_transform", "cross_text_features", "cross_text_features_", "data", "def", "device", "dim", "does", "dtype", "else", "embeddings", "encoder", "encoder_attention_mask", "eps", "expand_as", "extend_image_masks", "extend_text_masks", "for", "forward", "forward_post", "forward_pre", "full", "get_cls_features", "get_extended_attention_mask", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "i", "if", "image_embeds", "image_embeds_with_ln", "image_features", "image_link_tower", "image_token_type_embeddings", "image_token_type_idx", "in", "init_layernorm_from_vision_encoder", "input_ids", "input_shape", "inputs_embeds", "instead", "int", "interpolate_pos_encoding", "is", "labels", "layer", "layer_norm_eps", "layer_outputs_image", "layer_outputs_text", "len", "link_layer_index", "ln", "ln_post", "long", "nn", "not", "num_hidden_layers", "ones", "output_attentions", "output_hidden_states", "pass", "permute", "pixel_mask", "pixel_values", "pooler_output", "post_init", "r", "raise", "range", "resblocks", "return", "return_dict", "self", "set_input_embeddings", "share_cross_modal_transformer_layers", "share_layernorm", "share_link_tower_layers", "size", "split_index", "super", "sure", "text_config", "text_embeds", "text_features", "text_link_tower", "text_model", "text_token_type_embeddings", "to", "token_type_embeddings", "token_type_ids", "torch", "transformer", "tuple", "type", "use", "use_return_dict", "v", "value", "vision_config", "vision_model", "visual", "weight", "zeros"], "bridgetower/modeling_bridgetower.py:BridgeTowerPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "bridgetower/modeling_bridgetower.py:BridgeTowerMLMHead": ["False", "Linear", "ModelMLMHead", "ModelPredictionHeadTransform", "Module", "None", "Parameter", "__init__", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "if", "is", "mlm_score", "nn", "not", "return", "self", "super", "text_config", "torch", "transform", "vocab_size", "weight", "x", "zeros"], "bridgetower/modeling_bridgetower.py:BridgeTowerITMHead": ["Linear", "ModelITMHead", "Module", "__init__", "class", "def", "fc", "forward", "hidden_size", "itm_score", "nn", "return", "self", "super", "x"], "bridgetower/modeling_bridgetower.py:BridgeTowerForMaskedLM": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelMLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "decoder", "def", "device", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "image_embeds", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "mlm_logits", "mlm_score", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "post_init", "r", "return", "return_dict", "self", "set_output_embeddings", "super", "text_config", "text_features", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "bridgetower/modeling_bridgetower.py:BridgeTowerForImageAndTextRetrieval": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "Model", "ModelForImageAndTextRetrieval", "ModelITMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "def", "device", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "image_embeds", "input_ids", "inputs_embeds", "is", "itm_loss", "itm_score", "labels", "logits", "loss", "loss_fct", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_mask", "pixel_values", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict"], "bridgetower/modeling_bridgetower.py:BridgeTowerContrastiveHead": ["Linear", "ModelContrastiveHead", "Module", "__init__", "class", "def", "embed_size", "fc", "forward", "hidden_size", "nn", "return", "self", "super", "x"], "bridgetower/modeling_bridgetower.py:BridgeTowerForContrastiveLearning": ["FloatTensor", "LongTensor", "Model", "ModelContrastiveHead", "ModelContrastiveOutput", "ModelForContrastiveLearning", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "True", "Union", "__init__", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "class", "config", "contrastive_hidden_size", "cross_embeds", "cross_entropy", "cross_modal_image_transform", "def", "device", "dim", "dtype", "else", "exp", "expand_as", "forward", "forward_post", "full", "functional", "head_mask", "hidden_size", "hidden_states", "hidden_states_cross_modal", "hidden_states_img", "hidden_states_txt", "if", "image_embeds", "image_embeds_with_ln", "image_to_cross_loss", "image_token_type_embeddings", "input_ids", "inputs_embeds", "is", "itc_cross_modal_head", "itc_image_head", "itc_loss", "itc_text_head", "labels", "len", "logit_scale", "logit_scale_init_value", "logits", "logits_image_to_cross", "logits_text_to_cross", "logits_text_to_image", "long", "loss", "matmul", "nn", "normalize", "not", "output", "output_attentions", "output_hidden_states", "outputs", "p", "pixel_mask", "pixel_values", "pooler_output", "post_init", "r", "return", "return_dict", "return_loss", "self", "stack", "super", "t", "tensor", "text_embeds", "text_to_cross_loss", "text_to_image_loss", "to", "token_type_embeddings", "token_type_ids", "torch", "tuple", "use_return_dict", "vision_model", "visual", "weight"], "autoformer/modeling_autoformer.py:AutoFormerDecoderOutput": ["Cache", "FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "trend", "tuple"], "autoformer/modeling_autoformer.py:AutoformerModelOutput": ["Cache", "FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "last_hidden_state", "loc", "past_key_values", "r", "scale", "static_features", "torch", "trend", "tuple"], "autoformer/modeling_autoformer.py:AutoformerFeatureEmbedder": ["Embedding", "ModelFeatureEmbedder", "Module", "ModuleList", "None", "Tensor", "__init__", "c", "cardinalities", "cat", "cat_feature_slice", "cat_feature_slices", "chunk", "class", "d", "def", "dim", "else", "embed", "embedders", "embedding_dims", "features", "for", "forward", "if", "in", "int", "len", "list", "nn", "num_features", "return", "self", "squeeze", "super", "torch", "zip"], "autoformer/modeling_autoformer.py:AutoformerStdScaler": ["ModelConfig", "ModelStdScaler", "Module", "Tensor", "True", "__init__", "clamp_min", "class", "config", "data", "def", "denominator", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "minimum_scale", "nn", "observed_indicator", "return", "scale", "scaling_dim", "self", "sqrt", "sum", "super", "torch", "tuple", "variance"], "autoformer/modeling_autoformer.py:AutoformerMeanScaler": ["ModelConfig", "ModelMeanScaler", "Module", "None", "Tensor", "True", "__init__", "abs", "batch_observations", "batch_sum", "clamp", "class", "config", "data", "def", "default_scale", "dim", "else", "forward", "hasattr", "if", "is", "keepdim", "min", "minimum_scale", "nn", "not", "num_observed", "observed_indicator", "ones_like", "return", "scale", "scaled_data", "scaling_dim", "self", "squeeze", "sum", "super", "torch", "ts_sum", "tuple", "where", "zeros_like"], "autoformer/modeling_autoformer.py:AutoformerNOPScaler": ["False", "ModelConfig", "ModelNOPScaler", "Module", "None", "Optional", "Tensor", "True", "__init__", "class", "config", "data", "def", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "mean", "nn", "observed_indicator", "ones_like", "requires_grad", "return", "scale", "scaling_dim", "self", "super", "torch", "tuple", "zeros_like"], "autoformer/modeling_autoformer.py:weighted_average": ["Model_average", "Model_tensor", "None", "Optional", "Tensor", "clamp", "def", "dim", "else", "if", "input_tensor", "is", "mean", "min", "not", "return", "sum", "sum_weights", "torch", "weights", "where", "zeros_like"], "autoformer/modeling_autoformer.py:nll": ["Distribution", "Model", "Tensor", "def", "distributions", "input", "log_prob", "return", "target", "torch"], "autoformer/modeling_autoformer.py:AutoformerSinusoidalPositionalEmbedding": ["Embedding", "False", "FloatTensor", "ModelSinusoidalPositionalEmbedding", "None", "Optional", "Parameter", "Size", "Tensor", "__init__", "_init_weight", "arange", "array", "bsz", "class", "cos", "def", "device", "dim", "dtype", "else", "embedding_dim", "empty", "for", "forward", "if", "in", "input_ids_shape", "int", "is", "j", "long", "n_pos", "nn", "no_grad", "np", "num_positions", "out", "padding_idx", "past_key_values_length", "pos", "position_enc", "position_ids", "power", "range", "requires_grad", "return", "self", "sentinel", "seq_len", "shape", "sin", "super", "torch", "weight"], "autoformer/modeling_autoformer.py:AutoformerValueEmbedding": ["False", "Linear", "ModelValueEmbedding", "Module", "__init__", "bias", "class", "d_model", "def", "feature_size", "forward", "in_features", "nn", "out_features", "return", "self", "super", "value_projection", "x"], "autoformer/modeling_autoformer.py:AutoformerSeriesDecompositionLayer": ["AvgPool1d", "ModelConfig", "ModelSeriesDecompositionLayer", "Module", "__init__", "avg", "cat", "class", "config", "def", "dim", "end", "forward", "front", "kernel_size", "moving_average", "nn", "num_of_pads", "padding", "permute", "repeat", "return", "self", "stride", "super", "torch", "x", "x_padded", "x_seasonal", "x_trend"], "autoformer/modeling_autoformer.py:AutoformerLayernorm": ["LayerNorm", "ModelConfig", "ModelLayernorm", "Module", "__init__", "bias", "class", "config", "d_model", "def", "dim", "forward", "layernorm", "mean", "nn", "repeat", "return", "self", "shape", "super", "torch", "unsqueeze", "x", "x_hat"], "autoformer/modeling_autoformer.py:AutoformerAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Head", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "a", "and", "arange", "attention_mask", "attn_output", "attn_weights", "attn_weights_reshaped", "autocorrelation_factor", "autocorrelations", "autocorrelations_mean_on_bsz", "autocorrelations_mean_on_head_channel", "be", "bias", "bool", "bsz", "but", "by", "cache_position", "cat", "channel", "class", "conj", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "delays_agg", "deprecate_kwarg", "device", "dim", "dims", "divisible", "dropout", "else", "embed_dim", "f", "fft", "float", "for", "forward", "gather", "get", "got", "head_dim", "hidden_states", "i", "if", "in", "index", "init_index", "int", "irfft", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_states_fft", "key_value_states", "keys", "layer", "layer_head_mask", "layer_idx", "layers", "log", "mask", "math", "mean", "must", "n", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "past_key_value", "past_key_values", "proj_shape", "q_proj", "queries_time_length", "query_states", "query_states_fft", "raise", "range", "repeat", "reshape", "return", "rfft", "roll", "scaling", "self", "self_attention_cache", "shifts", "should", "single", "size", "softmax", "src_len", "stack", "super", "tgt_len", "time_length", "tmp_delay", "tmp_values", "to", "top_k", "top_k_autocorrelations", "top_k_autocorrelations_at_delay", "top_k_delays_index", "topk", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "value_states_roll_delay", "values", "values_time_length", "version", "view", "weights", "zeros", "zeros_like"], "autoformer/modeling_autoformer.py:AutoformerEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelLayernorm", "ModelSeriesDecompositionLayer", "Optional", "_", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_weights", "autocorrelation_factor", "bool", "clamp", "clamp_value", "class", "config", "d_model", "decomp1", "decomp2", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "layer_head_mask", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "autoformer/modeling_autoformer.py:AutoformerDecoderLayer": ["ACT2FN", "Cache", "Conv1d", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLayernorm", "ModelSeriesDecompositionLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "autocorrelation_factor", "bias", "bool", "cache_position", "circular", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "decomp1", "decomp2", "decomp3", "def", "deprecate_kwarg", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "feature_size", "final_layer_norm", "forward", "functional", "hidden_states", "if", "in_channels", "is", "is_decoder", "kernel_size", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "out_channels", "output_attentions", "outputs", "p", "padding", "padding_mode", "past_key_value", "past_key_values", "permute", "residual", "residual_trend", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "stride", "super", "torch", "training", "transpose", "trend1", "trend2", "trend3", "trend_projection", "tuple", "use_cache", "version"], "autoformer/modeling_autoformer.py:AutoformerPreTrainedModel": ["Conv1d", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "Module", "None", "PreTrainedModel", "Tensor", "True", "Union", "_attn_implementation", "_init_weight", "_init_weights", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "attention_mask", "base_model_prefix", "bias", "class", "config", "data", "def", "dtype", "elif", "else", "fill_", "flash", "flex_attention", "if", "in", "init_std", "inputs_embeds", "is", "is_causal", "isinstance", "main_input_name", "make_flex_block_causal_mask", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_values", "return", "sdpa", "self", "std", "supports_gradient_checkpointing", "torch", "weight", "zero_"], "autoformer/modeling_autoformer.py:AutoformerEncoder": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModelValueEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "_", "__init__", "_update_full_mask", "all_attentions", "attention_mask", "attentions", "be", "bool", "but", "class", "config", "context_length", "d_model", "def", "dropout", "dropout_probability", "else", "embed_pos", "embed_positions", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "feature_size", "for", "forward", "functional", "gradient_checkpointing", "head_mask", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "needs", "nn", "not", "output_attentions", "output_hidden_states", "p", "post_init", "prediction_length", "r", "raise", "rand", "range", "return", "return_dict", "self", "should", "size", "specified", "super", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "value_embedding"], "autoformer/modeling_autoformer.py:AutoformerDecoder": ["Cache", "DynamicCache", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelPreTrainedModel", "ModelSinusoidalPositionalEmbedding", "ModelValueEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "attention_mask", "attentions", "attn_mask", "be", "bool", "but", "cache_position", "checkpointing", "class", "config", "context_length", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "dropout", "dropout_probability", "dtype", "e", "else", "embed_pos", "embed_positions", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "feature_size", "for", "forward", "from_legacy_cache", "functional", "g", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "it", "label_length", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "needs", "nn", "not", "of", "output_attentions", "output_hidden_states", "p", "pass", "past_key_values", "past_key_values_length", "post_init", "prediction_length", "r", "raise", "rand", "range", "removed", "residual_trend", "return", "return_dict", "seasonality_projection", "self", "should", "size", "specified", "super", "tgt_len", "to", "torch", "training", "trend", "tuple", "use_cache", "use_return_dict", "v", "v4", "value_embedding", "warning", "warning_once", "will", "with", "zip"], "autoformer/modeling_autoformer.py:AutoformerModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderOutput", "ModelEncoder", "ModelFeatureEmbedder", "ModelMeanScaler", "ModelModel", "ModelModelOutput", "ModelNOPScaler", "ModelPreTrainedModel", "ModelSeriesDecompositionLayer", "ModelStdScaler", "None", "Optional", "Tensor", "True", "Union", "ValueError", "_", "__init__", "_past_length", "abs", "and", "append", "attention_mask", "attentions", "auto_docstring", "begin_index", "bool", "cache_position", "cannot", "cardinalities", "cardinality", "cat", "class", "config", "context", "context_length", "create_network_inputs", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input", "decoder_outputs", "decomposition_layer", "def", "device", "dim", "does", "elif", "else", "embedded_cat", "embedder", "embedding_dimension", "embedding_dims", "enc_input", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_index", "expand", "expanded_static_feat", "f", "feature", "features", "for", "forward", "found", "further", "future_time_features", "future_values", "get_encoder", "get_lagged_subsequences", "go", "head_mask", "hidden_states", "history", "if", "in", "indices", "input", "input_size", "inputs", "inputs_embeds", "int", "is", "isinstance", "label_length", "lag", "lag_index", "lagged_sequence", "lagged_values", "lags", "lags_sequence", "lags_shape", "last_hidden_state", "len", "length", "lengths", "list", "loc", "log", "log1p", "log_abs_loc", "log_scale", "match", "max", "mean", "not", "num_static_categorical_features", "observed_context", "ones_like", "only", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_observed_mask", "past_time_features", "past_values", "post_init", "prediction_length", "property", "r", "raise", "repeat", "reshape", "reshaped_lagged_sequence", "return", "return_dict", "scale", "scaler", "scaling", "seasonal_input", "self", "sequence", "sequence_length", "shape", "shift", "squeeze", "stack", "static_categorical_features", "static_feat", "static_features", "static_real_features", "std", "subsequences_length", "super", "temporal_features", "than", "time", "time_feat", "torch", "transformer_inputs", "trend", "trend_init", "trend_input", "tuple", "unsqueeze", "use_cache", "use_return_dict", "while", "zeros"], "autoformer/modeling_autoformer.py:AutoformerForPrediction": ["Cache", "Distribution", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelForPrediction", "ModelModel", "ModelPreTrainedModel", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleTSPredictionOutput", "Seq2SeqTSPredictionOutput", "StudentTOutput", "Tensor", "True", "Union", "Unknown", "ValueError", "_", "__init__", "auto_docstring", "bool", "cat", "class", "config", "context_length", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input", "decoder_last_hidden", "decoder_output", "decoder_outputs", "decomposition_layer", "def", "device", "dim", "distr", "distribution", "distribution_output", "distributions", "elif", "else", "enc_last_hidden", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "event_shape", "expand", "expanded_static_feat", "f", "feature_size", "features", "for", "forward", "function", "future_observed_mask", "future_samples", "future_time_features", "future_values", "generate", "get_decoder", "get_encoder", "get_lagged_subsequences", "get_parameter_projection", "head_mask", "if", "ignore", "in", "input_size", "inputs_embeds", "is", "jit", "keepdim", "label_length", "lagged_sequence", "lags_shape", "last_hidden_state", "len", "list", "loc", "loss", "loss_weights", "mean", "min", "model", "negative_binomial", "nll", "no_grad", "normal", "not", "num_parallel_samples", "ones_like", "output", "output_attentions", "output_distribution", "output_hidden_states", "output_params", "outputs", "p", "parameter_projection", "params", "past_key_values", "past_observed_mask", "past_time_features", "past_values", "post_init", "prediction_length", "prediction_loss", "r", "raise", "repeat", "repeat_interleave", "repeated_enc_last_hidden", "repeated_features", "repeated_loc", "repeated_past_values", "repeated_scale", "repeats", "reshape", "reshaped_lagged_sequence", "return", "return_dict", "sample", "scale", "seasonal_input", "self", "sequence", "sequences", "shape", "sliced_params", "static_categorical_features", "static_feat", "static_features", "static_real_features", "student_t", "subsequences_length", "super", "target_shape", "time_features", "torch", "trailing_n", "trend", "trend_init", "trend_input", "tuple", "unsqueeze", "use_cache", "use_return_dict", "weighted_average", "weights", "zeros"], "granitemoehybrid/modeling_granitemoehybrid.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "granitemoehybrid/modeling_granitemoehybrid.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "granitemoehybrid/modeling_granitemoehybrid.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "granitemoehybrid/modeling_granitemoehybrid.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_multiplier", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "errors", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "return", "scaling", "self", "sin", "size", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "granitemoehybrid/modeling_granitemoehybrid.py:HybridMambaAttentionDynamicCache": ["Any", "False", "LongTensor", "ModelConfig", "ModelMambaAttentionDynamicCache", "None", "Optional", "Tensor", "_", "__init__", "append", "batch_size", "beam_idx", "cache_kwargs", "cat", "class", "config", "conv_kernel_size", "conv_states", "def", "device", "dict", "dim", "dtype", "else", "float16", "for", "get_seq_length", "has_previous_state", "hidden_size", "i", "if", "in", "index_select", "int", "is_compileable", "key_cache", "key_states", "layer_idx", "layers_block_type", "len", "mamba", "mamba_d_conv", "mamba_d_head", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "not", "num_hidden_layers", "range", "reorder_cache", "return", "self", "shape", "ssm_state_size", "ssm_states", "str", "tensor", "to", "torch", "transformer_layers", "tuple", "update", "value_cache", "value_states", "zeros"], "granitemoehybrid/modeling_granitemoehybrid.py:pad_tensor_by_size": ["Model", "Model_shape", "Model_size", "Model_tensor_by_size", "Tensor", "constant", "def", "else", "functional", "if", "input_tensor", "int", "len", "mode", "nn", "return", "shape", "torch", "value"], "granitemoehybrid/modeling_granitemoehybrid.py:reshape_into_chunks": ["Model", "Model_into_chunks", "chunk_size", "def", "else", "if", "input_tensor", "len", "pad_size", "pad_tensor_by_size", "return", "shape"], "granitemoehybrid/modeling_granitemoehybrid.py:segment_sum": ["Model_sum", "None", "bool", "chunk_size", "cumsum", "def", "device", "diagonal", "dim", "dtype", "expand", "inf", "input_tensor", "mask", "masked_fill", "ones", "return", "size", "tensor_segsum", "torch", "tril"], "granitemoehybrid/modeling_granitemoehybrid.py:apply_mask_to_padding_states": ["Model_mask_to_padding_states", "None", "and", "attention_mask", "def", "dtype", "hidden_states", "if", "is", "not", "return", "shape", "to"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridMambaLayer": ["A", "ACT2FN", "AILab", "A_cumsum", "A_log", "B", "B_decay", "C", "C_reshaped", "C_times_states", "Conv1d", "D", "D_residual", "Dao", "Falling", "False", "G", "GPU", "G_intermediate", "HybridMambaAttentionDynamicCache", "IntTensor", "L", "Linear", "LongTensor", "M", "M_intermediate", "Model", "ModelConfig", "ModelMambaLayer", "ModelRMSNormGated", "Module", "None", "NotImplementedError", "Optional", "Parameter", "Please", "Tensor", "The", "To", "True", "Y_diag", "Y_off", "_", "__init__", "a", "act", "activation", "and", "apply_mask_to_padding_states", "arange", "attention_mask", "available", "back", "batch_size", "be", "because", "bias", "bmm", "cache_device", "cache_params", "cache_position", "cat", "causal", "causal_conv1d", "causal_conv1d_fn", "causal_conv1d_update", "chunk_size", "clamp", "class", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_dim", "conv_kernel_size", "conv_states", "copy_", "cuda", "cuda_kernels_forward", "cumsum", "dA", "dB", "dBx", "decay_chunk", "decay_states", "def", "device", "dim", "dims", "dt", "dt_bias", "dt_limit", "dt_limit_kwargs", "dt_softplus", "dtype", "else", "eps", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "groups_time_state_size", "has_previous_state", "head_dim", "headdim", "hidden_act", "hidden_size", "hidden_states", "hidden_states_B_C", "hidden_states_B_C_transposed", "hidden_states_reshaped", "https", "if", "implementation", "in", "in_channels", "in_proj", "inf", "input_states", "install", "int", "intermediate_size", "is", "is_fast_path_available", "kernel_size", "kwargs", "layer_idx", "layer_norm_epsilon", "log", "logger", "mamba", "mamba_chunk_scan_combined", "mamba_chunk_size", "mamba_conv_bias", "mamba_d_conv", "mamba_d_head", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "mamba_proj_bias", "mamba_split_conv1d_scan_combined", "mamba_ssm", "model", "n_groups", "naive", "new_states", "ngroups", "nn", "norm", "norm_before_gate", "not", "num_heads", "of", "on", "one", "ones", "out", "out_channels", "out_proj", "outproj_bias", "outproj_weight", "output_size", "pad", "pad_size", "pad_tensor_by_size", "padding", "path", "permute", "previous_states", "projected_states", "projection_size", "raise", "repeat_interleave", "requires", "reshape", "reshape_into_chunks", "return", "return_final_states", "rms_norm_eps", "rmsnorm_eps", "rmsnorm_weight", "roll", "running", "scan_output", "segment_sum", "selective_state_update", "self", "seq_idx", "seq_len", "shape", "shifts", "silu", "softplus", "spaces", "split", "squeeze", "ssm_state", "ssm_state_size", "ssm_states", "ssm_states_reshaped", "state", "state_decay_out", "state_decay_out_permuted", "states", "sum", "super", "support", "swish", "t", "the", "time_step_limit", "time_step_max", "time_step_min", "to", "torch", "torch_forward", "training", "transpose", "type", "use_bias", "use_conv_bias", "use_precomputed_states", "used", "variance_epsilon", "view", "warning_once", "weight", "when", "will", "x", "y", "z", "zeros_like"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridRMSNormGated": ["ModelRMSNormGated", "Module", "None", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "functional", "gate", "hidden_size", "hidden_states", "if", "input_dtype", "is", "keepdim", "mean", "nn", "not", "ones", "pow", "return", "rsqrt", "self", "silu", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridMLP": ["ACT2FN", "False", "Linear", "ModelConfig", "ModelMLP", "Module", "Tensor", "__init__", "activation", "bias", "chunk", "chunked_hidden_states", "class", "config", "def", "dim", "forward", "hidden_act", "hidden_size", "hidden_states", "input_linear", "input_size", "nn", "output_linear", "return", "self", "shared_intermediate_size", "super", "torch"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteFlashAttentionKwargs": ["False", "IntTensor", "LongTensor", "ModelFlashAttentionKwargs", "TypedDict", "class", "cu_seq_lens_k", "cu_seq_lens_q", "int", "max_length_k", "max_length_q", "seq_idx", "torch", "total"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridParallelExperts": ["F", "ModelParallelExperts", "Module", "None", "Parameter", "__init__", "append", "cat", "class", "def", "dim", "empty", "expert_size", "for", "forward", "i", "in", "input_list", "input_size", "inputs", "int", "linear", "nn", "num_experts", "output_list", "output_size", "range", "results", "return", "self", "split", "super", "torch", "weight"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridTopKGating": ["False", "Linear", "ModelTopKGating", "Module", "_", "__init__", "batch_gates", "batch_index", "bias", "class", "def", "device", "dim", "div", "dtype", "expert_size", "flatten", "float", "forward", "gates", "hidden_states", "index_sorted_experts", "input_size", "int", "layer", "logits", "long", "nn", "num_experts", "return", "rounding_mode", "scatter", "self", "size", "softmax", "sort", "sum", "super", "tolist", "top_k", "top_k_experts", "top_k_gates", "top_k_indices", "top_k_logits", "topk", "torch", "trunc", "type_as", "zeros"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridMoE": ["ACT2FN", "ModelConfig", "ModelMoE", "ModelParallelExperts", "ModelTopKGating", "Module", "None", "_", "__init__", "activation", "batch_gates", "batch_index", "bsz", "chunk", "chunked_hidden_states", "class", "config", "def", "device", "dim", "dtype", "emb_size", "expert_inputs", "expert_outputs", "expert_size", "forward", "hidden_act", "hidden_size", "hidden_states", "index_add", "input_linear", "input_size", "intermediate_size", "layer_input", "layer_output", "length", "nn", "num_experts", "num_experts_per_tok", "num_local_experts", "output_linear", "reshape", "return", "router", "router_logits", "self", "size", "super", "top_k", "torch", "view", "zeros"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFlashAttentionKwargs", "ModelMLP", "ModelMambaLayer", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "block_sparse_moe", "bool", "cache_params", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "getattr", "has_experts", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is", "kwargs", "layer_idx", "layer_type", "layers_block_type", "mamba", "moe_hidden_states", "new_name", "not", "num_local_experts", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_embeddings", "post_attention_layernorm", "residual", "residual_multiplier", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "shared_mlp", "super", "torch", "tuple", "use_cache", "version"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridPreTrainedModel": ["A_log", "D", "False", "ModelConfig", "ModelDecoderLayer", "ModelMambaLayer", "ModelParallelExperts", "ModelPreTrainedModel", "ModelRMSNormGated", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_is_stateful", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "arange", "base_model_prefix", "class", "config", "data", "def", "dt_bias", "elif", "fill_", "if", "initializer_range", "isinstance", "log", "mean", "model", "module", "normal_", "num_heads", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "torch", "weight"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "Because", "BlockMask", "Cache", "Embedding", "False", "FloatTensor", "HybridMambaAttentionDynamicCache", "LongTensor", "Model", "ModelConfig", "ModelDecoderLayer", "ModelFlashAttentionKwargs", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "_update_mamba_mask", "a", "all", "all_hidden_states", "all_router_logits", "all_self_attns", "an", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "cache", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "embedding_multiplier", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "has_previous_state", "head_dim", "hidden_size", "hidden_states", "if", "in", "incompatible", "initialized", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_mask", "layer_outputs", "layer_type", "layers", "list", "logger", "make_flex_block_causal_mask", "mamba", "mamba_mask", "mask_length", "masked_fill", "max_position_embeddings", "min", "min_dtype", "must", "nn", "no", "norm", "not", "npu", "num_attention_heads", "num_heads", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embedding_type", "position_embeddings", "position_ids", "post_init", "provided", "raise", "range", "requires", "reshape", "return", "return_dict", "returned", "rms_norm_eps", "rope", "rope_theta", "rotary_emb", "router_logits", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "vocab_size", "warning_once", "was", "will", "with", "xpu"], "granitemoehybrid/modeling_granitemoehybrid.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "index", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "granitemoehybrid/modeling_granitemoehybrid.py:GraniteMoeHybridForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "HybridMambaAttentionDynamicCache", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "class", "config", "contiguous", "cumsum", "def", "device", "dtype", "elif", "else", "empty_past_kv", "float", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "key", "kwargs", "labels", "list", "lm_head", "load_balancing_loss_func", "logits", "logits_scaling", "logits_to_keep", "long", "loss", "loss_function", "masked_fill_", "model", "model_inputs", "nn", "not", "num_experts", "num_experts_per_tok", "num_local_experts", "or", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "router_aux_loss_coef", "router_logits", "self", "shape", "slice", "slice_indices", "super", "to", "torch", "tuple", "update", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLModelOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelModelOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLRotaryEmbedding": ["False", "ModelRotaryEmbedding", "ModelTextConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen2_vl/modeling_qwen2_vl.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen2_vl/modeling_qwen2_vl.py:apply_multimodal_rotary_pos_emb": ["Model_multimodal_rotary_pos_emb", "cat", "cos", "def", "dim", "enumerate", "for", "i", "in", "k", "k_embed", "m", "mrope_section", "q", "q_embed", "return", "rotate_half", "sin", "split", "torch", "unsqueeze", "unsqueeze_dim"], "qwen2_vl/modeling_qwen2_vl.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "k", "k_embed", "orig_k_dtype", "orig_q_dtype", "q", "q_embed", "return", "rotate_half", "sin", "to", "torch", "tuple", "unsqueeze"], "qwen2_vl/modeling_qwen2_vl.py:VisionRotaryEmbedding": ["False", "ModelRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "qwen2_vl/modeling_qwen2_vl.py:PatchEmbed": ["Conv3d", "False", "ModelEmbed", "Model_size", "Module", "None", "Tensor", "__init__", "bias", "class", "def", "dtype", "embed_dim", "forward", "hidden_states", "in_channels", "int", "kernel_size", "nn", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_Model_size", "to", "torch", "view", "weight"], "qwen2_vl/modeling_qwen2_vl.py:PatchMerger": ["GELU", "LayerNorm", "Linear", "ModelMerger", "Module", "None", "Sequential", "Tensor", "__init__", "class", "context_dim", "def", "dim", "eps", "forward", "hidden_size", "int", "ln_q", "mlp", "nn", "return", "self", "spatial_merge_size", "super", "torch", "view", "x"], "qwen2_vl/modeling_qwen2_vl.py:VisionMlp": ["ACT2FN", "Linear", "ModelMlp", "Module", "None", "Tensor", "__init__", "act", "class", "def", "dim", "fc1", "fc2", "forward", "hidden_act", "hidden_dim", "int", "nn", "return", "self", "str", "super", "torch", "x"], "qwen2_vl/modeling_qwen2_vl.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen2_vl/modeling_qwen2_vl.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen2_vl/modeling_qwen2_vl.py:VisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "ModelModelConfig", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_Model", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cos", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "flash_attention_2", "for", "forward", "head_dim", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "permute", "position_embeddings", "proj", "q", "qkv", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "sin", "split", "splits", "super", "tensor", "tolist", "torch", "training", "transpose", "tuple", "unbind", "unsqueeze", "v", "value_states", "zip"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLVisionBlock": ["GradientCheckpointingLayer", "LayerNorm", "ModelVisionBlock", "None", "Optional", "Tensor", "VisionAttention", "VisionMlp", "__init__", "attn", "attn_implementation", "class", "config", "cu_seqlens", "def", "dim", "embed_dim", "eps", "forward", "hidden_act", "hidden_dim", "hidden_states", "int", "kwargs", "mlp", "mlp_hidden_dim", "mlp_ratio", "norm1", "norm2", "position_embeddings", "return", "rotary_pos_emb", "sdpa", "self", "str", "super", "torch", "tuple"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelRotaryEmbedding", "ModelTextConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "apply_multimodal_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "errors", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "logger", "make", "mrope_section", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "rope_scaling", "rotary_emb", "scaling", "self", "sin", "size", "sliding_attention", "sliding_window", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLDecoderLayer": ["Attention", "Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelTextConfig", "None", "Optional", "Sliding", "Tensor", "Unpack", "Window", "__init__", "_attn_implementation", "and", "attention_mask", "attention_type", "be", "bool", "but", "cache_position", "class", "config", "def", "deprecate_kwarg", "enabled", "encountered", "eps", "f", "flash_attention_2", "for", "forward", "hidden_size", "hidden_states", "if", "implemented", "input_layernorm", "int", "is", "kwargs", "layer_idx", "layer_types", "logger", "may", "mlp", "new_name", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "results", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "unexpected", "use_cache", "use_sliding_window", "version", "warning_once"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLPreTrainedModel": ["ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VisionTransformerPretrainedModel": ["F", "False", "ModelPreTrainedModel", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionTransformerPretrainedModel", "ModuleList", "None", "PatchEmbed", "PatchMerger", "Tensor", "VisionRotaryEmbedding", "_", "__init__", "_no_split_modules", "append", "arange", "auto_docstring", "blk", "blocks", "cat", "class", "config", "context_dim", "cos", "cu_seqlens", "cumsum", "def", "depth", "device", "dim", "dtype", "else", "emb", "embed_dim", "expand", "fc2", "flatten", "for", "forward", "get_device", "get_dtype", "gradient_checkpointing", "grid_thw", "h", "head_dim", "hidden_size", "hidden_states", "hpos_ids", "if", "in", "in_channels", "int32", "is_tracing", "jit", "kwargs", "max", "max_grid_size", "merger", "mlp", "nn", "num_heads", "pad", "patch_embed", "patch_size", "permute", "pos_ids", "position_embeddings", "r", "range", "repeat", "repeat_interleave", "reshape", "return", "rot_pos_emb", "rotary_pos_emb", "rotary_pos_emb_full", "self", "sin", "spatial_merge_size", "stack", "super", "t", "temporal_patch_size", "torch", "unsqueeze", "value", "w", "weight", "wpos_ids"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModelTextConfig", "ModelTextModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "checkpointing", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "elif", "else", "embed_tokens", "eps", "exactly", "expand", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "isinstance", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layer_types", "layers", "logger", "mask_kwargs", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "return_dict", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "text_position_ids", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warning_once", "with"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLModel": ["Cache", "False", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelPreTrainedModel", "ModelTextModel", "ModelVisionTransformerPretrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "_", "__init__", "_checkpoint_conversion_mapping", "_from_config", "accepts_loss_kwargs", "all", "and", "append", "arange", "argwhere", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "batch_size", "bool", "cache_position", "cat", "class", "config", "cumsum", "decoder", "def", "delta", "device", "dim", "do", "dtype", "ed", "ed_image", "ed_video", "else", "enumerate", "expand", "expand_as", "f", "features", "flatten", "for", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_video_features", "grid_thw", "h", "h_index", "hidden_states", "i", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_index", "image_mask", "image_nums", "image_token_id", "in", "index", "input_ids", "input_tokens", "inputs_embeds", "is", "item", "keepdim", "kwargs", "language_model", "last_hidden_state", "len", "list", "llm_grid_h", "llm_grid_t", "llm_grid_w", "llm_pos_ids_list", "llm_positions", "long", "masked_fill_", "masked_scatter", "match", "max", "max_position_ids", "model", "mrope_position_deltas", "n_image_tokens", "n_video_tokens", "not", "numel", "ones", "ones_like", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prod", "r", "raise", "range", "remain_images", "remain_videos", "repeat_interleave", "reshape", "return", "return_dict", "rope_deltas", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "spatial_merge_size", "special_image_mask", "special_video_mask", "split", "split_sizes", "squeeze", "st", "st_idx", "stack", "sum", "super", "t", "t_index", "tensor", "text_config", "text_len", "to", "to_tuple", "tokens", "tolist", "torch", "total_input_ids", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "value", "video", "video_embeds", "video_features", "video_grid_thw", "video_index", "video_mask", "video_nums", "video_token_id", "view", "vision_config", "vision_start_indices", "vision_start_token_id", "vision_tokens", "visual", "w", "w_index", "zeros"], "qwen2_vl/modeling_qwen2_vl.py:Qwen2VLForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "If", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "__init__", "_checkpoint_conversion_mapping", "_expand_dict_for_generation", "_expand_dict_for_generation_visual", "_expand_inputs_for_generation", "_get_image_nums_and_video_nums", "_repeat_interleave_samples", "_tied_weights_keys", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "decoder", "def", "defined", "delta", "device", "dict", "dict_to_expand", "dim", "dims", "dtype", "elif", "else", "encoder_outputs", "expand", "expand_as", "expand_size", "for", "forward", "get", "get_decoder", "get_image_features", "get_input_embeddings", "get_rope_index", "get_seq_length", "get_video_features", "hidden_size", "hidden_states", "if", "image_grid_thw", "image_mask", "image_nums", "image_token_id", "in", "input_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "is_torchdynamo_compiling", "isinstance", "key", "kwargs", "labels", "language_model", "lengths", "list", "lm_head", "logits", "long", "loss", "loss_function", "make", "model", "model_inputs", "model_kwargs", "nn", "not", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prefill_compiled_stage", "prefill_noncompiled_stage", "prepare_inputs_for_generation", "prod", "property", "r", "raise", "repeat", "repeat_args", "repeat_interleave", "repeat_times", "result", "return", "return_dict", "roll", "rope_deltas", "sample", "samples", "second_per_grid_ts", "self", "seq_length", "set_decoder", "set_input_embeddings", "shape", "shifts", "split", "str", "sum", "super", "sure", "tensor", "text_config", "text_positions", "that", "torch", "tuple", "use_cache", "value", "video_grid_thw", "video_mask", "video_nums", "video_token_id", "view", "vision_first_mask", "vision_positions", "vision_start_mask", "vision_start_token_id", "visual", "visual_keys", "vocab_size", "weight", "x"], "dbrx/modeling_dbrx.py:DbrxRotaryEmbedding": ["False", "ModelRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "autocast", "base", "cat", "class", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "if", "int64", "inv_freq", "inv_freq_expanded", "max_position_embeddings", "mps", "nn", "no_grad", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "self", "seq_len", "shape", "sin", "super", "tensor", "to", "torch", "transpose", "type", "with", "x"], "dbrx/modeling_dbrx.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "dbrx/modeling_dbrx.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "dbrx/modeling_dbrx.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "dbrx/modeling_dbrx.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "_", "attention_mask", "batch_size", "cat", "compute_device", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_probabilities", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "sum", "tensor", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "dbrx/modeling_dbrx.py:DbrxAttention": ["Any", "Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "Wqkv", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_mask", "attn_config", "attn_output", "attn_pdrop", "attn_weights", "base", "be", "bias", "block_idx", "bool", "bsz", "but", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "clamp", "class", "clip_qkv", "config", "contiguous", "cos", "creating", "d_model", "def", "deprecate_kwarg", "dim", "dropout", "dtype", "during", "else", "errors", "f", "float32", "forward", "functional", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key_states", "kv_n_heads", "kwargs", "lead", "logger", "make", "math", "matmul", "max", "max_position_embeddings", "max_seq_len", "max_val", "min", "min_val", "n_heads", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "of", "out_proj", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "qkv_states", "query_states", "raise", "recommended", "repeat_kv", "reshape", "return", "rope_theta", "rotary_emb", "self", "shape", "should", "sin", "size", "softmax", "split", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "value_states", "version", "view", "warning_once", "when", "will", "without"], "dbrx/modeling_dbrx.py:DbrxFlashAttention2": ["Any", "Attention", "Cache", "False", "Flash", "Implicitly", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "StaticCache", "Tensor", "The", "ValueError", "We", "Wqkv", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "an", "and", "apply_rotary_pos_emb", "args", "as", "at", "attention_mask", "attn_implementation", "attn_output", "attn_pdrop", "attn_weights", "back", "be", "block_idx", "bool", "bsz", "cache", "cache_kwargs", "cache_position", "cast", "casted", "clamp", "class", "clip_qkv", "com", "compatible", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dim", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attention_2", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "github", "hasattr", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "https", "huggingface", "if", "implementation", "in", "info", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "isinstance", "issue", "it", "key_states", "kwargs", "layer", "layers", "logger", "make", "max", "mean", "might", "min", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "open", "or", "out_proj", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "qkv_states", "query_states", "raise", "related", "reshape", "return", "rotary_emb", "sdpa", "seems", "self", "setting", "silently", "sin", "size", "split", "states", "static", "super", "supported", "sure", "target_dtype", "the", "this", "time", "to", "torch", "training", "transformers", "transpose", "tuple", "type", "upcasted", "update", "use", "use_cache", "use_top_left_mask", "value_states", "version", "view", "warning_once", "will", "with", "you"], "dbrx/modeling_dbrx.py:DbrxSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "Wqkv", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "attn_pdrop", "back", "be", "block_idx", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "clamp", "class", "clip_qkv", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "dim", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_size", "hidden_states", "if", "implementation", "is", "is_causal", "key_states", "loading", "logger", "manual", "max", "min", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "onwards", "out_proj", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "qkv_states", "query_states", "removed", "repeat_kv", "required", "return", "rotary_emb", "scaled_dot_product_attention", "self", "seq_len", "shape", "sin", "size", "specifying", "split", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "dbrx/modeling_dbrx.py:DbrxNormAttentionNorm": ["Any", "Cache", "False", "LayerNorm", "LongTensor", "ModelConfig", "ModelNormAttentionNorm", "Model_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "attention_mask", "attn", "attn_weights", "bias", "block_idx", "bool", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "dtype", "forward", "functional", "hidden_states", "int", "kwargs", "new_name", "nn", "norm_1", "norm_2", "output_attentions", "p", "past_key_value", "past_key_values", "position_ids", "resid_pdrop", "residual_states", "return", "self", "super", "to", "torch", "training", "tuple", "use_cache", "version"], "dbrx/modeling_dbrx.py:DbrxRouter": ["False", "Linear", "LongTensor", "ModelRouter", "Module", "None", "Optional", "Tensor", "True", "__init__", "and", "bias", "class", "def", "dim", "dtype", "else", "empty_like", "float", "float32", "forward", "hidden_size", "hidden_states", "if", "int", "is", "keepdim", "layer", "moe_jitter_eps", "moe_normalize_expert_weights", "moe_num_experts", "moe_top_k", "nn", "norm", "not", "p", "return", "self", "shape", "softmax", "super", "to", "top_experts", "top_weights", "top_weights_scale", "topk", "torch", "training", "tuple", "uniform_", "view", "weights"], "dbrx/modeling_dbrx.py:DbrxExpertGLU": ["ACT2FN", "ModelExpertGLU", "Module", "Parameter", "Tensor", "__init__", "act_fn_name", "activation_fn", "class", "def", "dict", "down_proj", "empty", "expert_v1", "expert_w1", "expert_w2", "ffn_act_fn", "ffn_hidden_size", "forward", "gate_proj", "get", "hidden_size", "int", "intermediate_states", "matmul", "moe_num_experts", "name", "nn", "return", "self", "silu", "super", "t", "torch", "up_proj", "v1", "w1", "w2", "x"], "dbrx/modeling_dbrx.py:DbrxExperts": ["LongTensor", "ModelExpertGLU", "ModelExperts", "Module", "None", "Tensor", "__init__", "bsz", "chunk", "class", "continue", "def", "dict", "dim", "expert_idx", "expert_mask", "expert_out", "expert_tokens", "ffn_act_fn", "ffn_hidden_size", "for", "forward", "functional", "hidden_size", "if", "in", "index_add_", "int", "mlp", "moe_num_experts", "nn", "num_classes", "one_hot", "out", "permute", "q_len", "range", "reshape", "return", "self", "shape", "squeeze", "super", "token_idx", "token_list", "top_experts", "top_weights", "topk_idx", "topk_list", "torch", "v1", "v1_chunked", "view", "w1", "w1_chunked", "w2", "w2_chunked", "weights", "where", "x", "zeros_like"], "dbrx/modeling_dbrx.py:DbrxFFN": ["ModelConfig", "ModelExperts", "ModelFFN", "ModelRouter", "Module", "Tensor", "__init__", "class", "config", "d_model", "def", "experts", "ffn_act_fn", "ffn_config", "ffn_hidden_size", "forward", "hidden_size", "moe_jitter_eps", "moe_normalize_expert_weights", "moe_num_experts", "moe_top_k", "nn", "out", "return", "router", "self", "super", "top_experts", "top_weights", "torch", "tuple", "weights", "x"], "dbrx/modeling_dbrx.py:DbrxBlock": ["Any", "Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelBlock", "ModelConfig", "ModelFFN", "ModelNormAttentionNorm", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "block_idx", "bool", "cache_position", "class", "config", "d_model", "def", "deprecate_kwarg", "dropout", "ffn", "forward", "functional", "hidden_size", "hidden_states", "if", "int", "kwargs", "new_name", "nn", "norm_attn_norm", "output_attentions", "output_router_logits", "outputs", "p", "past_key_value", "past_key_values", "position_ids", "resid_pdrop", "resid_states", "return", "router_logits", "self", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "dbrx/modeling_dbrx.py:DbrxPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "ModelBlock", "ModelConfig", "ModelExpertGLU", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "transformer", "v1", "w1", "w2", "weight", "zero_"], "dbrx/modeling_dbrx.py:DbrxModel": ["AttentionMaskConverter", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "LayerNorm", "LongTensor", "ModelBlock", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_router_logits", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "block", "block_idx", "block_outputs", "blocks", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "d_model", "def", "device", "diagonal", "dim", "dropout", "dtype", "else", "emb_pdrop", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "functional", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "n_layers", "nn", "norm_f", "not", "npu", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "p", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_ids", "post_init", "raise", "range", "reshape", "return", "return_dict", "router_logits", "sdpa", "self", "sequence_length", "set_input_embeddings", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "value", "vocab_size", "warning_once", "with", "wte", "xpu"], "dbrx/modeling_dbrx.py:DbrxForCausalLM": ["Cache", "Embedding", "False", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "None", "Optional", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "class", "config", "decoder", "def", "device", "else", "ffn_config", "forward", "get_decoder", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "moe_loss_weight", "moe_num_experts", "moe_top_k", "new_embeddings", "nn", "not", "num_experts", "num_experts_per_tok", "output", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "router_logits", "self", "set_decoder", "set_input_embeddings", "set_output_embeddings", "slice", "slice_indices", "super", "to", "torch", "transformer", "tuple", "use_cache", "use_return_dict", "value", "vocab_size"], "deberta/modeling_deberta.py:DebertaLayerNorm": ["ModelLayerNorm", "Module", "Parameter", "True", "__init__", "bias", "class", "def", "dtype", "eps", "float", "forward", "hidden_states", "input_type", "keepdim", "mean", "nn", "ones", "pow", "return", "self", "size", "sqrt", "super", "to", "torch", "variance", "variance_epsilon", "weight", "y", "zeros"], "deberta/modeling_deberta.py:DebertaSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelLayerNorm", "ModelSelfOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super"], "deberta/modeling_deberta.py:build_relative_position": ["Model_relative_position", "None", "arange", "def", "device", "dtype", "k_ids", "key_layer", "key_size", "long", "q_ids", "query_layer", "query_size", "rel_pos_ids", "repeat", "return", "size", "torch", "unsqueeze", "view"], "deberta/modeling_deberta.py:c2p_dynamic_expand": ["Model_dynamic_expand", "Model_pos", "def", "expand", "query_layer", "relative_pos", "return", "size"], "deberta/modeling_deberta.py:p2c_dynamic_expand": ["Model_dynamic_expand", "c2p_pos", "def", "expand", "key_layer", "query_layer", "return", "size"], "deberta/modeling_deberta.py:pos_dynamic_expand": ["Model_dynamic_expand", "Model_index", "def", "expand", "key_layer", "p2c_att", "return", "size"], "deberta/modeling_deberta.py:scaled_size_sqrt": ["Model_size_sqrt", "Tensor", "def", "dtype", "float", "int", "query_layer", "return", "scale_factor", "size", "sqrt", "tensor", "torch"], "deberta/modeling_deberta.py:build_rpos": ["Model_relative_position", "Model_rpos", "Tensor", "def", "else", "if", "key_layer", "query_layer", "relative_pos", "return", "size", "torch"], "deberta/modeling_deberta.py:compute_attention_span": ["Model_attention_span", "Tensor", "def", "int", "key_layer", "max", "max_relative_positions", "min", "query_layer", "return", "size", "tensor", "torch"], "deberta/modeling_deberta.py:uneven_size_corrected": ["Model_size_corrected", "Tensor", "def", "dim", "else", "gather", "if", "index", "key_layer", "p2c_att", "pos_dynamic_expand", "pos_index", "query_layer", "relative_pos", "return", "size", "torch", "unsqueeze"], "deberta/modeling_deberta.py:DisentangledSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Model_att_bias", "Module", "None", "Optional", "Parameter", "Relative", "Tensor", "The", "ValueError", "__init__", "a", "all_head_size", "and", "att_span", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "be", "bias", "bool", "build_relative_position", "build_rpos", "c2p", "c2p_att", "c2p_dynamic_expand", "c2p_pos", "cat", "chunk", "clamp", "class", "compute_attention_span", "config", "context_layer", "contiguous", "def", "device", "dim", "dropout", "dtype", "elif", "else", "f", "finfo", "float", "for", "forward", "functional", "gather", "getattr", "head_logits_proj", "head_weights_proj", "heads", "hidden", "hidden_dropout_prob", "hidden_size", "hidden_states", "i", "ids", "if", "in", "in_proj", "index", "int", "is", "k", "key_layer", "len", "long", "masked_fill", "matmul", "max_position_embeddings", "max_relative_positions", "min", "multiple", "must", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "p2c", "p2c_att", "p2c_dynamic_expand", "p2c_pos", "permute", "pos_att_type", "pos_dropout", "pos_key_layer", "pos_proj", "pos_q_proj", "pos_query_layer", "position", "q", "q_bias", "qkvw", "qp", "query_layer", "query_states", "r_pos", "raise", "range", "rel_att", "rel_embeddings", "relative_attention", "relative_pos", "return", "scale", "scale_factor", "scaled_size_sqrt", "score", "self", "size", "softmax", "super", "t", "talking_head", "the", "to", "torch", "transpose", "transpose_for_scores", "tuple", "uneven_size_corrected", "unsqueeze", "v", "v_bias", "value_layer", "view", "weight", "ws", "x", "zeros"], "deberta/modeling_deberta.py:DebertaEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Linear", "ModelEmbeddings", "ModelLayerNorm", "Module", "None", "True", "__init__", "arange", "bias", "class", "config", "def", "device", "dim", "dropout", "dtype", "else", "embed_proj", "embedding_size", "embeddings", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_biased_input", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "squeeze", "super", "to", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros", "zeros_like"], "deberta/modeling_deberta.py:DebertaAttention": ["DisentangledSelfAttention", "False", "ModelAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "att_matrix", "attention_mask", "attention_output", "bool", "class", "config", "def", "else", "forward", "hidden_states", "if", "is", "nn", "output", "output_attentions", "query_states", "rel_embeddings", "relative_pos", "return", "self", "self_output", "super", "torch", "tuple"], "deberta/modeling_deberta.py:DebertaIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deberta/modeling_deberta.py:DebertaOutput": ["Dropout", "LayerNorm", "Linear", "ModelLayerNorm", "ModelOutput", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super"], "deberta/modeling_deberta.py:DebertaLayer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "att_matrix", "attention", "attention_mask", "attention_output", "bool", "class", "config", "def", "else", "forward", "hidden_states", "if", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "query_states", "rel_embeddings", "relative_pos", "return", "self", "super", "torch", "tuple"], "deberta/modeling_deberta.py:DebertaEncoder": ["BaseModelOutput", "Embedding", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "_", "__init__", "all_attentions", "all_hidden_states", "and", "att_m", "attention_mask", "attentions", "bool", "build_relative_position", "class", "config", "def", "dim", "elif", "else", "enumerate", "extended_attention_mask", "for", "forward", "get_attention_mask", "get_rel_embedding", "get_rel_pos", "getattr", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_module", "max_position_embeddings", "max_relative_positions", "next_kv", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "query_states", "range", "rel_embeddings", "relative_attention", "relative_pos", "return", "return_dict", "self", "squeeze", "super", "torch", "tuple", "unsqueeze", "v", "weight"], "deberta/modeling_deberta.py:DebertaPreTrainedModel": ["DisentangledSelfAttention", "Embedding", "LayerNorm", "LegacyModelLMPredictionHead", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelLayerNorm", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_keys_to_ignore_on_load_unexpected", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "position_embeddings", "q_bias", "self", "std", "supports_gradient_checkpointing", "v_bias", "weight", "zero_"], "deberta/modeling_deberta.py:DebertaModel": ["BaseModelOutput", "False", "Model", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "and", "append", "at", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoded_layers", "encoder", "encoder_outputs", "for", "forward", "function", "get_attention_mask", "get_input_embeddings", "get_rel_embedding", "get_rel_pos", "have", "heads_to_prune", "hidden_states", "if", "implemented", "in", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "layer", "layers", "long", "mask", "model", "new_embeddings", "not", "ones", "or", "output_attentions", "output_hidden_states", "position_ids", "post_init", "prune", "query_states", "raise", "range", "rel_embeddings", "rel_pos", "relative_pos", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "warn_if_padding_and_no_attention_mask", "word_embeddings", "z_steps", "zeros"], "deberta/modeling_deberta.py:LegacyDebertaPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "embedding_size", "eps", "forward", "getattr", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "deberta/modeling_deberta.py:LegacyDebertaLMPredictionHead": ["False", "Linear", "ModelModelLMPredictionHead", "ModelModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "embedding_size", "forward", "getattr", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "deberta/modeling_deberta.py:LegacyDebertaOnlyMLMHead": ["ModelModelLMPredictionHead", "ModelModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "deberta/modeling_deberta.py:DebertaLMPredictionHead": ["ACT2FN", "LayerNorm", "Linear", "ModelLMPredictionHead", "Module", "Parameter", "True", "__init__", "bias", "class", "config", "def", "dense", "elementwise_affine", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "matmul", "nn", "return", "self", "str", "super", "t", "torch", "transform_act_fn", "vocab_size", "weight", "word_embeddings", "zeros"], "deberta/modeling_deberta.py:DebertaOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "__init__", "class", "config", "def", "forward", "lm_head", "nn", "prediction_scores", "return", "self", "sequence_output", "super", "word_embeddings"], "deberta/modeling_deberta.py:DebertaForMaskedLM": ["CrossEntropyLoss", "LegacyModelOnlyMLMHead", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "dense", "else", "embeddings", "forward", "get_output_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "legacy", "lm_head", "lm_predictions", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight", "word_embeddings"], "deberta/modeling_deberta.py:ContextPooler": ["ACT2FN", "Dropout", "Linear", "ModelPooler", "Model_token", "Module", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_size", "hidden_states", "nn", "output_dim", "pooled_output", "pooler_dropout", "pooler_hidden_act", "pooler_hidden_size", "property", "return", "self", "super"], "deberta/modeling_deberta.py:DebertaForSequenceClassification": ["BCEWithLogitsLoss", "ContextPooler", "CrossEntropyLoss", "Dropout", "Linear", "LogSoftmax", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "cls_dropout", "config", "def", "dim", "drop_out", "dropout", "dtype", "elif", "else", "encoder_layer", "expand", "float", "forward", "gather", "get_input_embeddings", "getattr", "hidden_dropout_prob", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "label_index", "labeled_logits", "labels", "log_softmax", "logits", "long", "loss", "loss_fct", "loss_fn", "mean", "multi_label_classification", "new_embeddings", "nn", "nonzero", "not", "num_labels", "or", "output", "output_attentions", "output_dim", "output_hidden_states", "outputs", "pooled_output", "pooler", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "set_input_embeddings", "single_label_classification", "size", "squeeze", "sum", "super", "tensor", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deberta/modeling_deberta.py:DebertaForTokenClassification": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deberta/modeling_deberta.py:DebertaForQuestionAnswering": ["CrossEntropyLoss", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "cohere2_vision/modeling_cohere2_vision.py:Cohere2VisionMultiModalProjector": ["Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "SiLU", "True", "__init__", "act", "alignment_intermediate_size", "batch_size", "bias", "channels", "chunk", "class", "config", "def", "dim", "downsample_factor", "feature_dim", "forward", "gate", "height", "hidden_size", "hidden_states", "image_features", "int", "intermediate_size", "linear_1", "linear_2", "nn", "permute", "pixel_shuffle", "reshape", "return", "self", "seq_length", "shape", "super", "text_config", "vision_config", "width", "x"], "cohere2_vision/modeling_cohere2_vision.py:Cohere2VisionModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "cohere2_vision/modeling_cohere2_vision.py:Cohere2VisionCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "cohere2_vision/modeling_cohere2_vision.py:Cohere2VisionPreTrainedModel": ["Attention", "DecoderLayer", "False", "ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "past_key_values", "supports_gradient_checkpointing"], "cohere2_vision/modeling_cohere2_vision.py:Cohere2VisionModel": ["AutoModel", "Cache", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "check_model_inputs", "class", "config", "decoder", "def", "device", "do", "dtype", "else", "exactly", "expand_as", "f", "features", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_token_id", "input_ids", "inputs_embeds", "is", "kwargs", "language_model", "last_hidden_state", "long", "masked_scatter", "match", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "raise", "return", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "special_image_mask", "specify", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "value", "vision_config", "vision_tower"], "cohere2_vision/modeling_cohere2_vision.py:Cohere2VisionForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "check_model_inputs", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "super", "text_config", "torch", "tuple", "use_cache", "value", "vision_tower", "vocab_size", "weight"], "plbart/modeling_plbart.py:PLBartScaledWordEmbedding": ["Embedding", "Model", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "plbart/modeling_plbart.py:PLBartPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "False", "Model", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_ignore_causal_mask_sdpa", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "model", "not", "npu", "ones", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "supports_gradient_checkpointing", "target_length", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "xpu"], "plbart/modeling_plbart.py:PLBartLearnedPositionalEmbedding": ["Embedding", "Model", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "else", "embedding_dim", "expand", "forward", "if", "input_ids", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "seq_len", "shape", "super", "torch", "unsqueeze", "weight"], "plbart/modeling_plbart.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "plbart/modeling_plbart.py:PLBartAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "plbart/modeling_plbart.py:PLBartEncoderLayer": ["ACT2FN", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "None", "Optional", "__init__", "activation_dropout", "activation_fn", "activation_function", "and", "any", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "int", "isinf", "isnan", "layer_head_mask", "layer_idx", "max", "min", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training", "tuple"], "plbart/modeling_plbart.py:PLBartEncoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "__init__", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "input", "input_ids", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "weight"], "plbart/modeling_plbart.py:PLBartDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "plbart/modeling_plbart.py:PLBartDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with", "zip"], "plbart/modeling_plbart.py:shift_tokens_right": ["Model_tokens_right", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_tokens", "def", "defined", "dim", "gather", "has", "if", "index_of_eos", "input_ids", "int", "is", "masked_fill_", "model", "ne", "pad_token_id", "prev_output_tokens", "raise", "return", "self", "squeeze", "sum", "to", "torch", "unsqueeze"], "plbart/modeling_plbart.py:PLBartModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "init_weights", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "list", "math", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "r", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "shift_tokens_right", "sqrt", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "plbart/modeling_plbart.py:PLBartForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "head_mask", "if", "init_weights", "input_ids", "inputs_embeds", "int", "is", "labels", "list", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "prepare_decoder_input_ids_from_labels", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight", "zeros"], "plbart/modeling_plbart.py:PLBartClassificationHead": ["Dropout", "Linear", "Model", "Module", "Tensor", "__init__", "class", "def", "dense", "dropout", "float", "forward", "hidden_states", "inner_dim", "input_dim", "int", "nn", "num_classes", "out_proj", "p", "pooler_dropout", "return", "self", "super", "tanh", "torch"], "plbart/modeling_plbart.py:PLBartForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "classification_head", "classifier_dropout", "config", "cross_attentions", "cross_attn_head_mask", "currently", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_states", "if", "input", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "len", "list", "logits", "long", "loss", "loss_fct", "model", "multi_label_classification", "must", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "same", "self", "sentence_representation", "single_label_classification", "size", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "plbart/modeling_plbart.py:PLBartDecoderWrapper": ["Model", "ModelDecoder", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "plbart/modeling_plbart.py:PLBartForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelDecoderWrapper", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "layoutlm/modeling_layoutlm.py:LayoutLMEmbeddings": ["Dropout", "Embedding", "False", "IndexError", "LayerNorm", "ModelEmbeddings", "ModelLayerNorm", "Module", "None", "The", "__init__", "arange", "as", "bbox", "be", "class", "config", "coordinate", "def", "device", "dropout", "dtype", "e", "else", "embeddings", "eps", "except", "expand", "forward", "from", "h_position_embeddings", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "left_position_embeddings", "long", "lower_position_embeddings", "max_2d_position_embeddings", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embeddings", "position_ids", "raise", "range", "register_buffer", "return", "right_position_embeddings", "self", "seq_length", "should", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "try", "type_vocab_size", "upper_position_embeddings", "values", "vocab_size", "w_position_embeddings", "within", "word_embeddings", "words_embeddings", "x_position_embeddings", "y_position_embeddings", "zeros"], "layoutlm/modeling_layoutlm.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "layoutlm/modeling_layoutlm.py:LayoutLMSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "bool", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_states", "kwargs", "multiple", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "query", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_states", "view"], "layoutlm/modeling_layoutlm.py:LayoutLMSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "layoutlm/modeling_layoutlm.py:LayoutLMAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "layoutlm/modeling_layoutlm.py:LayoutLMIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "layoutlm/modeling_layoutlm.py:LayoutLMOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "layoutlm/modeling_layoutlm.py:LayoutLMLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "layoutlm/modeling_layoutlm.py:LayoutLMEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "layoutlm/modeling_layoutlm.py:LayoutLMPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "layoutlm/modeling_layoutlm.py:LayoutLMPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "layoutlm/modeling_layoutlm.py:LayoutLMLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "layoutlm/modeling_layoutlm.py:LayoutLMOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "layoutlm/modeling_layoutlm.py:LayoutLMPreTrainedModel": ["Embedding", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelLayerNorm", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "layoutlm/modeling_layoutlm.py:LayoutLMModel": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "both", "can_return_tuple", "cannot", "class", "config", "def", "device", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "finfo", "for", "forward", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "long", "min", "next", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "parameters", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "unsqueeze", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "layoutlm/modeling_layoutlm.py:LayoutLMForMaskedLM": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bbox", "bias", "bool", "can_return_tuple", "class", "cls", "config", "decoder", "def", "else", "embeddings", "forward", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight", "word_embeddings"], "layoutlm/modeling_layoutlm.py:LayoutLMForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "True", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "can_return_tuple", "class", "classifier", "config", "def", "dropout", "dtype", "elif", "else", "embeddings", "forward", "get_input_embeddings", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "word_embeddings"], "layoutlm/modeling_layoutlm.py:LayoutLMForTokenClassification": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TokenClassifierOutput", "True", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "can_return_tuple", "class", "classifier", "config", "def", "dropout", "else", "embeddings", "forward", "get_input_embeddings", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "word_embeddings"], "layoutlm/modeling_layoutlm.py:LayoutLMForQuestionAnswering": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "True", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "can_return_tuple", "clamp", "class", "config", "contiguous", "def", "dim", "else", "embeddings", "end_logits", "end_loss", "end_positions", "forward", "get_input_embeddings", "has_visual_segment_embedding", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "word_embeddings"], "clvp/modeling_clvp.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "clvp/modeling_clvp.py:clvp_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "return", "similarity", "speech_loss", "t", "torch"], "clvp/modeling_clvp.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "clvp/modeling_clvp.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim", "v", "v_embed"], "clvp/modeling_clvp.py:_pad_extra_bos_eos_tokens": ["None", "True", "_pad_extra_bos_eos_tokens", "add_bos_token", "add_eos_token", "attention_mask", "bos_token_id", "concatenate", "def", "device", "dtype", "each_input_id", "else", "enumerate", "eos_token_id", "for", "functional", "i", "if", "in", "input_ids", "is", "isin_mps_friendly", "min", "modified_input_ids", "nn", "not", "pad", "pad_token_id", "pos", "return", "shape", "sum", "tensor", "torch", "value", "where", "zeros"], "clvp/modeling_clvp.py:ClvpEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "embeds", "hidden_states", "last_hidden_state", "pooler_output", "r", "torch", "tuple"], "clvp/modeling_clvp.py:ClvpOutput": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelOutput", "None", "Optional", "class", "decoder_hidden_states", "logits_per_speech", "logits_per_text", "loss", "r", "speech_embeds", "speech_encoder_hidden_states", "speech_ids", "speech_model_output", "text_embeds", "text_encoder_hidden_states", "text_model_output", "torch"], "clvp/modeling_clvp.py:ClvpRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "clvp/modeling_clvp.py:ClvpRotaryPositionalEmbedding": ["FloatTensor", "ModelRotaryPositionalEmbedding", "Module", "None", "__init__", "and", "arange", "cached_rotary_positional_embedding", "cached_sequence_length", "cat", "class", "config", "def", "device", "dim", "dtype", "einsum", "embeddings", "float", "forward", "freqs", "hidden_states", "i", "if", "ij", "int64", "inv_freq", "is", "j", "max", "nn", "not", "num_attention_heads", "projection_dim", "register_buffer", "return", "self", "sequence_length", "shape", "super", "time_stamps", "torch", "type_as", "unsqueeze"], "clvp/modeling_clvp.py:ClvpSelfAttention": ["Attention", "Cache", "False", "FloatTensor", "Linear", "LongTensor", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "ValueError", "_", "__init__", "_shape", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_position", "cat", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "embed_dim", "f", "forward", "functional", "got", "hasattr", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_pass", "key_rot", "key_states", "layer_idx", "mask", "matmul", "max_position_embeddings", "max_positions", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "of", "ones", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "persistent", "position_ids", "provided", "q_proj", "query_pass", "query_rot", "query_states", "raise", "register_buffer", "reshape", "return", "rotary_emb_dim", "rotary_pos_emb", "scale", "self", "seq_len", "shape", "should", "sin", "size", "softmax", "squeeze", "src_len", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tril", "tuple", "update", "use_attention_bias", "use_cache", "v_proj", "value_pass", "value_rot", "value_states", "version", "view", "when"], "clvp/modeling_clvp.py:ClvpGatedLinearUnit": ["ACT2FN", "FloatTensor", "Linear", "ModelGatedLinearUnit", "Module", "__init__", "activation_fn", "chunk", "class", "config", "def", "dim", "forward", "gate", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "proj", "return", "self", "super", "torch"], "clvp/modeling_clvp.py:ClvpEncoderMLP": ["Dropout", "FloatTensor", "Linear", "ModelEncoderMLP", "ModelGatedLinearUnit", "Module", "__init__", "class", "config", "def", "dropout", "dropout_layer", "fc1", "fc2", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "clvp/modeling_clvp.py:ClvpEncoderLayer": ["False", "FloatTensor", "LongTensor", "ModelConfig", "ModelEncoderLayer", "ModelEncoderMLP", "ModelRMSNorm", "ModelSelfAttention", "Module", "Optional", "__init__", "attention_mask", "attn_weights", "bool", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "input_rmsnorm", "layer_norm_eps", "mlp", "nn", "output_attentions", "position_ids", "post_attention_rmsnorm", "residual", "return", "rotary_pos_emb", "self", "self_attn", "super", "torch", "tuple"], "clvp/modeling_clvp.py:ClvpSequenceSummary": ["Callable", "Dropout", "FloatTensor", "Identity", "Linear", "LongTensor", "ModelConfig", "ModelSequenceSummary", "Module", "None", "NotImplementedError", "Optional", "__init__", "activation", "activation_string", "and", "attn", "class", "cls_index", "config", "def", "dim", "dtype", "elif", "else", "expand", "first", "first_dropout", "forward", "full_like", "gather", "get_activation", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "is", "last", "last_dropout", "long", "mean", "nn", "num_classes", "num_labels", "output", "r", "raise", "return", "self", "shape", "size", "squeeze", "summary", "summary_activation", "summary_first_dropout", "summary_last_dropout", "summary_proj_to_labels", "summary_type", "summary_use_proj", "super", "torch", "unsqueeze"], "clvp/modeling_clvp.py:ClvpDecoderMLP": ["ACT2FN", "Conv1D", "Dropout", "FloatTensor", "ModelDecoderMLP", "Module", "Optional", "__init__", "act", "activation_function", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "hidden_size", "hidden_states", "intermediate_size", "nn", "resid_pdrop", "return", "self", "super", "torch", "tuple"], "clvp/modeling_clvp.py:ClvpDecoderLayer": ["Cache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelDecoderLayer", "ModelDecoderMLP", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attn", "attn_output", "attn_outputs", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "feed_forward_hidden_states", "forward", "head_mask", "hidden_size", "hidden_states", "if", "inner_dim", "input_layernorm", "is", "layer_idx", "layer_norm_epsilon", "mlp", "n_inner", "new_name", "nn", "not", "output_attentions", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "residual", "return", "self", "super", "torch", "tuple", "use_cache", "version"], "clvp/modeling_clvp.py:ClvpConditioningEncoder": ["Conv1d", "Embedding", "False", "FloatTensor", "Found", "GroupNorm", "LongTensor", "ModelConditioningEncoder", "ModelConfig", "ModelSelfAttention", "Module", "ModuleList", "None", "Number", "Optional", "Please", "The", "True", "ValueError", "You", "_", "__init__", "_pad_extra_bos_eos_tokens", "a", "affine", "and", "at", "attention_mask", "audios", "batch_size", "be", "bos_token_id", "both", "but", "cannot", "channels", "checkpoint", "class", "compute_groupnorm_groups", "concat", "config", "consider", "cumsum", "decoder_config", "def", "device", "different", "dim", "dtype", "either", "elif", "else", "enumerate", "eos_token_id", "eps", "f", "feature_size", "for", "forward", "gradient_checkpointing", "greater", "group_norms", "groups", "have", "hidden_size", "i", "if", "in", "input_features", "input_ids", "inputs_embeds", "int", "is", "it", "kernel_size", "long", "max_text_tokens", "mel_attn_block", "mel_attn_blocks", "mel_conv", "mel_spec", "must", "nn", "not", "num_groups", "num_mel_attn_blocks", "number", "of", "ones", "or", "position_embeds", "position_ids", "raise", "range", "repeat", "residual_mel_spec", "return", "same", "self", "seq_length", "shape", "size", "specify", "super", "text_config", "text_embeds", "text_position_embedding", "text_token_embedding", "texts", "than", "the", "time", "to", "torch", "training", "transpose", "unsqueeze", "using", "utils", "vocab_size", "vs", "while"], "clvp/modeling_clvp.py:ClvpPreTrainedModel": ["Conv1D", "Conv1d", "Embedding", "GroupNorm", "LayerNorm", "Linear", "Model", "ModelConditioningEncoder", "ModelConfig", "ModelEncoder", "ModelEncoderMLP", "ModelForCausalLM", "ModelModelForConditionalGeneration", "ModelPreTrainedModel", "ModelRMSNorm", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_skip_keys_device_placement", "base_model_prefix", "bias", "c_proj", "class", "config", "data", "def", "elif", "else", "factor", "fc1", "fc2", "fc_std", "fill_", "for", "get_text_config", "getattr", "hidden_size", "if", "in", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "logit_scale", "logit_scale_init_value", "math", "mean", "mel_conv", "module", "name", "named_parameters", "nn", "normal_", "not", "num_hidden_layers", "p", "past_key_values", "proj", "projection", "self", "sqrt", "std", "supports_gradient_checkpointing", "weight", "zero_"], "clvp/modeling_clvp.py:ClvpEncoder": ["BaseModelOutput", "Embedding", "False", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelEncoderOutput", "ModelPreTrainedModel", "ModelRotaryPositionalEmbedding", "ModelSequenceSummary", "ModuleList", "None", "Optional", "Union", "ValueError", "You", "_", "__call__", "__init__", "_prepare_4d_attention_mask", "all_attentions", "and", "arange", "at", "attention_mask", "attentions", "bias", "bool", "both", "cannot", "checkpoint", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embeds", "encoder_layer", "encoder_states", "enumerate", "eps", "final_layer_norm", "for", "forward", "get_input_embeddings", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "layer_norm_eps", "layer_outputs", "layers", "long", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "post_init", "projection", "projection_dim", "r", "raise", "range", "return", "return_dict", "rotary_pos_emb", "same", "self", "sequence_summary", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_embedding", "torch", "training", "tuple", "unsqueeze", "use_return_dict", "use_rotary_embedding", "utils", "v", "value", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "clvp/modeling_clvp.py:ClvpDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "__call__", "__init__", "_prepare_4d_causal_attention_mask", "_prune_heads", "a", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "an", "and", "arange", "at", "attention_mask", "attentions", "attn", "auto_docstring", "be", "block", "bool", "both", "cache_position", "cannot", "checkpoint", "checkpointing", "class", "config", "cross_attentions", "def", "deprecated", "device", "drop", "dtype", "e", "either", "elif", "else", "embd_pdrop", "enumerate", "eps", "for", "forward", "from_legacy_cache", "g", "get_head_mask", "get_input_embeddings", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "input_embeds_layer", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "items", "last_hidden_state", "layer", "layer_idx", "layer_norm", "layer_norm_epsilon", "layers", "logger", "long", "max_position_embeddings", "new_embeddings", "nn", "not", "num_hidden_layers", "of", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "pass", "past_key_values", "past_key_values_length", "position_embeds", "position_embeds_layer", "position_ids", "post_init", "prune_heads", "raise", "range", "removed", "return", "return_dict", "same", "self", "set_input_embeddings", "shape", "should", "size", "specify", "super", "the", "time", "to", "token_type_embeds", "token_type_ids", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "utils", "v", "v4", "view", "vocab_size", "warn_if_padding_and_no_attention_mask", "warning_once", "will", "with"], "clvp/modeling_clvp.py:ClvpModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "FloatTensor", "LongTensor", "ModelDecoder", "ModelDecoderConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "decoder", "decoder_outputs", "def", "else", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_embeds_layer", "input_ids", "inputs_embeds", "is", "last_hidden_state", "not", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "return", "return_dict", "self", "set_input_embeddings", "super", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value"], "clvp/modeling_clvp.py:ClvpForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "FloatTensor", "GenerationMixin", "LayerNorm", "Linear", "LongTensor", "Make", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "__init__", "_maybe_initialize_input_ids_for_generation", "_prepare_model_inputs", "allowed", "alongside", "and", "arange", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "bos_token_id", "cache_position", "class", "concat", "conditioning_embeds", "config", "contiguous", "cross_attentions", "cumsum", "decoder", "def", "device", "dict", "dim", "dtype", "either", "elif", "else", "f", "fill_value", "final_norm", "for", "forward", "full", "get", "get_input_embeddings", "get_output_embeddings", "hasattr", "head_mask", "hidden_size", "hidden_states", "if", "in", "input_embeds_layer", "input_ids", "input_ids_length", "input_name", "inputs", "inputs_embeds", "inputs_kwarg", "int", "is", "items", "k", "kwargs", "labels", "lm_head", "lm_logits", "logits", "long", "loss", "loss_fct", "main_input_name", "mel_start_token_embedding", "model", "model_inputs", "model_kwargs", "new_embeddings", "nn", "not", "ones", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pass", "passed", "past_key_values", "pop", "position_embeds_layer", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "raise", "repeat", "return", "return_dict", "self", "set_input_embeddings", "shape", "shift_labels", "shift_logits", "size", "str", "super", "sure", "tensor", "to", "token_type_ids", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "view", "vocab_size", "were", "which"], "clvp/modeling_clvp.py:ClvpModelForConditionalGeneration": ["Either", "False", "FloatTensor", "Found", "GenerationConfig", "GenerationMixin", "LongTensor", "Maximum", "ModelConditioningEncoder", "ModelConfig", "ModelDecoderConfig", "ModelEncoder", "ModelEncoderConfig", "ModelForCausalLM", "ModelModelForConditionalGeneration", "ModelOutput", "ModelPreTrainedModel", "Model_loss", "None", "Optional", "Parameter", "Please", "Tensor", "True", "TypeError", "Union", "ValueError", "__init__", "_pad_extra_bos_eos_tokens", "_validate_model_kwargs", "add_bos_token", "and", "argmax", "attention_mask", "auto_docstring", "be", "bool", "bos_token_id", "but", "cache_position", "class", "conditioning_embeds", "conditioning_encoder", "conditioning_encoder_inputs_embeds", "config", "continue", "copy", "decoder_config", "decoder_fixing_codes", "decoder_hidden_states", "decoder_outputs", "deepcopy", "def", "device", "dim", "dtype", "each_seq_stop_token_index", "else", "enumerate", "eos_token_id", "exp", "expected", "f", "fix_speech_decoder_output", "for", "forward", "functional", "generate", "generation_config", "get_speech_features", "get_text_features", "hidden_states", "i", "if", "in", "input_features", "input_ids", "inputs_embeds", "int", "is", "isinstance", "keepdim", "kwargs", "length", "logit_scale", "logit_scale_init_value", "logits_per_speech", "logits_per_text", "long", "loss", "make", "mask", "masked_fill", "matmul", "max_text_tokens", "maximum", "model_kwargs", "must", "ndim", "nn", "no_grad", "norm", "not", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "p", "pad", "pad_to_max_mel_tokens", "padding_needed", "post_init", "provided", "r", "raise", "reached", "return", "return_dict", "return_dict_in_generate", "return_loss", "self", "sequence", "sequence_length", "sequences", "shape", "speech_config", "speech_decoder_model", "speech_embeds", "speech_encoder_hidden_states", "speech_encoder_model", "speech_ids", "speech_model_output", "speech_outputs", "stm", "stop_token_indices", "sum", "super", "sure", "t", "tensor", "text_config", "text_embeds", "text_encoder_hidden_states", "text_encoder_inputs_embeds", "text_encoder_model", "text_model_output", "text_outputs", "that", "the", "to", "torch", "tuple", "type", "update", "use_return_dict", "validate", "value", "where"], "qwen3_moe/modeling_qwen3_moe.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen3_moe/modeling_qwen3_moe.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen3_moe/modeling_qwen3_moe.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen3_moe/modeling_qwen3_moe.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "not", "return", "self", "super", "up_proj", "x"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeSparseMoeBlock": ["F", "False", "Linear", "ModelMLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "batch_size", "bias", "class", "config", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_hit", "expert_idx", "expert_layer", "expert_mask", "experts", "final_hidden_states", "float", "for", "forward", "functional", "gate", "greater", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "keepdim", "moe_intermediate_size", "nn", "nonzero", "norm_topk_prob", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "squeeze", "sum", "super", "to", "top_k", "top_x", "topk", "torch", "view", "where", "zeros"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeDecoderLayer": ["Cache", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "and", "attention_mask", "cache_position", "class", "config", "decoder_sparse_step", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "in", "input_layernorm", "int", "intermediate_size", "isinstance", "kwargs", "layer_idx", "mlp", "mlp_only_layers", "new_name", "not", "num_experts", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "version"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoePreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelSparseMoeBlock", "OutputRecorder", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "index", "model", "past_key_values", "router_logits", "supports_gradient_checkpointing"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeModel": ["Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "mask_function", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_window", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "qwen3_moe/modeling_qwen3_moe.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_gate_logits", "def", "device", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "gate_logits", "if", "in", "int", "is", "isinstance", "layer_gate", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "reshape", "return", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeForCausalLM": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "device", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "num_experts", "num_experts_per_tok", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "router_aux_loss_coef", "router_logits", "self", "slice", "slice_indices", "super", "to", "torch", "use_cache", "vocab_size", "weight"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "qwen3_moe/modeling_qwen3_moe.py:Qwen3MoeForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTEmbeddings": ["Conv2d", "Model", "ModelConfig", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "batch_size", "bool_masked_pos", "cat", "class", "cls_token", "config", "def", "dim", "dtype", "embeddings", "empty", "expand", "flatten", "forward", "hidden_size", "if", "is", "kernel_size", "mask_token", "nn", "not", "num_channels", "num_register_tokens", "patch_embeddings", "patch_size", "pixel_values", "randn", "register_tokens", "return", "self", "shape", "stride", "super", "target_dtype", "to", "torch", "transpose", "unsqueeze", "weight", "where", "zeros"], "dinov3_vit/modeling_dinov3_vit.py:get_patches_center_coordinates": ["Model_patches_center_coordinates", "Tensor", "arange", "coords", "coords_h", "coords_w", "def", "device", "dim", "dtype", "flatten", "ij", "indexing", "int", "meshgrid", "num_patches_h", "num_patches_w", "return", "stack", "torch"], "dinov3_vit/modeling_dinov3_vit.py:augment_patches_center_coordinates": ["Model_patches_center_coordinates", "None", "Optional", "Tensor", "coords", "def", "device", "dtype", "empty", "exp", "float", "if", "is", "jitter", "jitter_hw", "jitter_range", "log", "not", "np", "rescale", "rescale_hw", "rescale_range", "return", "shift", "shift_hw", "torch", "uniform_"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTRopePositionEmbedding": ["False", "Model", "ModelConfig", "Module", "None", "Tensor", "_", "__init__", "and", "angles", "arange", "augment_patches_center_coordinates", "autocast", "base", "class", "config", "cos", "cpu", "def", "device", "device_type", "dtype", "else", "enabled", "flatten", "float32", "forward", "get_patches_center_coordinates", "head_dim", "height", "hidden_size", "if", "image_size", "inv_freq", "isinstance", "jitter", "math", "mps", "nn", "num_attention_heads", "num_patches_h", "num_patches_w", "patch_coords", "patch_size", "persistent", "pi", "pixel_values", "pos_embed_jitter", "pos_embed_rescale", "pos_embed_shift", "register_buffer", "rescale", "return", "rope_theta", "self", "shape", "shift", "sin", "str", "super", "tile", "to", "torch", "training", "tuple", "type", "width", "with"], "dinov3_vit/modeling_dinov3_vit.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "dinov3_vit/modeling_dinov3_vit.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "dinov3_vit/modeling_dinov3_vit.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "Tensor", "cat", "cos", "def", "dim", "k", "k_patches", "k_prefix_tokens", "kwargs", "num_patches", "num_prefix_tokens", "num_tokens", "q", "q_patches", "q_prefix_tokens", "return", "rotate_half", "shape", "sin", "split", "torch", "tuple"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "class", "config", "contiguous", "cos", "def", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "forward", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "key_bias", "key_states", "kwargs", "nn", "not", "num_attention_heads", "num_heads", "o_proj", "patches", "position_embeddings", "proj_bias", "q_proj", "query_bias", "query_states", "reshape", "return", "scaling", "self", "sin", "size", "super", "torch", "training", "transpose", "tuple", "v_proj", "value_bias", "value_states", "view"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTLayerScale": ["Model", "Module", "None", "Parameter", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_size", "hidden_state", "lambda1", "layerscale_value", "nn", "ones", "return", "self", "super", "torch"], "dinov3_vit/modeling_dinov3_vit.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTDropPath": ["Model", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTMLP": ["ACT2FN", "Linear", "Model", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTGatedMLP": ["ACT2FN", "Linear", "Model", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTLayer": ["GradientCheckpointingLayer", "Identity", "LayerNorm", "Model", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelGatedMLP", "ModelMLP", "ModelScale", "None", "Optional", "Tensor", "_", "__init__", "attention", "attention_mask", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm_eps", "layer_scale1", "layer_scale2", "mlp", "nn", "norm1", "norm2", "position_embeddings", "residual", "return", "self", "super", "torch", "tuple", "use_gated_mlp"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelLayerScale", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "dtype", "elif", "fill_", "float32", "hidden_states", "if", "init", "initializer_range", "is", "isinstance", "lambda1", "layerscale_value", "main_input_name", "mask_token", "mean", "module", "nn", "not", "num_register_tokens", "pixel_values", "register_tokens", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "dinov3_vit/modeling_dinov3_vit.py:DINOv3ViTModel": ["BaseModelOutputWithPooling", "False", "LayerNorm", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelRopePositionEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dtype", "else", "embeddings", "enumerate", "eps", "for", "forward", "get_input_embeddings", "gradient_checkpointing", "head_mask", "hidden_size", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_norm_eps", "nn", "norm", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "pooled_output", "pooler_output", "position_embeddings", "post_init", "r", "range", "return", "rope_embeddings", "self", "sequence_output", "super", "to", "torch", "weight"], "pvt/modeling_pvt.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "pvt/modeling_pvt.py:PvtDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "pvt/modeling_pvt.py:PvtPatchEmbeddings": ["Conv2d", "Dropout", "F", "False", "Iterable", "LayerNorm", "Make", "ModelConfig", "ModelPatchEmbeddings", "Module", "None", "Parameter", "Tensor", "Union", "ValueError", "_", "__init__", "abc", "and", "batch_size", "bilinear", "bool", "cat", "channel", "class", "cls_token", "collections", "config", "configuration", "def", "dim", "dimension", "dropout", "else", "embeddings", "eps", "expand", "flatten", "forward", "height", "hidden_dropout_prob", "hidden_size", "if", "image_size", "in", "int", "interpolate", "interpolate_pos_encoding", "interpolated_embeddings", "is", "is_tracing", "isinstance", "jit", "kernel_size", "layer_norm", "layer_norm_eps", "match", "mode", "nn", "not", "num_channels", "num_patches", "of", "one", "p", "patch_embed", "patch_size", "permute", "pixel", "pixel_values", "position_embeddings", "projection", "raise", "randn", "reshape", "return", "self", "set", "shape", "size", "stride", "super", "sure", "that", "the", "torch", "transpose", "tuple", "values", "width", "with", "zeros"], "pvt/modeling_pvt.py:PvtSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "int", "nn", "return", "self", "super", "torch"], "pvt/modeling_pvt.py:PvtEfficientSelfAttention": ["Conv2d", "Dropout", "False", "LayerNorm", "Linear", "ModelConfig", "ModelEfficientSelfAttention", "Module", "Tensor", "The", "ValueError", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "eps", "f", "float", "forward", "functional", "heads", "height", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "kernel_size", "key", "key_layer", "layer_norm", "layer_norm_eps", "math", "matmul", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "num_channels", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "self", "seq_len", "sequence_reduction", "sequences_reduction_ratio", "shape", "size", "softmax", "sqrt", "stride", "super", "the", "torch", "transpose", "transpose_for_scores", "tuple", "value", "value_layer", "view", "width"], "pvt/modeling_pvt.py:PvtAttention": ["False", "ModelAttention", "ModelConfig", "ModelEfficientSelfAttention", "ModelSelfOutput", "Module", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "float", "forward", "heads", "height", "hidden_size", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "sequences_reduction_ratio", "set", "super", "torch", "tuple", "union", "value", "width"], "pvt/modeling_pvt.py:PvtFFN": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelFFN", "Module", "None", "Optional", "Tensor", "__init__", "class", "config", "def", "dense1", "dense2", "dropout", "else", "forward", "hidden_act", "hidden_dropout_prob", "hidden_features", "hidden_states", "if", "in_features", "int", "intermediate_act_fn", "is", "isinstance", "nn", "not", "out_features", "return", "self", "str", "super", "torch"], "pvt/modeling_pvt.py:PvtLayer": ["False", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelFFN", "ModelLayer", "Module", "Tensor", "__init__", "attention", "attention_output", "bool", "class", "config", "def", "drop_path", "else", "eps", "float", "forward", "height", "hidden_features", "hidden_size", "hidden_states", "if", "in_features", "int", "layer_norm_1", "layer_norm_2", "layer_norm_eps", "layer_output", "mlp", "mlp_hidden_size", "mlp_output", "mlp_ratio", "nn", "num_attention_heads", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "sequences_reduction_ratio", "super", "torch", "width"], "pvt/modeling_pvt.py:PvtEncoder": ["BaseModelOutput", "False", "FloatTensor", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelLayer", "ModelPatchEmbeddings", "Module", "ModuleList", "None", "Optional", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "append", "attentions", "batch_size", "block", "block_layer", "blocks", "bool", "class", "cls_token", "config", "contiguous", "cpu", "cur", "def", "depths", "device", "drop_path", "drop_path_decays", "drop_path_rate", "else", "embedding_layer", "embeddings", "enumerate", "eps", "for", "forward", "height", "hidden_size", "hidden_sizes", "hidden_states", "i", "idx", "if", "image_size", "in", "is", "j", "last_hidden_state", "layer_norm", "layer_norm_eps", "layer_outputs", "layers", "len", "linspace", "mlp_ratio", "mlp_ratios", "nn", "not", "num_attention_heads", "num_blocks", "num_channels", "num_encoder_blocks", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_size", "patch_sizes", "permute", "pixel_values", "range", "reshape", "return", "return_dict", "self", "sequence_reduction_ratios", "sequences_reduction_ratio", "shape", "stride", "strides", "sum", "super", "tolist", "torch", "tuple", "v", "width", "zip"], "pvt/modeling_pvt.py:PvtPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPatchEmbeddings", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "trunc_normal_", "weight", "zero_"], "pvt/modeling_pvt.py:PvtModel": ["BaseModelOutput", "FloatTensor", "ModelConfig", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_prune_heads", "attention", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "encoder", "encoder_outputs", "for", "forward", "heads", "heads_to_prune", "hidden_states", "if", "in", "is", "items", "last_hidden_state", "layer", "not", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "prune_heads", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "pvt/modeling_pvt.py:PvtForImageClassification": ["Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "tapas/modeling_tapas.py:TableQuestionAnsweringOutput": ["FloatTensor", "ModelOutput", "ModelQuestionAnsweringOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "logits_aggregation", "loss", "r", "torch", "tuple"], "tapas/modeling_tapas.py:TapasEmbeddings": ["Dropout", "Embedding", "IndexMap", "LayerNorm", "ModelEmbeddings", "Module", "None", "ProductIndexMap", "__init__", "arange", "as_tensor", "batch_dims", "class", "col_index", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "enumerate", "eps", "expand", "f", "first_position", "first_position_per_segment", "for", "forward", "full_index", "gather", "getattr", "hidden_dropout_prob", "hidden_size", "i", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "len", "long", "max_position_embeddings", "min", "name", "nn", "not", "number_of_token_type_embeddings", "pad_token_id", "padding_idx", "position", "position_embeddings", "position_ids", "range", "reduce_min", "reset_position_index_per_cell", "return", "row_index", "self", "seq_length", "setattr", "size", "super", "token_type_embeddings_", "token_type_ids", "torch", "type_vocab_sizes", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "tapas/modeling_tapas.py:TapasSelfAttention": ["Dropout", "EncoderDecoderCache", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "True", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "cache_position", "class", "config", "context_layer", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "dim", "dropout", "else", "embedding_size", "encoder_hidden_states", "f", "forward", "functional", "get", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "key", "key_layer", "keys", "layer_idx", "layers", "math", "matmul", "multiple", "new_context_layer_shape", "new_name", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "past_key_value", "past_key_values", "permute", "query", "query_layer", "raise", "return", "self", "self_attention_cache", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "update", "value", "value_layer", "values", "version", "view"], "tapas/modeling_tapas.py:TapasSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "tapas/modeling_tapas.py:TapasAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "cache_position", "class", "config", "def", "dense", "dim", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "layer_idx", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "tapas/modeling_tapas.py:TapasIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "tapas/modeling_tapas.py:TapasOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "tapas/modeling_tapas.py:TapasLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "bool", "by", "cache_position", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "crossattention", "decoder", "def", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_idx", "layer_output", "layers", "model", "not", "output", "output_attentions", "outputs", "passed", "past_key_values", "raise", "return", "self", "self_attention_outputs", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "with"], "tapas/modeling_tapas.py:TapasEncoder": ["BaseModelOutput", "DynamicCache", "EncoderDecoderCache", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Passing", "Transformers", "True", "You", "__init__", "a", "all_attentions", "all_hidden_states", "an", "and", "attention_mask", "attentions", "be", "cache_position", "class", "config", "def", "deprecated", "e", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "from_legacy_cache", "g", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "layer_outputs", "logger", "nn", "not", "num_hidden_layers", "of", "output_attentions", "output_hidden_states", "pass", "past_key_values", "range", "removed", "return", "return_dict", "self", "should", "super", "tuple", "use_cache", "v", "v4", "warning_once", "will"], "tapas/modeling_tapas.py:TapasPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "tapas/modeling_tapas.py:TapasPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "tapas/modeling_tapas.py:TapasLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "tapas/modeling_tapas.py:TapasOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "tapas/modeling_tapas.py:TapasPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_supports_param_buffer_assignment", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "tapas/modeling_tapas.py:TapasModel": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "items", "last_hidden_state", "layer", "len", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "type_vocab_sizes", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "tapas/modeling_tapas.py:TapasForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelConfig", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "base_model_prefix", "bias", "bool", "class", "cls", "config", "decoder", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "predictions", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view", "vocab_size", "weight"], "tapas/modeling_tapas.py:TapasForQuestionAnswering": ["Bernoulli", "Dropout", "EPSILON_ZERO_DIVISION", "False", "FloatTensor", "IndexMap", "Linear", "LongTensor", "Make", "Model", "ModelConfig", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "ProductIndexMap", "TableQuestionAnsweringOutput", "True", "Union", "ValueError", "You", "_", "__init__", "_calculate_aggregate_mask", "_calculate_aggregation_loss", "_calculate_regression_loss", "_single_column_cell_selection_loss", "a", "aggregate", "aggregate_mask", "aggregation", "aggregation_classifier", "aggregation_labels", "aggregation_loss_weight", "allow_empty_column_selection", "and", "answer_loss", "answers", "are", "as_tensor", "assert", "attention_mask", "attentions", "auto_docstring", "average_logits_per_cell", "batch_dims", "batch_size", "bool", "calculate", "calculate_loss", "cell_index", "cell_mask", "cell_selection_preference", "class", "col_index", "column_ids", "column_logits", "column_output_bias", "column_output_weights", "column_ranks", "compute_column_logits", "compute_token_logits", "config", "def", "device", "dim", "disable_per_token_loss", "dist_per_token", "distributions", "dropout", "dtype", "elif", "else", "empty", "float", "float32", "float_answer", "forward", "gather", "have", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "index", "indices", "init", "init_cell_selection_weights_to_zero", "initializer_range", "input_ids", "input_mask_float", "input_shape", "inputs_embeds", "inv_column_ranks", "is", "is_supervised", "labels", "large_answer_loss_mask", "len", "log_prob", "logits", "logits_aggregation", "logits_per_cell", "long", "loss", "mask", "max_num_columns", "max_num_rows", "mean", "min", "nn", "normal_", "not", "num_aggregation_labels", "num_segments", "numeric", "numeric_relations", "numeric_values", "numeric_values_scale", "of", "ones", "ones_like", "or", "order", "output", "output_attentions", "output_bias", "output_hidden_states", "output_weights", "outputs", "pass", "per_example_additional_loss", "pooled_output", "position_ids", "positive_label_weight", "post_init", "prev_labels", "r", "raise", "reduce_mean", "regression", "return", "return_dict", "row_ids", "row_index", "scale", "segment_ids", "select_one_column", "selection_loss_per_example", "selection_loss_per_token", "self", "sequence_output", "shape", "size", "specify", "std", "sum", "super", "sure", "table_mask", "table_mask_float", "temperature", "the", "to", "token_type_ids", "token_types", "torch", "total_loss", "tuple", "type_vocab_sizes", "use_answer_as_supervision", "use_return_dict", "values", "weight", "where", "zeros", "zeros_like"], "tapas/modeling_tapas.py:TapasForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "dtype", "elif", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "tapas/modeling_tapas.py:AverageApproximationFunction": ["Enum", "FIRST_ORDER", "ModelApproximationFunction", "RATIO", "SECOND_ORDER", "class", "enum", "first_order", "ratio", "second_order", "str"], "tapas/modeling_tapas.py:IndexMap": ["ModelMap", "__init__", "as_tensor", "batch_dims", "batch_shape", "class", "def", "device", "indices", "num_segments", "return", "self", "size", "torch"], "tapas/modeling_tapas.py:ProductIndexMap": ["IndexMap", "ModelIndexMap", "ValueError", "__init__", "and", "batch_dims", "be", "class", "def", "div", "float", "floor", "fmod", "if", "index", "indices", "inner_index", "long", "must", "num_segments", "outer_index", "project_inner", "project_outer", "raise", "return", "rounding_mode", "same", "self", "super", "the", "torch", "type"], "tapas/modeling_tapas.py:gather": ["Model", "batch_dims", "def", "else", "expand", "if", "index", "indices", "len", "name", "return", "segmented_Model", "shape", "size", "torch", "unsqueeze", "values", "view"], "tapas/modeling_tapas.py:flatten": ["IndexMap", "Model", "_", "arange", "batch_dims", "batch_shape", "batch_size", "def", "device", "end", "for", "in", "index", "indices", "len", "list", "name", "num_segments", "offset", "prod", "range", "return", "segmented_Model", "size", "start", "tensor", "torch", "unsqueeze", "view"], "tapas/modeling_tapas.py:range_index_map": ["IndexMap", "Model_index_map", "aModel", "as_tensor", "assert", "batch_dims", "batch_shape", "cat", "cpu", "def", "device", "dim", "dtype", "else", "end", "for", "if", "in", "indices", "int", "is_tensor", "len", "list", "long", "multiples", "name", "new_shape", "new_tensor", "num_segments", "ones_like", "repeat", "return", "size", "start", "tolist", "torch", "unsqueeze", "view", "x"], "tapas/modeling_tapas.py:_segment_reduce": ["False", "_segment_reduce", "as_tensor", "batch_shape", "cat", "clone", "def", "device", "dim", "dtype", "flat_index", "flat_values", "flatten", "flattened_shape", "float", "include_self", "index", "indices", "int", "len", "long", "name", "new_shape", "num_segments", "out", "output_index", "output_values", "range_index_map", "reduce", "reshape", "return", "scatter_reduce", "segment_means", "segment_reduce_fn", "size", "src", "to", "tolist", "torch", "values", "vector_shape", "view", "zeros"], "tapas/modeling_tapas.py:reduce_sum": ["Model_sum", "_segment_Model", "def", "index", "name", "return", "segmented_Model_sum", "sum", "values"], "tapas/modeling_tapas.py:reduce_mean": ["Model_mean", "_segment_Model", "def", "index", "mean", "name", "return", "segmented_Model_mean", "values"], "tapas/modeling_tapas.py:reduce_max": ["Model_max", "_segment_Model", "amax", "def", "index", "name", "return", "segmented_Model_max", "values"], "tapas/modeling_tapas.py:reduce_min": ["Model_min", "_segment_Model", "amin", "def", "index", "name", "return", "segmented_Model_min", "values"], "tapas/modeling_tapas.py:compute_column_logits": ["CLOSE_ENOUGH_TO_LOG_ZERO", "EPSILON_ZERO_DIVISION", "Model_column_logits", "_", "allow_empty_column_selection", "as_tensor", "bs", "bsj", "cell_count", "cell_index", "cell_logits", "cell_logits_index", "cell_mask", "column_index", "column_logits", "column_output_bias", "column_output_weights", "def", "device", "dtype", "einsum", "eq", "float32", "if", "indices", "is_padding", "j", "logical_and", "not", "out_index", "project_inner", "reduce_mean", "reduce_sum", "return", "sequence_output", "token_logits", "torch"], "tapas/modeling_tapas.py:_single_column_cell_selection_loss": ["Bernoulli", "CLOSE_ENOUGH_TO_LOG_ZERO", "Categorical", "EPSILON_ZERO_DIVISION", "_", "_single_column_cell_selection_loss", "argmax", "as_tensor", "cell_dist", "cell_index", "cell_log_prob", "cell_loss", "cell_mask", "col_index", "column_dist", "column_id_for_cells", "column_label", "column_logits", "column_loss_per_example", "column_mask", "def", "device", "dim", "distributions", "dtype", "eq", "float32", "gather", "indices", "labels", "labels_index", "labels_per_cell", "labels_per_column", "log_prob", "logits", "logits_per_cell", "long", "max", "new_logits_per_cell", "no_cell_selected", "project_inner", "reduce_max", "reduce_mean", "reduce_sum", "return", "selected_column_id", "selected_column_mask", "selection_loss_per_example", "size", "sum", "token_logits", "torch", "type", "unsqueeze", "view", "where", "zeros_like"], "tapas/modeling_tapas.py:compute_token_logits": ["Model_token_logits", "bs", "bsj", "def", "einsum", "j", "logits", "output_bias", "output_weights", "return", "sequence_output", "temperature", "torch"], "tapas/modeling_tapas.py:_calculate_aggregate_mask": ["Categorical", "FloatTensor", "_calculate_aggregate_mask", "aggregate_mask", "aggregate_mask_init", "aggregation_classifier", "aggregation_ops_total_mass", "answer", "categorical", "cell_selection_preference", "def", "detach", "device", "dim", "dist_aggregation", "distributions", "dtype", "float32", "is_cell_supervision_available", "is_pred_cell_selection", "isnan", "labels", "logical_and", "logical_not", "logits", "logits_aggregation", "pooled_output", "probs", "return", "size", "sum", "to", "torch", "type", "view", "where", "zeros_like"], "tapas/modeling_tapas.py:_calculate_aggregation_loss_known": ["_calculate_aggregation_loss_known", "aggregate_mask", "aggregation_labels", "def", "dim", "dtype", "else", "float32", "functional", "if", "log_probs", "log_softmax", "logits_aggregation", "long", "nn", "num_aggregation_labels", "num_classes", "one_hot", "one_hot_labels", "per_example_aggregation_intermediate", "return", "sum", "target_aggregation", "torch", "type", "use_answer_as_supervision", "zeros_like"], "tapas/modeling_tapas.py:_calculate_aggregation_loss_unknown": ["Categorical", "_calculate_aggregation_loss_unknown", "aggregate_mask", "aggregation_ops_total_mass", "categorical", "def", "dim", "dist_aggregation", "distributions", "log", "logits", "logits_aggregation", "probs", "return", "sum", "torch"], "tapas/modeling_tapas.py:_calculate_aggregation_loss": ["_calculate_aggregation_loss", "_calculate_aggregation_loss_known", "_calculate_aggregation_loss_unknown", "aggregate_mask", "aggregation_labels", "aggregation_loss_weight", "def", "if", "logits_aggregation", "num_aggregation_labels", "per_example_aggregation_loss", "return", "use_answer_as_supervision"], "tapas/modeling_tapas.py:_calculate_expected_result": ["AverageApproximationFunction", "EPSILON_ZERO_DIVISION", "FIRST_ORDER", "Invalid", "RATIO", "RelaxedBernoulli", "RelaxedOneHotCategorical", "SECOND_ORDER", "True", "ValueError", "_calculate_expected_result", "aggregation_op_only_probs", "aggregation_temperature", "all_results", "average_approximation_function", "average_result", "avg_approximation", "cat", "config", "count_result", "def", "dim", "dist_per_cell", "distributions", "elif", "else", "ex", "expected_result", "f", "functional", "gumbel_dist", "if", "input_mask_float", "isnan", "keepdim", "logits", "logits_aggregation", "multiplier", "nn", "numeric_values", "numeric_values_masked", "numeric_values_scale", "pointwise_var", "probs", "raise", "return", "sample", "scaled_probability_per_cell", "softmax", "square", "sum", "sum_result", "temperature", "torch", "unsqueeze", "use_gumbel_for_aggregation", "use_gumbel_for_cells", "var", "where", "zeros_like"], "tapas/modeling_tapas.py:huber_loss": ["Model_loss", "abs", "def", "delta", "errors", "float", "input", "return", "target", "torch", "where"], "tapas/modeling_tapas.py:_calculate_regression_loss": ["EPSILON_ZERO_DIVISION", "None", "_calculate_expected_result", "_calculate_regression_loss", "abs", "aggregate_mask", "answer", "answer_loss_cutoff", "answer_loss_importance", "answer_masked", "config", "def", "delta", "detach", "dist_per_cell", "dtype", "else", "expected_result", "float32", "huber_loss", "huber_loss_delta", "if", "input_mask_float", "is", "isnan", "large_answer_loss_mask", "logits_aggregation", "max", "normalized_answer_masked", "normalized_expected_result", "normalizer", "numeric_values", "numeric_values_scale", "ones_like", "per_example_answer_loss", "per_example_answer_loss_scaled", "return", "torch", "use_normalized_answer_loss", "where", "zeros_like"], "visual_bert/modeling_visual_bert.py:VisualBertEmbeddings": ["Dropout", "Embedding", "False", "Found", "LayerNorm", "Linear", "Model", "ModelEmbeddings", "Model_embedding_dim", "Model_embeddings", "Model_embeds", "Model_position_embeddings", "Model_position_ids", "Model_projection", "Model_token_type_embeddings", "Model_token_type_ids", "Module", "None", "Parameter", "Setting", "True", "ValueError", "__init__", "arange", "as", "avoid", "be", "by", "cat", "class", "clone", "config", "data", "def", "device", "dim", "divide", "dropout", "dtype", "else", "embeddings", "eps", "error", "expand", "f", "forward", "hidden_dropout_prob", "hidden_size", "if", "image_text_alignment", "image_text_alignment_mask", "in", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "length", "logger", "long", "max_position_embeddings", "nn", "not", "ones", "pad_token_id", "padding_idx", "persistent", "position", "position_embeddings", "position_ids", "raise", "register_buffer", "requires_grad", "return", "same", "self", "seq_length", "should", "size", "special_Model_initialize", "sum", "super", "the", "them", "to", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "unsqueeze", "values", "vocab_size", "warning", "weight", "word_embeddings", "zero", "zeros"], "visual_bert/modeling_visual_bert.py:VisualBertSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "functional", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "value", "value_layer", "view"], "visual_bert/modeling_visual_bert.py:VisualBertSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "visual_bert/modeling_visual_bert.py:VisualBertAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "visual_bert/modeling_visual_bert.py:VisualBertIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "visual_bert/modeling_visual_bert.py:VisualBertOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "visual_bert/modeling_visual_bert.py:VisualBertLayer": ["False", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "visual_bert/modeling_visual_bert.py:VisualBertEncoder": ["BaseModelOutput", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "visual_bert/modeling_visual_bert.py:VisualBertPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "visual_bert/modeling_visual_bert.py:VisualBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "visual_bert/modeling_visual_bert.py:VisualBertLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "visual_bert/modeling_visual_bert.py:VisualBertPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "visual_bert/modeling_visual_bert.py:VisualBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMPredictionHead", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "and", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "visual_bert/modeling_visual_bert.py:VisualBertForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "seq_relationship_logits", "torch", "tuple"], "visual_bert/modeling_visual_bert.py:VisualBertModel": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "ModelEmbeddings", "ModelEncoder", "ModelLayer", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "Model_attention_mask", "Model_embedding_output", "Model_embeds", "Model_input_shape", "Model_token_type_ids", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "additional_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "bypass_transformer", "cannot", "cat", "class", "combined_attention_mask", "concatenated_input", "config", "def", "device", "dim", "either", "elif", "else", "embedding_output", "embeddings", "encoded_outputs", "encoder", "encoder_outputs", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "image_text_alignment", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "text_embedding_output", "text_extended_attention_mask", "text_length", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings"], "visual_bert/modeling_visual_bert.py:VisualBertForPreTraining": ["CrossEntropyLoss", "FloatTensor", "Found", "LongTensor", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "Model_attention_mask", "Model_embeds", "Model_token_type_ids", "None", "Optional", "Tensor", "The", "Union", "ValueError", "__init__", "_tied_weights_keys", "and", "as", "attention", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "cls", "config", "decoder", "def", "elif", "else", "expected", "f", "forward", "get_output_embeddings", "have", "head_mask", "hidden_states", "if", "image_text_alignment", "input_ids", "inputs_embeds", "is", "labels", "length", "loss", "loss_fct", "mask", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "prediction_logits", "prediction_scores", "predictions", "provided", "r", "raise", "return", "return_dict", "same", "self", "sentence_image_labels", "sentence_image_loss", "seq_relationship_logits", "seq_relationship_score", "sequence", "sequence_output", "set_output_embeddings", "should", "size", "super", "token_type_ids", "torch", "total", "total_loss", "total_size", "tuple", "use_return_dict", "view", "vocab_size", "weight", "with"], "visual_bert/modeling_visual_bert.py:VisualBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "Model_attention_mask", "Model_embeds", "Model_token_type_ids", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "_", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "cls", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image_text_alignment", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "visual_bert/modeling_visual_bert.py:VisualBertForQuestionAnswering": ["Dropout", "FloatTensor", "KLDivLoss", "Linear", "LogSoftmax", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "Model_attention_mask", "Model_embeds", "Model_token_type_ids", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "batchmean", "bool", "class", "cls", "config", "contiguous", "def", "dim", "dropout", "else", "expand", "forward", "gather", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image_text_alignment", "index_to_gather", "input_ids", "inputs_embeds", "is", "labels", "log_softmax", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reduction", "reshaped_logits", "return", "return_dict", "self", "sequence_output", "size", "sum", "super", "token_type_ids", "torch", "tuple", "unsqueeze", "use_return_dict", "view"], "visual_bert/modeling_visual_bert.py:VisualBertForVisualReasoning": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForModelReasoning", "ModelModel", "ModelPreTrainedModel", "Model_attention_mask", "Model_embeds", "Model_token_type_ids", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "cls", "config", "contiguous", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "image_text_alignment", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "visual_bert/modeling_visual_bert.py:VisualBertRegionToPhraseAttention": ["Dropout", "Linear", "ModelRegionToPhraseAttention", "Module", "The", "ValueError", "_", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_probs_dropout_prob", "attention_scores", "batch_size", "class", "config", "def", "dropout", "dtype", "f", "finfo", "forward", "heads", "hidden", "hidden_size", "if", "int", "is", "key", "key_layer", "math", "matmul", "min", "multiple", "nn", "not", "num_attention_heads", "number", "of", "query", "query_layer", "raise", "return", "self", "seq_length", "shape", "size", "sqrt", "squeeze", "super", "the", "to", "torch", "transpose", "unsqueeze", "value", "view"], "visual_bert/modeling_visual_bert.py:VisualBertForRegionToPhraseAlignment": ["Dropout", "Flickr", "FloatTensor", "KLDivLoss", "LogSoftmax", "LongTensor", "Model", "ModelForRegionToPhraseAlignment", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "ModelRegionToPhraseAttention", "Model_attention_mask", "Model_embeds", "Model_features", "Model_token_type_ids", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "ValueError", "__init__", "_tied_weights_keys", "as", "attention", "attention_mask", "attentions", "auto_docstring", "batchmean", "be", "bias", "bool", "class", "cls", "config", "contiguous", "decoder", "def", "dim", "dropout", "else", "expand", "expanded_region_to_phrase_positions", "f", "features", "forward", "gather", "head_mask", "hidden_dropout_prob", "hidden_states", "if", "image_text_alignment", "input_ids", "inputs_embeds", "is", "labels", "length", "log_softmax", "logits", "long", "loss", "loss_fct", "mask", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "predictions", "r", "raise", "reduction", "region_to_phrase_position", "region_to_phrase_position_mask", "return", "return_dict", "same", "scores", "selected_positions", "self", "sequence_output", "should", "size", "super", "the", "token_type_ids", "torch", "tuple", "unsqueeze", "use_return_dict", "using", "when"], "internvl/modeling_internvl.py:InternVLVisionRMSNorm": ["ModelVisionRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "internvl/modeling_internvl.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "float", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "torch", "training", "transpose", "value", "value_states"], "internvl/modeling_internvl.py:InternVLVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "Identity", "Linear", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "_", "__init__", "_attn_implementation", "and", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "by", "class", "config", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "output", "proj_dropout", "projection_dropout", "projection_layer", "q_norm", "q_proj", "qk_norm", "query_states", "raise", "reshape", "return", "scale", "scaling", "self", "seq_len", "size", "super", "torch", "training", "transpose", "use_qk_norm", "v_proj", "value_states", "view"], "internvl/modeling_internvl.py:InternVLVisionModelOutputWithPooling": ["BaseModelOutputWithPooling", "ModelVisionModelOutputWithPooling", "class", "r"], "internvl/modeling_internvl.py:InternVLVisionPatchEmbeddings": ["Conv2d", "Make", "ModelVisionPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "batch_size", "channel", "class", "config", "configuration", "def", "dimension", "embeddings", "flatten", "forward", "height", "hidden_size", "if", "image_size", "in", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_height", "patch_shape", "patch_size", "patch_width", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "width", "with"], "internvl/modeling_internvl.py:InternVLVisionEmbeddings": ["BoolTensor", "Dropout", "False", "Iterable", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "abc", "align_corners", "and", "batch_size", "bicubic", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "collections", "config", "def", "dim", "dropout", "else", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "image_size", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "isinstance", "jit", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "patch_embeddings", "patch_height", "patch_pos_embed", "patch_size", "patch_width", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_absolute_position_embeddings", "use_mask_token", "view", "w", "width", "zeros"], "internvl/modeling_internvl.py:InternVLVisionMLP": ["ACT2FN", "Linear", "ModelVisionMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "internvl/modeling_internvl.py:InternVLVisionLayer": ["Dropout", "GradientCheckpointingLayer", "ModelVisionAttention", "ModelVisionConfig", "ModelVisionLayer", "ModelVisionMLP", "NORM2FN", "None", "Parameter", "Tensor", "True", "Union", "_", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "init_values", "is", "lambda_1", "lambda_2", "layer_norm_eps", "layer_output", "layer_scale_init_value", "layernorm_after", "layernorm_before", "mlp", "nn", "norm_type", "not", "ones", "requires_grad", "return", "self", "seq_len_dim", "super", "torch", "tuple"], "internvl/modeling_internvl.py:InternVLVisionEncoder": ["BaseModelOutput", "False", "ModelVisionConfig", "ModelVisionEncoder", "ModelVisionLayer", "Module", "ModuleList", "None", "Tensor", "Union", "__init__", "check_model_inputs", "class", "config", "def", "for", "forward", "gradient_checkpointing", "hidden_states", "i", "in", "last_hidden_state", "layer", "layer_module", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "internvl/modeling_internvl.py:InternVLVisionPreTrainedModel": ["ModelVisionAttention", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionLayer", "ModelVisionPreTrainedModel", "Model_vision", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "cls_token", "config", "data", "def", "elif", "fill_", "hidden_states", "if", "is", "isinstance", "lambda_1", "lambda_2", "layer_scale_init_value", "main_input_name", "mask_token", "module", "not", "pixel_values", "position_embeddings", "self", "super", "supports_gradient_checkpointing", "zero_"], "internvl/modeling_internvl.py:InternVLVisionModel": ["BoolTensor", "Identity", "LayerNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionModel", "ModelVisionModelOutputWithPooling", "ModelVisionPreTrainedModel", "None", "Optional", "Tensor", "Union", "_", "__init__", "attentions", "auto_docstring", "bool_masked_pos", "can_return_tuple", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "last_hidden_state", "layer_norm_eps", "layernorm", "nn", "patch_embeddings", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch", "tuple", "use_mean_pooling"], "internvl/modeling_internvl.py:InternVLPreTrainedModel": ["ModelConfig", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "past_key_values", "supports_gradient_checkpointing"], "internvl/modeling_internvl.py:InternVLMultiModalProjector": ["ACT2FN", "LayerNorm", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "class", "config", "def", "downsample_ratio", "forward", "hidden_size", "hidden_states", "image_features", "int", "layer_norm", "linear_1", "linear_2", "nn", "projector_hidden_act", "return", "self", "super", "text_config", "vision_config"], "internvl/modeling_internvl.py:InternVLModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "internvl/modeling_internvl.py:InternVLModel": ["AutoModel", "Cache", "FloatTensor", "Height", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "by", "cache_position", "can_return_tuple", "channels", "class", "config", "contiguous", "decoder", "def", "default", "device", "divisible", "do", "downsample_ratio", "downsampling", "dtype", "else", "exactly", "expand_as", "f", "feature_size", "features", "float", "for", "forward", "from_config", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "height", "hidden_states", "if", "image", "image_features", "image_hidden_states", "image_token_id", "input_ids", "inputs_embeds", "int", "is", "kwargs", "language_model", "last_hidden_state", "list", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_image_features", "n_image_tokens", "not", "numel", "of", "one", "or", "outputs", "past_key_values", "permute", "pixel_shuffle", "pixel_values", "position_ids", "post_init", "proper", "raise", "reshape", "return", "scale_factor", "self", "set_decoder", "set_input_embeddings", "shape", "size", "special_image_mask", "specify", "str", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "value", "view", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_features", "vision_model", "vision_tower", "width"], "internvl/modeling_internvl.py:InternVLCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "internvl/modeling_internvl.py:InternVLForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "cache_position", "can_return_tuple", "class", "config", "decoder", "def", "else", "forward", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "model_inputs", "multi_modal_projector", "nn", "not", "outputs", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "return", "self", "set_decoder", "set_input_embeddings", "slice", "slice_indices", "str", "super", "text_config", "torch", "tuple", "value", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "codegen/modeling_codegen.py:create_sinusoidal_positions": ["Model_sinusoidal_positions", "Tensor", "arange", "cat", "cos", "def", "dim", "dtype", "einsum", "float", "i", "int", "int64", "inv_freq", "j", "num_pos", "return", "sin", "sinusoid_inp", "torch"], "codegen/modeling_codegen.py:rotate_every_two": ["Model_every_two", "Tensor", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "codegen/modeling_codegen.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "Tensor", "cos", "def", "repeat_interleave", "return", "rotate_every_two", "sin", "tensor", "torch"], "codegen/modeling_codegen.py:CodeGenAttention": ["Cache", "Dropout", "False", "FloatTensor", "Input", "Instantiating", "Linear", "LongTensor", "ModelAttention", "Module", "None", "Optional", "Please", "Softmax", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_attn", "_merge_heads", "_split_heads", "a", "and", "apply_rotary_pos_emb", "attention_mask", "attn_dropout", "attn_head_size", "attn_output", "attn_pdrop", "attn_weights", "be", "bias", "bool", "but", "by", "cache_kwargs", "cache_position", "caching", "call", "cat", "causal_mask", "class", "config", "contiguous", "cos", "create_sinusoidal_positions", "creating", "def", "device", "dim", "dim_head", "divisible", "dtype", "during", "elif", "else", "embed_dim", "embed_positions", "errors", "f", "float32", "forward", "get_default_dtype", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "is", "k_pass", "k_rot", "key", "layer_idx", "layer_past", "lead", "len", "local_dim", "logger", "make", "matmul", "max_position_embeddings", "max_positions", "mp_num", "must", "n_head", "new_shape", "nn", "not", "num_attention_heads", "of", "one", "or", "out_proj", "output_attentions", "partial_rotation_size", "passing", "permute", "pos_embd_dim", "position_ids", "provide", "q_pass", "q_rot", "qkv", "qkv_proj", "qkv_split", "query", "raise", "rank", "recommended", "reshape", "reshaped", "resid_dropout", "resid_pdrop", "return", "rotary_dim", "scale_attn", "self", "shape", "should", "sin", "sincos", "size", "split", "sqrt", "super", "sure", "tensor", "the", "this", "to", "torch", "transpose", "tuple", "update", "use_cache", "used", "value", "view", "warning_once", "when", "will", "without", "x"], "codegen/modeling_codegen.py:CodeGenMLP": ["ACT2FN", "Dropout", "FloatTensor", "Linear", "ModelMLP", "Module", "Optional", "__init__", "act", "activation_function", "class", "config", "def", "dropout", "embed_dim", "fc_in", "fc_out", "forward", "hidden_states", "intermediate_size", "n_embd", "nn", "resid_pdrop", "return", "self", "super", "torch"], "codegen/modeling_codegen.py:CodeGenBlock": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "LongTensor", "ModelAttention", "ModelBlock", "ModelMLP", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attn", "attn_outputs", "attn_weights", "bool", "cache_position", "class", "config", "def", "else", "eps", "feed_forward_hidden_states", "forward", "head_mask", "hidden_states", "if", "inner_dim", "is", "layer_idx", "layer_norm_epsilon", "layer_past", "ln_1", "mlp", "n_embd", "n_inner", "nn", "not", "output_attentions", "position_ids", "residual", "return", "self", "super", "torch", "tuple", "use_cache"], "codegen/modeling_codegen.py:CodeGenPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelBlock", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "__init__", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "inputs", "is", "isinstance", "kwargs", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "transformer", "weight", "zero_"], "codegen/modeling_codegen.py:CodeGenModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "Dropout", "DynamicCache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelBlock", "ModelModel", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attentions", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "block", "bool", "cache_position", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "drop", "dtype", "else", "embd_pdrop", "embed_dim", "enumerate", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_head_mask", "get_input_embeddings", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "h", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_epsilon", "layer_past", "ln_f", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "n_ctx", "n_embd", "n_layer", "new_embeddings", "nn", "not", "npu", "num_attention_heads", "of", "one", "or", "output_attentions", "output_hidden_states", "output_shape", "outputs", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_ids", "post_init", "r", "raise", "range", "reshape", "return", "return_dict", "rotary_dim", "sdpa", "self", "seq_length", "sequence_length", "set_input_embeddings", "shape", "size", "specify", "staticmethod", "super", "target_length", "to", "token_type_embeds", "token_type_ids", "torch", "training", "triu", "tuple", "type", "unsqueeze", "use_cache", "use_return_dict", "using_compilable_cache", "v", "view", "vocab_size", "warning_once", "with", "wte", "xpu"], "codegen/modeling_codegen.py:CodeGenForCausalLM": ["Cache", "CausalLMOutputWithPast", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "def", "device", "dtype", "else", "float32", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_function", "n_embd", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "to", "token_type_ids", "torch", "transformer", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5MLP": ["ACT2FN", "Linear", "ModelConfig", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "use_bias", "x"], "ernie4_5/modeling_ernie4_5.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "stack", "torch", "x", "x1", "x2"], "ernie4_5/modeling_ernie4_5.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "ernie4_5/modeling_ernie4_5.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "ernie4_5/modeling_ernie4_5.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dim", "dtype", "float", "k", "k_embed", "original_dtype", "position_ids", "q", "q_embed", "repeat_interleave", "return", "rotate_half", "shape", "sin", "to", "unsqueeze", "unsqueeze_dim"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "use_bias", "v_proj", "value_states", "version", "view"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "ernie4_5/modeling_ernie4_5.py:Ernie4_5ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "eomt/modeling_eomt.py:EomtForUniversalSegmentationOutput": ["FloatTensor", "ModelForUniversalSegmentationOutput", "ModelOutput", "None", "Optional", "Tensor", "attentions", "class", "class_queries_logits", "hidden_states", "last_hidden_state", "list", "loss", "masks_queries_logits", "patch_offsets", "r", "torch", "tuple"], "eomt/modeling_eomt.py:sample_point": ["False", "Model_point", "Tensor", "True", "add_dim", "def", "dim", "functional", "grid_Model", "if", "input_features", "kwargs", "nn", "point_coordinates", "point_features", "return", "squeeze", "torch", "unsqueeze"], "eomt/modeling_eomt.py:pair_wise_dice_loss": ["Model_wise_dice_loss", "None", "T", "Tensor", "def", "denominator", "flatten", "inputs", "labels", "loss", "matmul", "numerator", "return", "sigmoid", "sum", "torch"], "eomt/modeling_eomt.py:pair_wise_sigmoid_cross_entropy_loss": ["BCEWithLogitsLoss", "Model_wise_sigmoid_cross_entropy_loss", "T", "Tensor", "criterion", "cross_entropy_loss_neg", "cross_entropy_loss_pos", "def", "height_and_width", "inputs", "labels", "loss", "loss_neg", "loss_pos", "matmul", "nn", "none", "ones_like", "r", "reduction", "return", "shape", "torch", "zeros_like"], "eomt/modeling_eomt.py:EomtHungarianMatcher": ["All", "False", "ModelHungarianMatcher", "Module", "None", "Tensor", "ValueError", "__init__", "align_corners", "and", "append", "array", "as_tensor", "assigned_indices", "batch_size", "be", "can", "class", "class_labels", "class_queries_logits", "cost_class", "cost_dice", "cost_mask", "cost_matrix", "costs", "cpu", "def", "device", "dtype", "float", "for", "forward", "i", "if", "in", "indices", "int", "int64", "j", "linear_sum_assignment", "list", "mask_labels", "masks_queries_logits", "matched_indices", "maximum", "minimum", "nan_to_num", "nn", "no_grad", "np", "num_points", "pair_wise_dice_loss", "pair_wise_sigmoid_cross_entropy_loss", "point_coordinates", "pred_coordinates", "pred_mask", "pred_probs", "raise", "rand", "range", "repeat", "return", "sample_point", "self", "shape", "softmax", "squeeze", "super", "t", "target_coordinates", "target_mask", "tensor", "to", "torch", "tuple"], "eomt/modeling_eomt.py:dice_loss": ["Model_loss", "Tensor", "def", "denominator", "flatten", "inputs", "int", "labels", "loss", "num_masks", "numerator", "probs", "r", "return", "sigmoid", "sum"], "eomt/modeling_eomt.py:sigmoid_cross_entropy_loss": ["BCEWithLogitsLoss", "Model_cross_entropy_loss", "Tensor", "criterion", "cross_entropy_loss", "def", "inputs", "int", "labels", "loss", "mean", "nn", "none", "num_masks", "r", "reduction", "return", "sum", "torch"], "eomt/modeling_eomt.py:EomtLoss": ["CrossEntropyLoss", "False", "ModelConfig", "ModelHungarianMatcher", "ModelLoss", "Module", "None", "Optional", "PartialState", "Tensor", "_", "__init__", "_get_predictions_permutation_indices", "_get_targets_permutation_indices", "_max_by_axis", "_pad_images_to_max_in_batch", "_shared_state", "abs", "align_corners", "arange", "array", "as_tensor", "aux_outputs", "auxiliary_predictions", "batch_indices", "batch_shape", "batch_size", "bool", "calculate_uncertainty", "cat", "clamp", "class", "class_labels", "class_queries_logits", "class_weight", "classes", "config", "copy_", "cost_class", "cost_dice", "cost_mask", "criterion", "def", "del", "device", "dice_loss", "dice_weight", "dict", "dim", "dtype", "empty_weight", "enumerate", "eos_coef", "f", "fill_value", "float", "for", "forward", "full", "full_like", "get_num_masks", "height", "i", "idx", "if", "importance_sample_ratio", "in", "index", "indices", "int", "int64", "is", "is_accelerate_available", "item", "items", "j", "k", "key", "lambda", "len", "list", "logits", "long", "loss_ce", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_labels", "loss_mask", "loss_masks", "losses", "mask_labels", "mask_weight", "masks_queries_logits", "matcher", "max", "max_size", "maxes", "min", "nn", "no_grad", "no_object_weight", "not", "np", "num_boxes", "num_labels", "num_masks", "num_points", "num_points_sampled", "num_processes", "num_queries", "num_random_points", "num_uncertain_points", "ones", "oversample_ratio", "padded_tensor", "padded_tensors", "padding_mask", "padding_masks", "point_coordinates", "point_labels", "point_logits", "point_uncertainties", "pred_logits", "pred_logits_transposed", "pred_masks", "predictions_indices", "rand", "reduce", "register_buffer", "requires_backends", "return", "sample_point", "sample_points_using_uncertainty", "scipy", "self", "shape", "shift", "sigmoid_cross_entropy_loss", "sizes", "squeeze", "src", "src_idx", "str", "sublist", "sum", "super", "target", "target_classes", "target_classes_o", "target_indices", "target_masks", "tensor", "tensors", "tgt", "tgt_idx", "topk", "torch", "train_num_points", "transpose", "tuple", "uncertainty_function", "uncertainty_scores", "update", "value", "view", "weight", "weight_dict", "width", "with", "world_size", "zeros", "zip"], "eomt/modeling_eomt.py:EomtPatchEmbeddings": ["Conv2d", "Expected", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "f", "flatten", "forward", "got", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "with"], "eomt/modeling_eomt.py:EomtEmbeddings": ["Dropout", "Embedding", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Parameter", "Tensor", "_", "__init__", "arange", "batch_size", "cat", "class", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "dtype", "embeddings", "expand", "forward", "hidden_dropout_prob", "hidden_size", "nn", "num_patches", "num_prefix_tokens", "num_register_tokens", "patch_embeddings", "patch_size", "persistent", "pixel_values", "position_embeddings", "position_ids", "projection", "randn", "register_buffer", "register_tokens", "return", "self", "shape", "super", "target_dtype", "to", "torch", "weight", "zeros"], "eomt/modeling_eomt.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "eomt/modeling_eomt.py:EomtAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "eomt/modeling_eomt.py:EomtLayerScale": ["ModelLayerScale", "Module", "None", "Parameter", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_size", "hidden_state", "lambda1", "layerscale_value", "nn", "ones", "return", "self", "super", "torch"], "eomt/modeling_eomt.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "eomt/modeling_eomt.py:EomtDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "eomt/modeling_eomt.py:EomtMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "None", "Tensor", "True", "__init__", "activation", "bias", "class", "config", "def", "else", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "hidden_size", "hidden_state", "if", "in_features", "int", "isinstance", "mlp_ratio", "nn", "out_features", "return", "self", "str", "super", "torch"], "eomt/modeling_eomt.py:EomtSwiGLUFFN": ["Linear", "ModelSwiGLUFFN", "Module", "None", "Tensor", "True", "__init__", "bias", "chunk", "class", "config", "def", "dim", "forward", "functional", "hidden", "hidden_features", "hidden_size", "hidden_state", "in_features", "int", "mlp_ratio", "nn", "out_features", "return", "self", "silu", "super", "torch", "weights_in", "weights_out", "x1", "x2"], "eomt/modeling_eomt.py:EomtLayer": ["GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelLayerScale", "ModelMLP", "ModelSwiGLUFFN", "None", "Optional", "Tensor", "_", "__init__", "attention", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "if", "layer_norm_eps", "layer_output", "layer_scale1", "layer_scale2", "mlp", "nn", "norm1", "norm2", "return", "self", "self_attention_output", "super", "torch", "use_swiglu_ffn"], "eomt/modeling_eomt.py:EomtLayerNorm2d": ["F", "LayerNorm", "ModelLayerNorm2d", "Tensor", "True", "__init__", "affine", "bias", "class", "def", "elementwise_affine", "eps", "forward", "hidden_state", "layer_norm", "nn", "normalized_shape", "num_channels", "permute", "return", "self", "super", "torch", "weight"], "eomt/modeling_eomt.py:EomtScaleLayer": ["ACT2FN", "Conv2d", "ConvTranspose2d", "False", "ModelConfig", "ModelLayerNorm2d", "ModelScaleLayer", "Module", "Tensor", "__init__", "activation", "bias", "class", "config", "conv1", "conv2", "def", "forward", "groups", "hidden_act", "hidden_size", "hidden_states", "kernel_size", "layernorm2d", "nn", "padding", "return", "self", "stride", "super", "torch"], "eomt/modeling_eomt.py:EomtScaleBlock": ["ModelConfig", "ModelScaleBlock", "ModelScaleLayer", "Module", "ModuleList", "Tensor", "_", "__init__", "block", "class", "config", "def", "for", "forward", "hidden_states", "in", "nn", "num_blocks", "num_upscale_blocks", "range", "return", "self", "super", "torch"], "eomt/modeling_eomt.py:EomtMaskHead": ["ACT2FN", "Linear", "ModelConfig", "ModelMaskHead", "Module", "Tensor", "__init__", "activation", "class", "config", "def", "fc1", "fc2", "fc3", "forward", "hidden_act", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch"], "eomt/modeling_eomt.py:EomtPreTrainedModel": ["Conv2d", "ConvTranspose2d", "Embedding", "False", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelLayerScale", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_", "_calculate_fan_in_and_fan_out", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_sdpa", "a", "attentions", "base_model_prefix", "bias", "bound", "class", "cls_token", "config", "data", "def", "dtype", "elif", "else", "fan_in", "fill_", "float32", "hasattr", "hidden_states", "if", "init", "initializer_range", "is", "isinstance", "kaiming_uniform_", "lambda1", "layerscale_value", "main_input_name", "math", "mean", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "register_tokens", "self", "sqrt", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "uniform_", "weight", "zero_"], "eomt/modeling_eomt.py:EomtForUniversalSegmentation": ["Embedding", "F", "LayerNorm", "Linear", "ModelConfig", "ModelEmbeddings", "ModelForUniversalSegmentation", "ModelForUniversalSegmentationOutput", "ModelLayer", "ModelLoss", "ModelMaskHead", "ModelPreTrainedModel", "ModelScaleBlock", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "_", "__init__", "_disable_attention_mask", "and", "attention_mask", "attn_mask", "attn_mask_probs", "auto_docstring", "auxiliary_predictions", "bchw", "bilinear", "bool", "bqc", "bqhw", "cat", "check_model_inputs", "class", "class_labels", "class_logits", "class_predictor", "class_queries_logits", "class_queries_logits_per_layer", "class_weight", "config", "criterion", "def", "device", "dice_weight", "dict", "dim", "dtype", "einsum", "embeddings", "encoder_start_tokens", "enumerate", "eps", "expand", "float", "for", "forward", "get_input_embeddings", "get_loss", "get_loss_dict", "grid_size", "have", "hidden_size", "hidden_states", "idx", "if", "image_size", "in", "interpolate", "interpolated_logits", "is", "items", "key", "kwargs", "last_hidden_state", "layer_module", "layer_norm_eps", "layernorm", "layers", "list", "logits", "loss", "loss_cross_entropy", "loss_dice", "loss_dict", "loss_key", "loss_mask", "main_input_name", "mask_head", "mask_labels", "mask_logits", "mask_weight", "masked_fill", "masks_queries_logits", "masks_queries_logits_per_layer", "mode", "nn", "norm_hidden_states", "not", "num_attention_heads", "num_blocks", "num_hidden_layers", "num_labels", "num_prefix_tokens", "num_queries", "num_query_tokens", "ones", "or", "patch_embeddings", "patch_offsets", "patch_size", "pixel_values", "post_init", "predict", "prefix_tokens", "prob", "query", "query_tokens", "r", "raise", "rand", "random_queries", "range", "register_buffer", "reshape", "return", "self", "sequence_output", "shape", "size", "specify", "staticmethod", "str", "sum", "super", "to", "torch", "training", "transpose", "upscale_block", "values", "view", "weight", "weight_dict", "zip"], "parakeet/modeling_parakeet.py:ParakeetEncoderRelPositionalEncoding": ["False", "Length", "ModelEncoderConfig", "ModelEncoderRelPositionalEncoding", "Module", "None", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "autocast", "base", "be", "class", "config", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "enabled", "equal", "expand", "f", "float", "forward", "freqs", "has", "hidden_size", "hidden_states", "if", "int64", "inv_freq", "inv_freq_expanded", "isinstance", "less", "max_position_embeddings", "mps", "nn", "no_grad", "or", "persistent", "pos_embed", "position_ids", "position_ids_expanded", "raise", "register_buffer", "reshape", "return", "self", "seq_length", "shape", "sin", "stack", "str", "super", "than", "to", "torch", "transpose", "type", "with"], "parakeet/modeling_parakeet.py:ParakeetEncoderFeedForward": ["ACT2FN", "Linear", "ModelEncoderConfig", "ModelEncoderFeedForward", "Module", "__init__", "activation", "activation_dropout", "attention_bias", "bias", "class", "config", "def", "dropout", "forward", "functional", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "linear1", "linear2", "nn", "p", "return", "self", "super", "training"], "parakeet/modeling_parakeet.py:ParakeetEncoderConvolutionModule": ["ACT2FN", "BatchNorm1d", "Conv1d", "ModelEncoderConfig", "ModelEncoderConvolutionModule", "Module", "None", "True", "__init__", "activation", "all", "all_masked_rows", "attention_mask", "bias", "channels", "class", "config", "conv_kernel_size", "def", "depthwise_conv", "dim", "else", "forward", "functional", "get", "getattr", "glu", "groups", "hidden_act", "hidden_size", "hidden_states", "if", "is", "kernel_size", "masked_fill", "module_config", "nn", "norm", "not", "padding", "pointwise_conv1", "pointwise_conv2", "return", "self", "silu", "stride", "super", "torch", "transpose"], "parakeet/modeling_parakeet.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "parakeet/modeling_parakeet.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "parakeet/modeling_parakeet.py:ParakeetEncoderAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelEncoderAttention", "ModelEncoderConfig", "Module", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_attn_implementation", "_rel_shift", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attention_scores", "attn_output", "attn_weights", "batch_size", "bias", "bias_u", "bias_v", "class", "config", "contiguous", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "float", "forward", "functional", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "inf", "input_shape", "int", "is", "is_causal", "k_proj", "key", "key_states", "kwargs", "layer_idx", "logical_not", "masked_fill_", "matrix_bd", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "pad", "past_key_value", "past_key_values", "permute", "position_embeddings", "position_length", "q_proj", "query", "query_length", "query_states", "query_states_with_bias_u", "query_states_with_bias_v", "relative_k_proj", "relative_key_states", "reshape", "return", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "value", "value_states", "version", "view", "zeros"], "parakeet/modeling_parakeet.py:ParakeetEncoderSubsamplingConv2D": ["Conv2d", "Linear", "ModelEncoderConfig", "ModelEncoderSubsamplingConv2D", "Module", "ModuleList", "None", "ReLU", "Tensor", "True", "__init__", "_get_output_length", "and", "append", "arange", "attention_mask", "bias", "channel_mask", "channels", "class", "config", "conv_layer", "current_lengths", "current_seq_length", "def", "device", "else", "for", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "i", "if", "in", "input_features", "input_lengths", "int", "is", "isinstance", "kernel_size", "layer", "layers", "linear", "log2", "math", "nn", "not", "num_layers", "num_mel_bins", "out_length", "output_lengths", "padding", "range", "reshape", "return", "self", "shape", "stride", "subsampling_conv_channels", "subsampling_conv_kernel_size", "subsampling_conv_stride", "subsampling_factor", "sum", "super", "torch", "transpose", "unsqueeze"], "parakeet/modeling_parakeet.py:ParakeetEncoderBlock": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelEncoderAttention", "ModelEncoderBlock", "ModelEncoderConfig", "ModelEncoderConvolutionModule", "ModelEncoderFeedForward", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attn_output", "class", "config", "conv", "conv_output", "def", "feed_forward1", "feed_forward2", "ff2_output", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "nn", "norm_conv", "norm_feed_forward1", "norm_feed_forward2", "norm_out", "norm_self_att", "normalized_hidden_states", "position_embeddings", "residual", "return", "self", "self_attn", "super", "torch"], "parakeet/modeling_parakeet.py:ParakeetPreTrainedModel": ["False", "ModelCTCConfig", "ModelEncoderAttention", "ModelEncoderBlock", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Tensor", "True", "_", "_can_compile_fullgraph", "_can_record_outputs", "_get_output_attention_mask", "_get_subsampling_output_length", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flat_attention_mask", "_supports_flex_attn", "_supports_sdpa", "add_pad", "all_paddings", "arange", "attention_mask", "attentions", "base_model_prefix", "bias_u", "bias_v", "class", "config", "data", "def", "device", "div", "dtype", "else", "encoder_config", "float", "floor", "for", "get_text_config", "getattr", "hasattr", "hidden_states", "if", "in", "initializer_range", "input_features", "input_lengths", "int", "is", "isinstance", "kernel_size", "lengths", "log2", "main_input_name", "math", "max", "max_length", "mean", "model", "module", "normal_", "not", "num_layers", "output_lengths", "range", "return", "self", "std", "stride", "subsampling_conv_kernel_size", "subsampling_conv_stride", "subsampling_factor", "sum", "super", "supports_gradient_checkpointing", "target_length", "to", "torch"], "parakeet/modeling_parakeet.py:ParakeetEncoder": ["BaseModelOutput", "False", "ModelEncoder", "ModelEncoderBlock", "ModelEncoderConfig", "ModelEncoderRelPositionalEncoding", "ModelEncoderSubsamplingConv2D", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_get_output_attention_mask", "attention_mask", "auto_docstring", "base_model_prefix", "can_return_tuple", "check_model_inputs", "class", "config", "def", "dropout", "dropout_positions", "dropout_probability", "else", "encode_positions", "encoder", "encoder_layer", "expand", "for", "forward", "functional", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_features", "input_scale", "is", "kwargs", "last_hidden_state", "layer_idx", "layerdrop", "layers", "math", "nn", "not", "num_hidden_layers", "p", "position_embeddings", "post_init", "r", "rand", "range", "return", "scale_input", "self", "shape", "sqrt", "subsampling", "super", "target_length", "to_drop", "torch", "training", "transpose", "unsqueeze"], "parakeet/modeling_parakeet.py:ParakeetGenerateOutput": ["FloatTensor", "LongTensor", "ModelGenerateOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "sequences", "torch", "tuple"], "parakeet/modeling_parakeet.py:ParakeetForCTC": ["CausalLMOutput", "Conv1d", "False", "LongTensor", "ModelCTCConfig", "ModelEncoder", "ModelForCTC", "ModelGenerateOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_get_output_attention_mask", "_get_subsampling_output_length", "argmax", "attention_mask", "attentions", "auto_docstring", "backends", "blank", "bool", "can_return_tuple", "class", "config", "ctc_head", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "dim", "dtype", "else", "enabled", "encoder", "encoder_config", "encoder_outputs", "flags", "flattened_targets", "float32", "forward", "functional", "generate", "hidden_size", "hidden_states", "if", "input_features", "input_lengths", "is", "kernel_size", "kwargs", "labels", "labels_mask", "last_hidden_state", "log_probs", "log_softmax", "logits", "long", "loss", "masked_select", "nn", "no_grad", "not", "ones_like", "outputs", "pad_token_id", "post_init", "r", "reduction", "return", "return_dict", "return_dict_in_generate", "self", "sequences", "shape", "sum", "super", "target_length", "target_lengths", "torch", "transpose", "vocab_size", "with", "zero_infinity"], "seggpt/modeling_seggpt.py:SegGptEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "intermediate_hidden_states", "last_hidden_state", "r", "torch", "tuple"], "seggpt/modeling_seggpt.py:SegGptImageSegmentationOutput": ["FloatTensor", "ModelImageModelmentationOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "pred_masks", "r", "torch", "tuple"], "seggpt/modeling_seggpt.py:SegGptPatchEmbeddings": ["Conv2d", "Input", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "forward", "height", "hidden_size", "if", "image", "image_size", "in", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_patches", "of", "one", "or", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "values", "width", "with"], "seggpt/modeling_seggpt.py:SegGptEmbeddings": ["BoolTensor", "Dropout", "Embedding", "F", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Modelment_token_input", "Modelment_token_prompt", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "batch_size", "be", "bicubic", "bool_masked_pos", "but", "cat", "class", "config", "def", "dim", "dropout", "either", "elif", "else", "embedding_type", "embeddings", "expand", "f", "forward", "got", "height", "hidden_dropout_prob", "hidden_size", "if", "input_embeddings", "instance", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask_token", "mode", "nn", "not", "num_patches", "num_positions", "or", "patch_embeddings", "patch_height", "patch_pos_embed", "patch_size", "patch_width", "permute", "pixel_values", "pos_embed", "position_embeddings", "pretrain_image_size", "pretrain_patch_size", "prompt_embeddings", "prompt_pixel_values", "raise", "randn", "reshape", "return", "self", "semantic", "shape", "should", "size", "str", "super", "torch", "torch_int", "type", "type_as", "type_embedding", "type_token_instance", "type_token_semantic", "unsqueeze", "w", "width", "zeros"], "seggpt/modeling_seggpt.py:SegGptAttention": ["F", "False", "Input", "Iterable", "Linear", "ModelAttention", "Module", "None", "Parameter", "Tensor", "ValueError", "_", "__init__", "abc", "add_decomposed_rel_pos", "arange", "attn", "attn_output", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bhwc", "bhwk", "bias", "class", "collections", "config", "def", "dim", "dtype", "einsum", "else", "encoding", "float32", "forward", "functional", "get_rel_pos", "head_dim", "height", "hidden_size", "hidden_states", "hkc", "if", "image_size", "input_size", "int", "interpolate", "is", "isinstance", "k_coords", "k_size", "key", "key_height", "key_width", "linear", "long", "max", "max_rel_dist", "mode", "must", "nn", "num_attention_heads", "output_attentions", "patch_size", "permute", "positional", "proj", "provided", "q_coords", "q_size", "qkv", "qkv_bias", "query", "query_height", "query_width", "raise", "rel_h", "rel_pos", "rel_pos_h", "rel_pos_resized", "rel_pos_w", "rel_w", "relative", "relative_coords", "relative_position_height", "relative_position_width", "reshape", "reshaped_query", "return", "scale", "self", "shape", "size", "softmax", "super", "to", "torch", "transpose", "tuple", "unbind", "use_relative_position_embeddings", "using", "value", "view", "width", "wkc", "zeros"], "seggpt/modeling_seggpt.py:SegGptMlp": ["ACT2FN", "Linear", "ModelMlp", "Module", "Tensor", "__init__", "act", "class", "config", "def", "forward", "hidden_act", "hidden_size", "hidden_states", "lin1", "lin2", "mlp_dim", "nn", "return", "self", "super", "torch"], "seggpt/modeling_seggpt.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "seggpt/modeling_seggpt.py:SegGptDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "seggpt/modeling_seggpt.py:SegGptLayer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelMlp", "None", "Tensor", "True", "Union", "__init__", "and", "attention", "attention_output", "bool", "cat", "class", "config", "def", "dim", "drop_path", "drop_path_rate", "else", "ensemble_cond", "eps", "expand_as", "feature_ensemble", "float", "forward", "hidden_size", "hidden_states", "if", "inputs", "int", "keepdim", "layer_norm_eps", "layernorm_after", "layernorm_before", "mean", "mlp", "nn", "num_prompts", "output_attentions", "outputs", "prompt", "reshape", "residual", "return", "self", "self_attention_outputs", "shape", "split", "super", "torch", "tuple"], "seggpt/modeling_seggpt.py:SegGptEncoder": ["False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderOutput", "ModelLayer", "Module", "ModuleList", "None", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "append", "attentions", "bool", "class", "config", "cpu", "def", "device", "dpr", "drop_path_rate", "else", "ensemble_cond", "enumerate", "eps", "feature_ensemble", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "intermediate_hidden_state_indices", "intermediate_hidden_states", "is", "item", "last_hidden_state", "layer_module", "layer_norm_eps", "layer_outputs", "layernorm", "layers", "linspace", "merge_index", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "v", "x"], "seggpt/modeling_seggpt.py:SegGptLayerNorm": ["LayerNorm", "ModelLayerNorm", "NotImplementedError", "Tensor", "Unsupported", "__init__", "channels_first", "channels_last", "class", "data", "data_format", "def", "else", "eps", "f", "features", "format", "forward", "if", "in", "kwargs", "nn", "normalized_shape", "not", "permute", "r", "raise", "return", "self", "super", "torch"], "seggpt/modeling_seggpt.py:SegGptDecoderHead": ["ACT2FN", "Conv2d", "FloatTensor", "ModelDecoderHead", "ModelLayerNorm", "Module", "True", "__init__", "act_fct", "bias", "channels_first", "class", "config", "conv", "data_format", "decoder_hidden_size", "def", "eps", "forward", "head", "hidden_act", "hidden_states", "kernel_size", "layer_norm_eps", "layernorm", "nn", "normalized_shape", "padding", "return", "self", "super", "torch"], "seggpt/modeling_seggpt.py:SegGptDecoder": ["FloatTensor", "Linear", "ModelDecoder", "ModelDecoderHead", "Module", "True", "_", "__init__", "_reshape_hidden_states", "batch_size", "bias", "class", "config", "decoder_embed", "decoder_hidden_size", "decoder_pred", "def", "forward", "hidden_size", "hidden_states", "intermediate_hidden_state_indices", "len", "nn", "patch_height", "patch_size", "patch_width", "permute", "reshape", "return", "self", "shape", "super", "torch"], "seggpt/modeling_seggpt.py:SegGptPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelLayerNorm", "ModelPreTrainedModel", "Modelment_token_input", "Modelment_token_prompt", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "dtype", "elif", "fill_", "float32", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "model", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "rel_pos_h", "rel_pos_w", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "type_token_instance", "type_token_semantic", "weight", "zero_"], "seggpt/modeling_seggpt.py:SegGptModel": ["BoolTensor", "False", "FloatTensor", "If", "It", "Labels", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelEncoderOutput", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_prune_heads", "a", "and", "attention", "auto_docstring", "be", "bool", "bool_masked_pos", "bool_masked_pos_ones", "bool_masked_pos_zeros", "but", "cat", "class", "config", "def", "default", "device", "dict", "dim", "dtype", "else", "embedding_output", "embedding_type", "embeddings", "encoder", "encoder_outputs", "expected_dtype", "feature_ensemble", "for", "forward", "get_input_embeddings", "heads", "heads_to_prune", "if", "in", "int", "is", "items", "labels", "layer", "list", "logger", "make", "model", "not", "num_patches", "ones", "output_attentions", "output_hidden_states", "patch_embeddings", "pixel_values", "post_init", "projection", "prompt_masks", "prompt_pixel_values", "provide", "provided", "prune_heads", "r", "re", "return", "return_dict", "self", "set", "str", "super", "sure", "the", "to", "torch", "training", "tuple", "unsqueeze", "use_return_dict", "value", "warning_once", "weight", "were", "will", "you", "zeros"], "seggpt/modeling_seggpt.py:patchify": ["Model", "Tensor", "batch_size", "def", "height", "int", "num_channels", "patch_height", "patch_size", "patch_width", "permute", "reshape", "return", "shape", "tensor", "torch", "width"], "seggpt/modeling_seggpt.py:unpatchify": ["Model", "Number", "Tensor", "ValueError", "and", "batch_size", "def", "does", "f", "height", "if", "int", "match", "not", "of", "patch", "patch_height", "patch_size", "patch_width", "patches", "permute", "raise", "reshape", "return", "shape", "tensor", "torch", "width"], "seggpt/modeling_seggpt.py:SegGptLoss": ["BoolTensor", "F", "FloatTensor", "ModelLoss", "Module", "None", "__init__", "beta", "bool_masked_pos", "cat", "class", "config", "def", "dim", "forward", "ground_truth", "labels", "loss", "mask", "nn", "none", "patch_size", "pred_masks", "prompt_masks", "reduction", "repeat", "return", "self", "shape", "smooth_l1_loss", "sum", "super", "torch", "unpatchify"], "seggpt/modeling_seggpt.py:SegGptForImageSegmentation": ["BoolTensor", "FloatTensor", "ModelConfig", "ModelDecoder", "ModelForImageModelmentation", "ModelImageModelmentationOutput", "ModelLoss", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "attentions", "auto_docstring", "bool", "bool_masked_pos", "bool_masked_pos_ones", "bool_masked_pos_zeros", "cat", "class", "config", "decoder", "def", "device", "dim", "dtype", "else", "embedding_type", "embeddings", "feature_ensemble", "forward", "hidden_states", "idx", "if", "intermediate_hidden_states", "is", "labels", "loss", "loss_fn", "model", "not", "num_patches", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "patch_embeddings", "pixel_values", "post_init", "pred_masks", "prompt_masks", "prompt_pixel_values", "r", "return", "return_dict", "self", "str", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "zeros"], "dia/modeling_dia.py:DiaPreTrainedModel": ["ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "input_ids", "main_input_name", "model", "supports_gradient_checkpointing"], "dia/modeling_dia.py:DiaMultiChannelEmbedding": ["Embedding", "False", "ModelDecoderConfig", "ModelMultiChannelEmbedding", "Module", "Tensor", "__init__", "arange", "audio_codes", "class", "config", "def", "device", "dim", "dtype", "embed", "embeds", "forward", "hidden_size", "long", "nn", "num_channels", "offsets", "persistent", "register_buffer", "return", "self", "shape", "squeeze", "sum", "super", "to", "tokens", "torch", "view", "vocab_size"], "dia/modeling_dia.py:DiaMLP": ["ACT2FN", "False", "FloatTensor", "Linear", "ModelMLP", "Module", "__init__", "activation_fn", "bias", "chunk", "class", "config", "def", "dim", "down_proj", "forward", "gate", "gate_up_proj", "hidden_act", "hidden_size", "hidden_states", "intermeModelte_size", "nn", "return", "self", "super", "torch", "up_states"], "dia/modeling_dia.py:DiaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "dia/modeling_dia.py:DiaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "dia/modeling_dia.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "dia/modeling_dia.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "dia/modeling_dia.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "dia/modeling_dia.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "dia/modeling_dia.py:DiaSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "Linear", "LongTensor", "ModelDecoderConfig", "ModelEncoderConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "dia/modeling_dia.py:DiaCrossAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Linear", "ModelCrossAttention", "ModelDecoderConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "class", "config", "contiguous", "cross_attention_cache", "cross_attention_states", "cross_head_dim", "cross_hidden_size", "cross_num_attention_heads", "cross_num_key_value_heads", "cross_shape", "def", "eager", "eager_attention_forward", "else", "forward", "get", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "is_updated", "k_proj", "key_states", "keys", "kwargs", "layer_idx", "layers", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_values", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "super", "torch", "transpose", "tuple", "update", "v_proj", "value_states", "values", "view"], "dia/modeling_dia.py:DiaEncoderLayer": ["False", "FlashAttentionKwargs", "GradientCheckpointingLayer", "ModelEncoderConfig", "ModelEncoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSelfAttention", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_mask", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_states", "int", "is_causal", "kwargs", "layer_idx", "mlp", "mlp_out", "norm_eps", "normed_states", "position_embeddings", "post_sa_norm", "pre_sa_norm", "residual", "return", "self", "self_attention", "self_attn_output", "self_attn_weights", "super", "torch", "tuple"], "dia/modeling_dia.py:DiaEncoder": ["BaseModelOutput", "Embedding", "False", "FlashAttentionKwargs", "ModelEncoder", "ModelEncoderConfig", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "Union", "Unpack", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_attentions", "arange", "attention_mask", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "device", "dtype", "elif", "else", "embedding", "encoder_layer", "encoder_states", "eps", "flash", "flex_attention", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "is_causal", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "make_flex_block_causal_mask", "nn", "norm", "norm_eps", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "position_embeddings", "position_ids", "range", "return", "rotary_embeddings", "sdpa", "self", "shape", "super", "torch", "tuple", "vocab_size"], "dia/modeling_dia.py:DiaDecoderLayer": ["EncoderDecoderCache", "GradientCheckpointingLayer", "LongTensor", "ModelCrossAttention", "ModelDecoderConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "ModelSelfAttention", "None", "Optional", "Tensor", "True", "__init__", "attention_mask", "cache_position", "class", "config", "cross_attention", "cross_attn_weights", "cross_states", "def", "embed_dim", "encoder_attention_mask", "encoder_hidden_states", "eps", "forward", "hidden_size", "hidden_states", "if", "int", "is_causal", "isinstance", "kwargs", "layer_idx", "mlp", "mlp_out", "norm_eps", "normed_states", "past_key_values", "position_embeddings", "pre_ca_norm", "pre_mlp_norm", "pre_sa_norm", "residual", "return", "self", "self_attention", "self_attention_cache", "self_attn_cache", "self_attn_output", "self_attn_weights", "super", "torch", "tuple"], "dia/modeling_dia.py:DiaDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "EncoderDecoderCache", "False", "FloatTensor", "LongTensor", "ModelDecoder", "ModelDecoderConfig", "ModelDecoderLayer", "ModelMultiChannelEmbedding", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Size", "Tensor", "Union", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_cross_attn_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "class", "config", "create_causal_mask", "cross_attentions", "def", "device", "dtype", "elif", "else", "embeddings", "encoder_attention_mask", "encoder_hidden_states", "eps", "flash", "flex_attention", "for", "forward", "get_seq_length", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "is", "is_causal", "is_torchdynamo_compiling", "isinstance", "kwargs", "last_hidden_state", "layer", "layer_idx", "layer_outputs", "layers", "make_flex_block_causal_mask", "mask_seq_length", "nn", "norm", "norm_eps", "not", "num_channels", "num_hidden_layers", "ones", "output_attentions", "output_hidden_states", "past_key_values", "past_key_values_length", "position_embeddings", "position_ids", "query_length", "r", "range", "return", "rotary_embeddings", "sdpa", "self", "seq_length", "shape", "size", "super", "tgt_len", "torch", "tuple", "vocab_size"], "dia/modeling_dia.py:DiaModel": ["BaseModelOutput", "DynamicCache", "EncoderDecoderCache", "False", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "Neither", "None", "Optional", "Seq2SeqModelOutput", "Setting", "True", "Union", "ValueError", "You", "__init__", "and", "attention_mask", "attentions", "auto_docstring", "been", "bool", "bos_token_id", "bsz", "cache_position", "cached", "can_return_tuple", "channels", "checkpointing", "class", "config", "cross_attentions", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_hidden_states", "decoder_input_ids", "decoder_outputs", "decoder_position_ids", "def", "device", "either", "elif", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "encodings", "fill_value", "forward", "found", "full", "get_encoder", "gradient", "has", "hidden_states", "ids", "if", "incompatible", "input_ids", "is", "is_gradient_checkpointing", "isinstance", "kwargs", "last_hidden_state", "len", "logger", "ndim", "not", "num_channels", "or", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "provide", "r", "raise", "reshape", "return", "self", "seq_len", "shape", "should", "size", "super", "text", "the", "torch", "training", "transpose", "tuple", "use_cache", "warning_once", "with"], "dia/modeling_dia.py:DiaForConditionalGeneration": ["BaseModelOutput", "EncoderDecoderCache", "False", "ForMaskedLM", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelGenerationMixin", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Union", "__init__", "attention_mask", "audio_logits", "auto_docstring", "base_model_prefix", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "contiguous", "cross_attentions", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_hidden_states", "decoder_input_ids", "decoder_position_ids", "def", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_encoder", "hidden_size", "if", "input_ids", "is", "kwargs", "labels", "last_hidden_state", "logits", "logits_dense", "loss", "loss_function", "loss_type", "model", "nn", "not", "num_channels", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "self", "shape", "super", "torch", "transpose", "tuple", "use_cache", "view", "vocab_size"], "pegasus_x/modeling_pegasus_x.py:DimensionInfo": ["ModelInfo", "batch_size", "block_size", "class", "dim_per_head", "global_len", "hidden_dim", "int", "num_blocks", "num_heads", "padded_seq_len", "seq_len"], "pegasus_x/modeling_pegasus_x.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "pegasus_x/modeling_pegasus_x.py:PegasusXScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "pegasus_x/modeling_pegasus_x.py:PegasusXSinusoidalPositionalEmbedding": ["ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "batch_size", "class", "cos", "def", "device", "div_term", "dtype", "embed_dim", "exp", "expand", "float", "forward", "half_d_feature", "if", "input_embeds", "int", "int64", "is", "log", "long", "max_scale", "nn", "no_grad", "np", "past_key_values_length", "pe", "position_ids", "return", "self", "seq_len", "shape", "sin", "super", "torch", "type_as", "zeros"], "pegasus_x/modeling_pegasus_x.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "pegasus_x/modeling_pegasus_x.py:PegasusXAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "pegasus_x/modeling_pegasus_x.py:PegasusXGlobalLocalAttention": ["BHGF", "BHGX", "BHNKF", "BHNKG", "BHNKX", "BHNXF", "BHXF", "DimensionInfo", "False", "Linear", "ModelGlobalLocalAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "batch_size", "be", "bias", "block_size", "blocked_local2global", "blocked_local2local", "blocked_local_k", "blocked_local_q", "blocked_local_v", "bool", "bsz", "by", "cat", "class", "compute_global_attention_representations", "compute_local_attention_representations", "contiguous", "def", "dim", "dim_per_head", "divisible", "dropout", "einsum", "else", "embed_dim", "extended_mask", "f", "float", "forward", "functional", "global", "global_and_local_k", "global_and_local_v", "global_attn_output", "global_attn_probs", "global_hidden_states", "global_k", "global_len", "global_q", "global_v", "got", "head_dim", "hidden_dim", "if", "int", "is_decoder", "k_proj", "local", "local2global_attn_output", "local2global_attn_probs", "local2local_attn_output", "local2local_attn_probs", "local_attn_output", "local_attn_probs", "local_k", "local_q", "local_v", "mask", "must", "nn", "num_blocks", "num_heads", "out_proj", "output_attentions", "p", "pad", "padded_seq_len", "permute", "q_proj", "raise", "return", "scaling", "self", "seq_len", "shape", "softmax", "super", "tensor", "token_hidden_states", "torch", "training", "transpose", "tuple", "v_proj", "value", "view"], "pegasus_x/modeling_pegasus_x.py:PegasusXEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderLayer", "ModelGlobalLocalAttention", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "block_size", "bool", "class", "classmethod", "cls", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "global_hidden_states", "global_residual", "global_self_attn_layer_norm", "hidden_states", "if", "mask_min_value", "min", "nn", "num_heads", "output_attentions", "outputs", "p", "pad", "pad_local_tokens", "pad_size", "padded_hidden_states", "padded_mask", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "stagger_blocks_this_layer", "super", "token_hidden_states", "torch", "training", "unpad_local_tokens", "value"], "pegasus_x/modeling_pegasus_x.py:PegasusXDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bias", "bool", "cache_position", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_decoder", "key_value_states", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "pegasus_x/modeling_pegasus_x.py:PegasusXPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "def", "device", "diagonal", "dim", "dtype", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "nn", "normal_", "not", "npu", "ones", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "query_length", "r", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "pegasus_x/modeling_pegasus_x.py:PegasusXEncoder": ["BaseModelOutput", "Embedding", "False", "LayerNorm", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Setting", "True", "ValueError", "You", "_", "__init__", "all_attentions", "and", "arange", "at", "attention_mask", "attentions", "batch_size", "block_size", "bool", "both", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_dim", "embed_global", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "expand", "f", "finfo", "for", "forward", "functional", "get_position_embeddings", "global_hidden_states", "gradient_checkpointing", "have", "hidden_states", "i", "idx", "if", "in", "info", "input_ids", "input_shape", "inputs_embeds", "int", "inverted_mask", "is", "last_hidden_state", "layer_norm", "layer_outputs", "layerdrop", "layers", "logger", "mask_min_value", "masked_fill", "math", "max_position_embeddings", "max_source_positions", "min", "new_num_position_embeddings", "nn", "not", "num_global_tokens", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad", "pad_len", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "resize_position_embeddings", "return", "return_dict", "same", "scale_embedding", "self", "seq_len", "shape", "size", "specify", "sqrt", "stagger_blocks_this_layer", "stagger_local_blocks", "super", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "value", "view", "vocab_size", "warn_if_padding_and_no_attention_mask"], "pegasus_x/modeling_pegasus_x.py:PegasusXDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Transformers", "True", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "batch_size", "be", "both", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "last_hidden_state", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "logger", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with"], "pegasus_x/modeling_pegasus_x.py:PegasusXModel": ["BaseModelOutput", "Cache", "Embedding", "FloatTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "get_position_embeddings", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "last_hidden_state", "len", "math", "max_position_embeddings", "new_num_position_embeddings", "nn", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "resize_position_embeddings", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "sqrt", "super", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "pegasus_x/modeling_pegasus_x.py:PegasusXForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "Union", "__init__", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "changed", "class", "config", "cross_attentions", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_encoder", "get_position_embeddings", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "max_position_embeddings", "model", "new_num_position_embeddings", "nn", "not", "num_embeddings", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "resize_position_embeddings", "return", "return_dict", "self", "shared", "shift_tokens_right", "since", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight"], "pegasus_x/modeling_pegasus_x.py:PegasusXDecoderWrapper": ["ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "speech_to_text/modeling_speech_to_text.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "speech_to_text/modeling_speech_to_text.py:Conv1dSubsampler": ["Model", "ModelSubsampler", "Module", "ModuleList", "__init__", "class", "config", "contiguous", "conv", "conv_channels", "conv_kernel_sizes", "conv_layers", "d_model", "def", "dim", "else", "enumerate", "for", "forward", "functional", "glu", "hidden_states", "i", "if", "in", "in_channels", "input_channels", "input_feat_per_channel", "input_features", "k", "kernel_size", "kernel_sizes", "mid_channels", "nn", "num_conv_layers", "num_layers", "out_channels", "padding", "return", "self", "stride", "super", "transpose"], "speech_to_text/modeling_speech_to_text.py:Speech2TextSinusoidalPositionalEmbedding": ["False", "ModelTextSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "cos", "create_position_ids_from_input_ids", "cumsum", "def", "detach", "device", "dim", "dtype", "emb", "emb_weights", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "speech_to_text/modeling_speech_to_text.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "speech_to_text/modeling_speech_to_text.py:Speech2TextAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Linear", "ModelTextAttention", "ModelTextConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "class", "config", "contiguous", "cross_attention_cache", "curr_past_key_value", "current_states", "def", "deprecate_kwarg", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "past_key_value", "past_key_values", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "values", "version", "view"], "speech_to_text/modeling_speech_to_text.py:Speech2TextEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelTextAttention", "ModelTextConfig", "ModelTextEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "speech_to_text/modeling_speech_to_text.py:Speech2TextDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelTextAttention", "ModelTextConfig", "ModelTextDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "speech_to_text/modeling_speech_to_text.py:Speech2TextPreTrainedModel": ["Conv1d", "Embedding", "False", "Linear", "LongTensor", "ModelTextConfig", "ModelTextPreTrainedModel", "None", "PreTrainedModel", "True", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "arange", "attention_mask", "base_model_prefix", "bias", "bsz", "class", "config", "cumsum", "data", "def", "device", "dtype", "elif", "feature_vector_length", "flip", "for", "i", "if", "in", "init_std", "input_features", "input_lengths", "is", "isinstance", "len", "long", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "num_conv_layers", "padding_idx", "range", "return", "self", "shape", "size", "std", "subsampled_lengths", "sum", "supports_gradient_checkpointing", "torch", "weight", "zero_", "zeros"], "speech_to_text/modeling_speech_to_text.py:Speech2TextEncoder": ["BaseModelOutput", "Conv1dSubsampler", "False", "LayerNorm", "ModelTextConfig", "ModelTextEncoder", "ModelTextEncoderLayer", "ModelTextPreTrainedModel", "ModelTextSinusoidalPositionalEmbedding", "ModuleList", "None", "Tensor", "The", "True", "Union", "_", "__init__", "_attn_implementation", "_get_feature_vector_attention_mask", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_attentions", "assert", "attention_mask", "attentions", "be", "but", "class", "config", "conv", "d_model", "def", "device", "dropout", "dropout_probability", "dtype", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "flash", "flex_attention", "for", "forward", "functional", "gradient_checkpointing", "head_mask", "hidden_states", "idx", "if", "in", "input_features", "inputs_embeds", "is", "is_causal", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "long", "make_flex_block_causal_mask", "math", "max_source_positions", "ne", "nn", "not", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "padding_mask", "post_init", "r", "rand", "range", "return", "return_dict", "scale_embedding", "sdpa", "self", "shape", "should", "size", "specified", "sqrt", "super", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "zeros"], "speech_to_text/modeling_speech_to_text.py:Speech2TextDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelTextConfig", "ModelTextDecoder", "ModelTextDecoderLayer", "ModelTextPreTrainedModel", "ModelTextSinusoidalPositionalEmbedding", "ModuleList", "None", "Passing", "Setting", "Size", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask", "_prepare_4d_causal_attention_mask_for_sdpa", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "assert", "at", "attention_mask", "attentions", "attn_mask", "be", "both", "but", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "flash_attention_2", "flex_attention", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "int", "is", "is_causal", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "make_flex_block_causal_mask", "mask_name", "math", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "query_length", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "sdpa", "self", "should", "size", "specified", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with", "zip"], "speech_to_text/modeling_speech_to_text.py:Speech2TextModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextDecoder", "ModelTextEncoder", "ModelTextModel", "ModelTextPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_get_feature_vector_attention_mask", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_features", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "not", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "shape", "super", "torch", "tuple", "use_cache", "use_return_dict", "value"], "speech_to_text/modeling_speech_to_text.py:Speech2TextForConditionalGeneration": ["Cache", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelTextConfig", "ModelTextForConditionalGeneration", "ModelTextModel", "ModelTextPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_start_token_id", "def", "else", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_features", "is", "labels", "lm_head", "lm_logits", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "post_init", "r", "return", "return_dict", "self", "shift_tokens_right", "super", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "nemotron/modeling_nemotron.py:_cast_if_autocast_enabled": ["_cast", "_cast_if_autocast_enabled", "amp", "args", "autocast_mode", "def", "device_type", "else", "get_autocast_dtype", "get_autocast_gpu_dtype", "hasattr", "if", "is_autocast_enabled", "not", "return", "target_dtype", "torch"], "nemotron/modeling_nemotron.py:NemotronLayerNorm1P": ["F", "False", "LayerNorm", "ModelLayerNorm1P", "None", "Size", "Tensor", "True", "Union", "__init__", "_cast_if_autocast_enabled", "args", "autocast", "bias", "bool", "class", "cpu", "def", "device", "device_type", "dtype", "elementwise_affine", "else", "enabled", "eps", "float", "forward", "if", "input", "int", "layer_norm", "list", "mps", "nn", "normalized_shape", "return", "self", "super", "torch", "type", "weight", "with"], "nemotron/modeling_nemotron.py:NemotronRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "nemotron/modeling_nemotron.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "nemotron/modeling_nemotron.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "position_ids", "q", "q_embed", "q_pass", "return", "rot_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "nemotron/modeling_nemotron.py:NemotronMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "nemotron/modeling_nemotron.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "nemotron/modeling_nemotron.py:NemotronAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "dropout", "dtype", "during", "errors", "f", "float32", "forward", "functional", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "lead", "logger", "make", "math", "matmul", "max_position_embeddings", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "output_attentions", "p", "partial_rotary_factor", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "recommended", "repeat_kv", "reshape", "return", "rope_theta", "self", "shape", "sin", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "nemotron/modeling_nemotron.py:NemotronFlashAttention2": ["Cache", "False", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "StaticCache", "Tensor", "The", "ValueError", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "an", "and", "apply_rotary_pos_emb", "args", "at", "attention_dropout", "attention_mask", "attn_implementation", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache", "cache_kwargs", "cache_position", "cast", "casted", "class", "com", "compatible", "config", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attention_2", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "github", "hasattr", "have", "head_dim", "hidden", "hidden_states", "https", "huggingface", "if", "implementation", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "isinstance", "issue", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "make", "mean", "might", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "o_proj", "open", "or", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "raise", "related", "reshape", "return", "sdpa", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "static", "super", "sure", "target_dtype", "the", "this", "time", "to", "torch", "training", "transformers", "transpose", "tuple", "type", "upcasted", "update", "use", "use_cache", "use_top_left_mask", "v_proj", "value_states", "version", "view", "warning_once", "weight", "will", "with", "you"], "nemotron/modeling_nemotron.py:NemotronSdpaAttention": ["Cache", "Falling", "False", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "apply_rotary_pos_emb", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_kwargs", "cache_position", "can", "causal_mask", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "scaled_dot_product_attention", "self", "shape", "sin", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "nemotron/modeling_nemotron.py:NemotronDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm1P", "ModelMLP", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "norm_eps", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "nemotron/modeling_nemotron.py:NemotronPreTrainedModel": ["Embedding", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm1P", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "nemotron/modeling_nemotron.py:NemotronModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "BlockMask", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm1P", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "all_hidden_states", "all_self_attns", "and", "any", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "finfo", "flash_attention_2", "flex_attention", "for", "forward", "full", "get_max_cache_shape", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "list", "logger", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "nn", "norm", "norm_eps", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "reshape", "return", "rotary_emb", "sdpa", "self", "sequence_length", "shape", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "type", "unsqueeze", "use_cache", "using_compilable_cache", "vocab_size", "warning_once", "with", "xpu"], "nemotron/modeling_nemotron.py:NemotronForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "nemotron/modeling_nemotron.py:NemotronForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "nemotron/modeling_nemotron.py:NemotronForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "nemotron/modeling_nemotron.py:NemotronForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class"], "lilt/modeling_lilt.py:LiltTextEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelTextEmbeddings", "Module", "None", "__init__", "absolute", "arange", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "incremental_indices", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "mask", "max_position_embeddings", "ne", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "sequence_length", "size", "super", "to", "token_type_embeddings", "token_type_ids", "torch", "type_as", "type_vocab_size", "unsqueeze", "vocab_size", "word_embeddings", "zeros"], "lilt/modeling_lilt.py:LiltLayoutEmbeddings": ["Dropout", "Embedding", "IndexError", "LayerNorm", "Linear", "ModelLayoutEmbeddings", "Module", "None", "The", "__init__", "as", "bbox", "be", "box_linear_embeddings", "box_position_embeddings", "cat", "channel_shrink_ratio", "class", "config", "coordinate", "def", "dim", "dropout", "e", "eps", "except", "forward", "from", "h_position_embeddings", "hidden_dropout_prob", "hidden_size", "in_features", "layer_norm_eps", "left_position_embeddings", "lower_position_embeddings", "max_2d_position_embeddings", "max_position_embeddings", "nn", "out_features", "pad_token_id", "padding_idx", "position_ids", "raise", "range", "return", "right_position_embeddings", "self", "should", "spatial_position_embeddings", "super", "torch", "try", "upper_position_embeddings", "values", "w_position_embeddings", "within", "x_position_embeddings", "y_position_embeddings"], "lilt/modeling_lilt.py:LiltSelfAttention": ["Dropout", "Embedding", "False", "Linear", "ModelSelfAttention", "Module", "None", "Softmax", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bhld", "bhlr", "bhrd", "channel_shrink_ratio", "class", "config", "context_layer", "contiguous", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "layer_idx", "layout_attention_probs", "layout_attention_scores", "layout_context_layer", "layout_inputs", "layout_key", "layout_key_layer", "layout_query", "layout_query_layer", "layout_value", "layout_value_layer", "long", "lrd", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "r", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "seq_length", "size", "sqrt", "super", "the", "tmp_attention_scores", "tmp_layout_attention_scores", "to", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "view", "x"], "lilt/modeling_lilt.py:LiltSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "lilt/modeling_lilt.py:LiltAttention": ["False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "channel_shrink_ratio", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "key", "layer_idx", "layout_attention_output", "layout_inputs", "layout_output", "len", "nn", "num_attention_heads", "ori_hidden_size", "output", "output_attentions", "outputs", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "lilt/modeling_lilt.py:LiltIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "lilt/modeling_lilt.py:LiltOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "lilt/modeling_lilt.py:LiltLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "channel_shrink_ratio", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_size", "hidden_states", "intermediate", "intermediate_output", "intermediate_size", "layer_idx", "layer_output", "layout_attention_output", "layout_feed_forward_chunk", "layout_inputs", "layout_intermediate", "layout_layer_output", "layout_output", "ori_hidden_size", "ori_intermediate_size", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "lilt/modeling_lilt.py:LiltEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "layout_inputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "lilt/modeling_lilt.py:LiltPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "lilt/modeling_lilt.py:LiltPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "lilt/modeling_lilt.py:LiltModel": ["BaseModelOutputWithPooling", "ModelEncoder", "ModelLayoutEmbeddings", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "ModelTextEmbeddings", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bbox", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "layout_embedding_output", "layout_embeddings", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "lilt/modeling_lilt.py:LiltForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "class", "classifier", "config", "def", "device", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "lilt/modeling_lilt.py:LiltForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "class", "classifier", "classifier_dropout", "config", "def", "device", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "lilt/modeling_lilt.py:LiltClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "lilt/modeling_lilt.py:LiltForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bbox", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "zamba/modeling_zamba.py:ZambaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "zamba/modeling_zamba.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "zamba/modeling_zamba.py:ZambaHybridDynamicCache": ["Any", "False", "LongTensor", "ModelHybridDynamicCache", "None", "Optional", "Tensor", "_", "__getitem__", "__init__", "__len__", "_buffers", "_modules", "_parameters", "append", "batch_size", "beam_idx", "cache_kwargs", "cache_shape", "cat", "class", "config", "conv_kernel_size", "conv_states", "def", "device", "dict", "dim", "dtype", "else", "float16", "for", "get_seq_length", "has_previous_state", "hidden_size", "hybrid", "i", "if", "in", "index_select", "int", "intermediate_size", "is_compileable", "key_cache", "key_states", "layer_idx", "layers_block_type", "len", "mamba_d_conv", "mamba_d_state", "mamba_expand", "n_mamba_heads", "not", "num_hidden_layers", "range", "reorder_cache", "return", "self", "shape", "ssm_state_size", "ssm_states", "str", "tensor", "to", "torch", "transformer_layers", "tuple", "update", "value_cache", "value_states", "zeros"], "zamba/modeling_zamba.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "zamba/modeling_zamba.py:ZambaAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "ModelHybridDynamicCache", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "attention_dropout", "attention_head_dim", "attention_hidden_size", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "class", "config", "contiguous", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "max_position_embeddings", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "zamba/modeling_zamba.py:ZambaMambaMixer": ["A", "ACT2FN", "AILab", "A_log", "B", "C", "CUDA", "Conv1d", "D", "Dao", "False", "Fast", "If", "Linear", "Make", "Mamba", "ModelConfig", "ModelHybridDynamicCache", "ModelMambaMixer", "Module", "None", "Parameter", "Tensor", "The", "To", "True", "ValueError", "_", "__init__", "a", "act", "activation", "all", "and", "append", "arange", "are", "attention_mask", "available", "batch_size", "because", "bias", "cache_params", "cat", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk", "class", "clone", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_kernel_size", "conv_state", "conv_states", "conv_weights", "copy_", "cuda", "cuda_kernels_forward", "def", "deltaB_u", "delta_softplus", "device", "dim", "dims", "discrete_A", "discrete_B", "discrete_time_step", "dt_proj_bias", "dt_proj_weight", "dt_softplus", "dtype", "else", "empty", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "has_previous_state", "hidden_mamba_act", "hidden_size", "hidden_states", "https", "i", "if", "implementation", "in", "in_channels", "in_proj", "input_states", "install", "installed", "intermediate_size", "is", "is_fast_path_available", "isinstance", "kernel_size", "kernels", "layer_idx", "lease", "log", "logger", "mamba", "mamba_conv_bias", "mamba_d_conv", "mamba_d_state", "mamba_dt_rank", "mamba_expand", "mamba_head_dim", "mamba_inner_fn", "mamba_proj_bias", "matmul", "model", "module", "n", "n_mamba_heads", "naive", "nn", "not", "of", "on", "one", "ones", "or", "out_channels", "out_proj", "pad", "padding", "path", "pip", "projected_states", "raise", "range", "reshape", "return", "return_last_state", "roll", "run", "s", "scan_output", "scan_outputs", "scan_outputs_", "selective_scan_fn", "selective_state_update", "self", "seq_len", "set", "shape", "shifts", "size", "slow_forward", "softplus", "spaces", "split", "squeeze", "ssm", "ssm_parameters", "ssm_state", "ssm_state_", "ssm_state_size", "ssm_states", "stack", "state", "sum", "super", "sure", "that", "the", "they", "time_proj_bias", "time_step", "time_step_rank", "to", "torch", "training", "transpose", "type", "unsqueeze", "use", "use_bias", "use_cache", "use_conv_bias", "use_fast_kernels", "use_mamba_kernels", "use_precomputed_states", "view", "want", "warning_once", "weight", "x_proj_weight", "you", "zeros"], "zamba/modeling_zamba.py:ZambaMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "zamba/modeling_zamba.py:ZambaAttentionDecoderLayer": ["False", "FlashAttentionKwargs", "FloatTensor", "ModelAttention", "ModelAttentionDecoderLayer", "ModelConfig", "ModelHybridDynamicCache", "ModelMLP", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "Unpack", "__init__", "attention_hidden_size", "attention_mask", "bool", "class", "concatenate", "config", "def", "deprecate_kwarg", "dim", "eps", "feed_forward", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "new_name", "nn", "original_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "pre_ff_layernorm", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "zamba/modeling_zamba.py:ZambaMambaDecoderLayer": ["False", "FloatTensor", "LongTensor", "ModelConfig", "ModelHybridDynamicCache", "ModelMambaDecoderLayer", "ModelMambaMixer", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_params", "cache_position", "causal_mask", "class", "config", "def", "deprecate_kwarg", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "is", "kwargs", "layer_idx", "mamba", "new_name", "nn", "not", "original_hidden_states", "output_attentions", "outputs", "past_key_value", "past_key_values", "residual", "return", "rms_norm_eps", "self", "self_attn_weights", "super", "torch", "transformer_hidden_states", "tuple", "use_cache", "version"], "zamba/modeling_zamba.py:ZambaHybridLayer": ["False", "FloatTensor", "Linear", "LongTensor", "ModelAttentionDecoderLayer", "ModelHybridDynamicCache", "ModelHybridLayer", "ModelMambaDecoderLayer", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_position", "causal_mask", "class", "def", "deprecate_kwarg", "forward", "hidden_states", "if", "int", "layer_idx", "layer_outputs", "linear", "mamba", "mamba_decoder", "new_name", "nn", "original_hidden_states", "output_attentions", "past_key_value", "past_key_values", "return", "self", "self_attn_weights", "shared_transf", "super", "torch", "transformer_hidden_states", "tuple", "use_cache", "version"], "zamba/modeling_zamba.py:ZambaPreTrainedModel": ["A", "A_log", "Conv1d", "D", "Embedding", "False", "Linear", "ModelAttentionDecoderLayer", "ModelConfig", "ModelMambaDecoderLayer", "ModelMambaMixer", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "arange", "base_model_prefix", "bias", "clamp", "class", "config", "contiguous", "copy_", "data", "def", "dt", "dt_init_std", "dt_proj_bias", "dt_proj_weight", "dtype", "elif", "exp", "expand", "expm1", "fill_", "float32", "hidden_size", "if", "init", "initializer_range", "intermediate_size", "inv_dt", "is", "isinstance", "log", "mamba_dt_rank", "mamba_expand", "mamba_head_dim", "math", "mean", "min", "model", "module", "n_mamba_heads", "nn", "normal_", "not", "padding_idx", "past_key_values", "rand", "reshape", "self", "ssm_state_size", "std", "supports_gradient_checkpointing", "time_step_floor", "time_step_max", "time_step_min", "torch", "uniform_", "weight", "x_proj_weight", "zero_"], "zamba/modeling_zamba.py:ZambaModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "Embedding", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelAttentionDecoderLayer", "ModelConfig", "ModelHybridDynamicCache", "ModelHybridLayer", "ModelMambaDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "__call__", "__init__", "_attn_implementation", "_gradient_checkpointing_func", "_tied_weights_keys", "_unmask_unattended", "_update_causal_mask", "a", "all_hidden_states", "all_self_attns", "an", "and", "append", "arange", "at", "attention_mask", "attentions", "auto_docstring", "be", "bias", "block", "bool", "both", "cache", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "down_proj", "dtype", "either", "elif", "else", "embed_tokens", "enumerate", "eps", "eq", "expand", "f", "feed_forward", "fill_value", "final_layernorm", "finfo", "flash_attention_2", "for", "forward", "full", "gate_proj", "gradient", "gradient_checkpointing", "has_previous_state", "hidden_size", "hidden_states", "hybrid", "i", "if", "in", "incompatible", "initialized", "input_ids", "input_layernorm", "input_tensor", "inputs_embeds", "is", "iter", "k_proj", "key", "last_hidden_state", "layer", "layer_id", "layer_idx", "layer_outputs", "layer_type", "layers", "layers_block_type", "linear_layers", "logger", "mamba", "mamba_layers", "mask_length", "masked_fill", "min", "min_dtype", "must", "next", "nn", "no", "not", "npu", "num_hidden_layers", "o_proj", "one", "original_hidden_states", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "position_ids", "post_init", "pre_ff_layernorm", "prefix_name", "provided", "q_proj", "raise", "range", "requires", "reshape", "return", "return_dict", "returned", "rms_norm_eps", "same", "sdpa", "self", "self_attn", "sequence_length", "shape", "shared_transf", "so", "specify", "super", "target_length", "the", "tied_keys", "time", "to", "to_tuple", "torch", "training", "triu", "tuple", "type", "unsqueeze", "up_proj", "use_cache", "use_return_dict", "v_proj", "vocab_size", "warning_once", "was", "weight", "will", "with", "xpu"], "zamba/modeling_zamba.py:ZambaForCausalLM": ["CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelHybridDynamicCache", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "contiguous", "cumsum", "def", "device", "dtype", "elif", "else", "empty_past_kv", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "key", "kwargs", "labels", "lm_head", "logits", "logits_to_keep", "long", "loss", "loss_function", "masked_fill_", "model", "model_inputs", "nn", "not", "num_logits_to_keep", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "shape", "slice", "slice_indices", "super", "torch", "tuple", "update", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "zamba/modeling_zamba.py:ZambaForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bias", "bool", "class", "config", "conjunction", "def", "defined", "detect", "device", "dtype", "elif", "else", "f", "forward", "handle", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "int32", "is", "labels", "last_non_pad_token", "list", "logger", "logits", "long", "loss", "loss_fct", "may", "model", "multi_label_classification", "nn", "no", "non_pad_mask", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "token_indices", "tokens", "torch", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "view", "warning_once", "will", "with"], "whisper/modeling_whisper.py:sinusoids": ["Model", "Number", "Tensor", "ValueError", "arange", "be", "by", "cat", "channels", "cos", "def", "dim", "divisible", "embeddings", "exp", "f", "float", "for", "got", "has", "if", "int", "inv_timescales", "length", "log", "log_timescale_increment", "math", "max_timescale", "of", "positional", "raise", "return", "scaled_time", "sin", "sinusoidal", "to", "torch", "view"], "whisper/modeling_whisper.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "whisper/modeling_whisper.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "whisper/modeling_whisper.py:WhisperPositionalEmbedding": ["Embedding", "ModelPositionalEmbedding", "None", "Optional", "__init__", "class", "def", "else", "embedding_dim", "forward", "if", "input_ids", "int", "is", "nn", "num_positions", "padding_idx", "past_key_values_length", "position_ids", "return", "self", "shape", "super", "weight"], "whisper/modeling_whisper.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "and", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "ndim", "nn", "not", "p", "query", "return", "scaling", "shape", "size", "softmax", "torch", "training", "transpose", "value", "view"], "whisper/modeling_whisper.py:WhisperAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kwargs", "layer_head_mask", "layer_idx", "layers", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "whisper/modeling_whisper.py:WhisperEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "whisper/modeling_whisper.py:WhisperDecoderLayer": ["ACT2FN", "EncoderDecoderCache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "whisper/modeling_whisper.py:WhisperPreTrainedModel": ["Conv1d", "Embedding", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelEncoder", "ModelEncoderLayer", "ModelForAudioClassification", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_get_feat_extract_output_lengths", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "copy_", "data", "def", "elif", "embed_positions", "fill_", "if", "init_std", "input_features", "input_lengths", "is", "isinstance", "layer_weights", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "num_hidden_layers", "padding_idx", "return", "self", "shape", "sinusoids", "std", "supports_gradient_checkpointing", "torch", "use_weighted_layer_sum", "weight", "zero_"], "whisper/modeling_whisper.py:WhisperEncoder": ["BaseModelOutput", "Conv1d", "Embedding", "False", "LayerNorm", "Make", "Model", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "Module", "ModuleList", "None", "The", "True", "ValueError", "_", "__init__", "_freeze_parameters", "_requires_grad", "all_attentions", "all_positions", "arange", "assert", "attention_mask", "attentions", "be", "but", "class", "config", "conv1", "conv2", "d_model", "def", "device", "dropout", "dropout_probability", "else", "embed_dim", "embed_positions", "embed_scale", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "expected_seq_length", "expects", "f", "features", "for", "forward", "found", "functional", "gelu", "get_input_embeddings", "gradient_checkpointing", "head_mask", "hidden_states", "idx", "if", "in", "input", "input_features", "inputs_embeds", "is", "it", "kernel_size", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "length", "math", "max_source_positions", "mel", "nn", "not", "num_embeddings", "num_mel_bins", "of", "output_attentions", "output_hidden_states", "p", "pad", "pad_token_id", "padding", "padding_idx", "param", "parameters", "permute", "post_init", "r", "raise", "rand", "range", "requires_grad", "requires_grad_", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shape", "should", "size", "specified", "sqrt", "stride", "super", "sure", "the", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "value"], "whisper/modeling_whisper.py:WhisperDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "LayerNorm", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPositionalEmbedding", "ModelPreTrainedModel", "ModuleList", "None", "Setting", "The", "True", "ValueError", "You", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "arange", "assert", "at", "attention_mask", "attentions", "attn_mask", "be", "both", "but", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "create_causal_mask", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "device", "dropout", "dropout_probability", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_hidden_states", "enumerate", "f", "for", "forward", "functional", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "incompatible", "input_embeds", "input_ids", "input_shape", "inputs_embeds", "is", "is_encoder_decoder", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "logger", "main_input_name", "mask_name", "math", "max_source_positions", "max_target_positions", "next_cache", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "past_key_values", "past_key_values_length", "position_ids", "positions", "post_init", "r", "raise", "rand", "range", "repeat", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warning_once", "with", "zip"], "whisper/modeling_whisper.py:WhisperModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_freeze_parameters", "_mask_input_features", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_position_ids", "def", "device", "dtype", "elif", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "expand", "forward", "freeze_encoder", "get_encoder", "get_input_embeddings", "getattr", "head_mask", "hidden_size", "hidden_states", "if", "input_features", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "min_masks", "not", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_length", "set_input_embeddings", "size", "super", "tensor", "torch", "training", "tuple", "use_cache", "use_return_dict", "value"], "whisper/modeling_whisper.py:WhisperForConditionalGeneration": ["Cache", "CrossEntropyLoss", "False", "FloatTensor", "Labels", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelGenerationMixin", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "Union", "ValueError", "__init__", "_freeze_parameters", "_tied_weights_keys", "allowed", "and", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cannot", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_position_ids", "decoder_start_token_id", "def", "device", "else", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "exceed", "f", "forward", "freeze_encoder", "get_decoder", "get_encoder", "get_input_embeddings", "get_output_embeddings", "head_mask", "if", "input_features", "is", "labels", "length", "lm_logits", "logits", "loss", "loss_fct", "max_target_positions", "maximum", "model", "new_embeddings", "nn", "not", "of", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "post_init", "proj_out", "r", "raise", "reshape", "return", "return_dict", "self", "sequence", "set_output_embeddings", "shape", "shift_tokens_right", "super", "the", "to", "tokens", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "whisper/modeling_whisper.py:WhisperDecoderWrapper": ["False", "ModelDecoder", "ModelDecoderWrapper", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "embed_tokens", "forward", "get_input_embeddings", "is_encoder_decoder", "kwargs", "return", "self", "set_input_embeddings", "super", "value"], "whisper/modeling_whisper.py:WhisperForCausalLM": ["BaseModelOutput", "Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelDecoderWrapper", "ModelForCausalLM", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "encoder_hidden_states", "encoder_outputs", "forward", "get_decoder", "get_input_embeddings", "get_output_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_encoder_decoder", "isinstance", "labels", "list", "logits", "loss", "loss_fct", "main_input_name", "model", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "proj_out", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "set_output_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "whisper/modeling_whisper.py:WhisperForAudioClassification": ["CrossEntropyLoss", "FloatTensor", "Linear", "LongTensor", "ModelEncoder", "ModelForAudioClassification", "ModelPreTrainedModel", "Module", "None", "Optional", "Parameter", "SequenceClassifierOutput", "Tensor", "True", "Union", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "attentions", "auto_docstring", "bool", "class", "classifier", "classifier_proj_size", "config", "def", "device", "dim", "elif", "else", "encoder", "encoder_outputs", "forward", "freeze_encoder", "functional", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_features", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "ones", "output", "output_attentions", "output_hidden_states", "pooled_output", "post_init", "projector", "r", "return", "return_dict", "self", "set_input_embeddings", "softmax", "stack", "sum", "super", "to", "torch", "tuple", "use_return_dict", "use_weighted_layer_sum", "value", "view"], "granite_speech/modeling_granite_speech.py:GraniteSpeechCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "granite_speech/modeling_granite_speech.py:GraniteSpeechEncoderProjector": ["AutoModel", "Linear", "ModelConfig", "ModelEncoderProjector", "Module", "None", "Parameter", "Tensor", "True", "__init__", "batch_size", "ceil", "class", "config", "constant", "data", "def", "dim", "downsample_rate", "encoder_attention_mask", "encoder_hidden_states", "forward", "from_config", "functional", "hidden_size", "hidden_states", "last_hidden_state", "linear", "math", "mean", "nblocks", "nn", "normal_", "num_queries", "pad", "projector_config", "qformer", "query", "query_embeds", "query_output", "query_proj", "return", "return_dict", "self", "seq_len", "size", "std", "super", "text_config", "torch", "view", "window_size", "zeros"], "granite_speech/modeling_granite_speech.py:GraniteSpeechConformerFeedForward": ["Dropout", "LayerNorm", "Linear", "ModelConformerFeedForward", "ModelEncoderConfig", "Module", "SiLU", "Tensor", "__init__", "class", "config", "def", "down_proj", "dropout", "feedforward_mult", "forward", "hidden_dim", "hidden_states", "nn", "pre_norm", "return", "self", "silu", "super", "torch", "up_proj"], "granite_speech/modeling_granite_speech.py:GraniteSpeechConformerAttention": ["Context", "Dropout", "Embedding", "F", "False", "LayerNorm", "Linear", "MATH", "ModelConformerAttention", "ModelEncoderConfig", "Module", "SDPBackend", "Tensor", "ValueError", "_", "__init__", "attention", "attention_dists", "attn_mask", "b", "bias", "bool", "bsz", "c", "ceil", "chunk", "class", "config", "context_size", "d", "def", "device", "dim", "dim_head", "dropout", "dtype", "einsum", "either", "exceeds", "finfo", "forward", "functional", "h", "hidden_dim", "hidden_states", "if", "inner_dim", "is", "key_states", "less", "m", "mask", "mask_value", "masked_fill_", "math", "max", "max_pos_emb", "nn", "num_blocks", "num_features", "num_heads", "ones", "or", "out", "pad", "pos_attn", "pre_norm", "query_states", "r", "raise", "rel_pos_emb", "remainder", "reshape", "return", "scale", "scaled_dot_product_attention", "sdpa_kernel", "self", "shape", "size", "super", "than", "the", "to_kv", "to_out", "to_q", "torch", "transpose", "value_states", "with"], "granite_speech/modeling_granite_speech.py:GraniteSpeechConformerDepthWiseConv1d": ["Conv1d", "F", "False", "ModelConformerDepthWiseConv1d", "Module", "Tensor", "__init__", "bias", "chan_in", "chan_out", "class", "conv", "def", "forward", "groups", "hidden_states", "int", "kernel_size", "nn", "pad", "pad_offset", "padding", "return", "self", "super", "torch"], "granite_speech/modeling_granite_speech.py:GraniteSpeechConformerConvModule": ["BatchNorm1d", "Conv1d", "Dropout", "GLU", "LayerNorm", "ModelConformerConvModule", "ModelConformerDepthWiseConv1d", "ModelEncoderConfig", "Module", "SiLU", "Tensor", "__init__", "batch_norm", "class", "config", "conv_expansion_factor", "conv_kernel_size", "def", "depth_conv", "dim", "down_conv", "dropout", "forward", "glu", "hidden_dim", "hidden_states", "inner_dim", "kernel_size", "nn", "norm", "permute", "return", "self", "silu", "super", "torch", "up_conv"], "granite_speech/modeling_granite_speech.py:GraniteSpeechConformerBlock": ["LayerNorm", "ModelConformerAttention", "ModelConformerBlock", "ModelConformerConvModule", "ModelConformerFeedForward", "ModelEncoderConfig", "Module", "Tensor", "__init__", "attention_dists", "attn", "class", "config", "conv", "def", "ff1", "ff2", "forward", "hidden_dim", "hidden_states", "nn", "post_norm", "return", "self", "super", "torch"], "granite_speech/modeling_granite_speech.py:GraniteSpeechCTCEncoder": ["False", "Linear", "ModelCTCEncoder", "ModelConformerBlock", "ModelEncoderConfig", "Module", "ModuleList", "Softmax", "Tensor", "True", "_", "__init__", "arange", "attention_dists", "bias", "clamp", "class", "clone", "config", "context_size", "def", "dim", "enumerate", "for", "forward", "hidden_dim", "hidden_states", "hidden_states_mid", "idx", "if", "in", "input_dim", "input_linear", "layer", "layers", "max_pos_emb", "nn", "num_layers", "out", "out_mid", "output_dim", "persistent", "range", "register_buffer", "relpos_dist", "return", "self", "seq", "start", "super", "torch", "view"], "granite_speech/modeling_granite_speech.py:GraniteSpeechPreTrainedModel": ["BatchNorm1d", "Conv1d", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderProjector", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_supports_flash_attn", "_supports_sdpa", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "query", "self", "std", "weight", "zero_"], "granite_speech/modeling_granite_speech.py:GraniteSpeechForConditionalGeneration": ["AutoModelForCausalLM", "Cache", "Config", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "LongTensor", "ModelCTCEncoder", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelEncoderProjector", "ModelForConditionalGeneration", "ModelPreTrainedModel", "None", "Number", "Optional", "Please", "Tensor", "Union", "ValueError", "You", "__init__", "_fix_state_dict_key_on_save", "_fix_state_dict_keys_on_save", "_get_adapter_name", "_hf_peft_config_loaded", "_tied_weights_keys", "a", "adapter", "adapter_name", "all", "and", "are", "args", "at", "attention_mask", "attentions", "audio", "audio_embeds", "audio_features", "audio_token_id", "auto_docstring", "base_layer", "base_model_name_or_path", "be", "bool", "both", "but", "cache_position", "cannot", "cause", "class", "clone", "config", "contiguous", "def", "device", "dim", "disable_adapters", "does", "dtype", "either", "else", "enable_adapters", "encoder", "encoder_config", "encoder_embeds", "exactly", "f", "features", "for", "forward", "from_config", "generate", "get_audio_features", "get_input_embeddings", "get_merged_audio_embeddings", "get_output_embeddings", "has_lora_adapter", "hidden_states", "if", "in", "incorrectly", "indicates", "input_features", "input_features_mask", "input_ids", "inputs", "inputs_embeds", "install", "installed", "int", "is", "is_audio_idx", "is_audio_index", "is_peft_available", "item", "items", "k", "key", "keys", "kwargs", "labels", "language_model", "list", "llm_input_ids", "lm_kwargs", "logger", "logits", "logits_to_keep", "lora", "lora_", "loss", "loss_fct", "masked_scatter", "match", "model", "model_inputs", "must", "new_embeddings", "nn", "not", "number", "of", "one", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "peft", "peft_config", "perform", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "present", "prev_val", "projected_embeds", "projector", "provided", "r", "raise", "reload", "replace", "return", "return_dict", "same", "save_directory", "save_pretrained", "self", "set_input_embeddings", "set_output_embeddings", "shape", "shift_attention_mask", "shift_labels", "shift_logits", "should", "size", "special_audio_mask", "specify", "state_dict", "staticmethod", "str", "sum", "super", "text_config", "that", "the", "this", "time", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "view", "warning", "when", "where", "will"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "None", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "else", "forward", "gate_proj", "hidden_act", "hidden_size", "if", "intermediate_size", "is", "nn", "return", "self", "super", "up_proj", "x"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3TopkRouter": ["F", "False", "ModelTopkRouter", "Module", "Parameter", "True", "__init__", "bool", "class", "config", "def", "denominator", "dim", "e_score_correction_bias", "empty", "expand", "float32", "forward", "gather", "get_topk_indices", "group_idx", "group_mask", "group_scores", "hidden_size", "hidden_states", "if", "k", "keepdim", "linear", "masked_fill", "n_group", "n_routed_experts", "nn", "no_grad", "norm_topk_prob", "num_experts_per_tok", "register_buffer", "reshape", "return", "routed_scaling_factor", "router_logits", "scatter_", "score_mask", "scores", "scores_for_choice", "self", "sigmoid", "sorted", "sum", "super", "top_k", "topk", "topk_group", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight", "zeros", "zeros_like"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3MoE": ["ModelMLP", "ModelMoE", "ModelTopkRouter", "Module", "ModuleList", "Tensor", "_", "__init__", "class", "config", "def", "dtype", "expert", "expert_idx", "expert_input", "expert_mask", "expert_output", "expert_weights", "experts", "final_hidden_states", "for", "forward", "functional", "gate", "hidden_states", "if", "in", "index_add_", "intermediate_size", "len", "mask", "moe", "moe_intermediate_size", "n_routed_experts", "n_shared_experts", "nn", "num_classes", "numel", "one_hot", "orig_shape", "permute", "r", "range", "residuals", "return", "self", "shape", "shared_experts", "super", "token_indices", "topk_indices", "topk_weights", "torch", "type", "unsqueeze", "view", "weight_indices", "weighted_output", "where", "zeros_like"], "deepseek_v3/modeling_deepseek_v3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "deepseek_v3/modeling_deepseek_v3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "deepseek_v3/modeling_deepseek_v3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "deepseek_v3/modeling_deepseek_v3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "deepseek_v3/modeling_deepseek_v3.py:apply_rotary_pos_emb_interleave": ["Model_rotary_pos_emb_interleave", "None", "b", "cos", "d", "def", "h", "k", "k_embed", "position_ids", "q", "q_embed", "r", "reshape", "return", "rotate_half", "s", "shape", "sin", "transpose", "unsqueeze", "unsqueeze_dim", "view"], "deepseek_v3/modeling_deepseek_v3.py:yarn_get_mscale": ["Model_get_mscale", "def", "if", "log", "math", "mscale", "return", "scale"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "F", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "apply_rotary_pos_emb", "apply_rotary_pos_emb_interleave", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "bias", "cache_kwargs", "cache_position", "cat", "class", "compressed_kv", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dim", "dropout", "eager", "eager_attention_forward", "else", "expand", "factor", "flash_attention_2", "forward", "get", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_pass", "k_rot", "key_shape", "key_states", "kv_a_layernorm", "kv_a_proj_with_mqa", "kv_b_proj", "kv_lora_rank", "kwargs", "layer_idx", "mscale", "mscale_all_dim", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "pad", "past_key_value", "past_key_values", "position_embeddings", "q_a_layernorm", "q_a_proj", "q_b_proj", "q_lora_rank", "q_pass", "q_proj", "q_rot", "q_states", "qk_head_dim", "qk_nope_head_dim", "qk_rope_head_dim", "query_shape", "query_states", "reshape", "return", "rope_interleave", "rope_scaling", "rope_theta", "scaling", "scaling_factor", "self", "seq_length", "shape", "sin", "split", "super", "torch", "training", "transpose", "tuple", "update", "v_head_dim", "value_states", "version", "view", "yarn_get_mscale"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelMoE", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "first_k_dense_replace", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3PreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelTopkRouter", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "initializer_range", "isinstance", "mean", "model", "module", "normal_", "past_key_values", "self", "std", "super", "supports_gradient_checkpointing", "weight"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_keys_to_ignore_on_load_unexpected", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "model", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "deepseek_v3/modeling_deepseek_v3.py:DeepseekV3ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "rwkv/modeling_rwkv.py:load_wkv_cuda_kernel": ["CUDA", "DEBUG", "DTmax", "Model", "Model_cuda_kernel", "Model_kernel", "Model_wkv_cuda_kernel", "Modeling", "None", "O3", "Path", "Xptxas", "__file__", "and", "at", "context", "context_length", "cpp", "cu", "cuda_kernel_files", "def", "device", "extra", "extra_cuda_cflags", "f", "flags", "for", "get_verbosity", "global", "if", "in", "info", "is", "kernel", "kernel_folder", "kernels", "length", "logger", "logging", "max_seq_length", "maxrregcount", "name", "not", "of", "parent", "res", "resolve", "return", "sources", "usage", "use_fast_math", "vectorization", "verbose", "wkv_", "wkv_cuda", "wkv_cuda_bf16", "wkv_op"], "rwkv/modeling_rwkv.py:RwkvLinearAttention": ["CUDA", "Calling", "Cannot", "False", "Function", "ModelLinearAttention", "Model_cuda_kernel", "None", "The", "ValueError", "a", "all", "and", "at", "attention", "autograd", "backward", "backward_bf16", "backward_func", "batch", "batch_size", "be", "bfloat16", "cat", "chunk", "class", "contiguous", "contiguous_format", "ctx", "cuda", "def", "device", "devices", "dim", "dtype", "else", "empty_like", "exp", "f", "float", "float16", "float32", "for", "forward", "forward_bf16", "forward_func", "forward_with_state", "forward_with_state_bf16", "g_key", "g_output", "g_state", "g_time_decay", "g_time_first", "g_value", "hidden", "hidden_size", "if", "in", "input_dtype", "is", "kernel", "key", "max_seq_length", "maximum", "memory_format", "min", "model", "multiple", "needs", "not", "of", "on", "or", "output", "process", "product", "raise", "requires", "return", "return_state", "round", "s", "same", "save_for_backward", "saved_tensors", "seq_len", "size", "squeeze", "state", "staticmethod", "tensors", "the", "this", "time", "time_decay", "time_first", "to", "tokens", "torch", "type", "unsqueeze", "use", "value", "with", "wkv", "zeros"], "rwkv/modeling_rwkv.py:rwkv_linear_attention_cpu": ["False", "Model_linear_attention_cpu", "None", "_", "current_index", "current_key", "current_value", "def", "den_state", "denominator", "dtype", "e1", "e2", "else", "exp", "float", "float32", "for", "if", "in", "is", "key", "max_for_output", "max_for_state", "max_state", "maximum", "not", "num_state", "numerator", "or", "output", "range", "return", "return_state", "seq_length", "size", "state", "time_decay", "time_first", "to", "torch", "value", "zeros_like"], "rwkv/modeling_rwkv.py:rwkv_linear_attention": ["False", "ModelLinearAttention", "Model_cuda_kernel", "Model_linear_attention", "Model_linear_attention_cpu", "None", "any", "apply", "cuda", "def", "device", "else", "for", "if", "in", "is", "key", "no_cuda", "one_token", "or", "return", "return_state", "size", "state", "t", "time_decay", "time_first", "type", "value"], "rwkv/modeling_rwkv.py:RwkvSelfAttention": ["CUDA", "Could", "Exception", "False", "Linear", "Model", "ModelSelfAttention", "Model_cuda_kernel", "Model_linear_attention", "Module", "None", "Parameter", "ZeroPad2d", "__init__", "and", "attention", "attention_hidden_size", "bias", "class", "config", "context_length", "custom", "def", "else", "empty", "except", "extract_key_value", "for", "forward", "hidden", "hidden_size", "if", "in", "info", "is", "is_ninja_available", "is_torch_cuda_available", "kernel", "kernel_loaded", "key", "layer_id", "layer_state", "load", "load_wkv_cuda_kernel", "logger", "max_seq_length", "nn", "not", "output", "receptance", "return", "return_state", "s", "self", "shifted", "sigmoid", "size", "state", "super", "the", "time_decay", "time_first", "time_mix_key", "time_mix_receptance", "time_mix_value", "time_shift", "torch", "try", "tuple", "use_cache", "value"], "rwkv/modeling_rwkv.py:RwkvFeedForward": ["False", "Linear", "ModelFeedForward", "Module", "None", "Parameter", "ZeroPad2d", "__init__", "and", "bias", "class", "config", "def", "else", "empty", "forward", "hidden", "hidden_size", "if", "intermediate_size", "is", "key", "layer_id", "nn", "not", "receptance", "relu", "return", "self", "shifted", "sigmoid", "size", "square", "state", "super", "time_mix_key", "time_mix_receptance", "time_shift", "torch", "value"], "rwkv/modeling_rwkv.py:RwkvBlock": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelBlock", "ModelFeedForward", "ModelSelfAttention", "None", "__init__", "attention", "class", "config", "def", "else", "eps", "feed_forward", "forward", "hidden", "hidden_size", "if", "layer_id", "layer_norm_epsilon", "ln1", "ln2", "nn", "output_attentions", "outputs", "pre_ln", "return", "self", "state", "super", "use_cache"], "rwkv/modeling_rwkv.py:RwkvPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelBlock", "ModelConfig", "ModelFeedForward", "ModelPreTrainedModel", "ModelSelfAttention", "Module", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_keep_in_fp32_modules", "_no_split_modules", "and", "attention_hidden_size", "base_model_prefix", "bias", "class", "config", "data", "decay_speed", "def", "device", "dtype", "elif", "fill_", "for", "gain", "h", "hidden_size", "i", "if", "in", "init", "is", "isinstance", "layer_id", "log", "math", "max", "module", "nn", "not", "num_hidden_layers", "ones_like", "orthogonal_", "pow", "range", "ratio_0_to_1", "ratio_1_to_almost0", "scale", "self", "shape", "sqrt", "supports_gradient_checkpointing", "tensor", "time_decay", "time_first", "time_mix_key", "time_mix_receptance", "time_mix_value", "time_weight", "torch", "vocab_size", "weight", "zero_", "zigzag"], "rwkv/modeling_rwkv.py:RwkvOutput": ["FloatTensor", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "list", "r", "state", "torch", "tuple"], "rwkv/modeling_rwkv.py:RwkvCausalLMOutput": ["FloatTensor", "ModelCausalLMOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "r", "state", "torch", "tuple"], "rwkv/modeling_rwkv.py:RwkvModel": ["Embedding", "False", "FloatTensor", "ImportError", "LayerNorm", "LongTensor", "ModelBlock", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Params4bit", "Please", "SCB", "Setting", "True", "Union", "ValueError", "You", "__init__", "_bnb_4bit_dequantize_and_rescale", "_rescale_layers", "all_hidden_states", "all_self_attentions", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "bitsandbytes", "block", "block_id", "blocks", "bnb", "bool", "both", "but", "cannot", "checkpointing", "class", "config", "cpu", "data", "def", "dequant_weights", "dequantize_4bit", "device", "div_", "dtype", "either", "elif", "else", "embeddings", "enumerate", "feed_forward", "float32", "for", "forward", "functional", "get_input_embeddings", "gradient", "gradient_checkpointing", "hasattr", "have", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input_ids", "inputs_embeds", "install", "int", "is", "is_bitsandbytes_available", "it", "last_hidden_state", "layer_id", "layers_are_rescaled", "list", "ln_out", "logger", "method", "model", "mul_", "new_embeddings", "nn", "no_grad", "not", "num_hidden_layers", "or", "output", "output_attentions", "output_hidden_states", "passed", "post_init", "quant_state", "quant_weight", "r", "raise", "range", "requires_grad", "rescale_every", "return", "return_dict", "same", "self", "set_input_embeddings", "setattr", "shape", "size", "specify", "state", "super", "target_layer", "the", "this", "time", "to", "torch", "training", "tuple", "unused", "use", "use_cache", "use_return_dict", "value", "vocab_size", "warning_once", "was", "weight", "with", "x", "zeros"], "rwkv/modeling_rwkv.py:RwkvForCausalLM": ["False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelCausalLMOutput", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "Model_outputs", "None", "Optional", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "def", "else", "for", "forward", "get_output_embeddings", "head", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "items", "key", "kwargs", "labels", "list", "logits", "loss", "loss_function", "model_inputs", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "self", "set_output_embeddings", "state", "super", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "bamba/modeling_bamba.py:BambaFlashAttentionKwargs": ["False", "IntTensor", "LongTensor", "ModelFlashAttentionKwargs", "TypedDict", "class", "cu_seq_lens_k", "cu_seq_lens_q", "int", "max_length_k", "max_length_q", "seq_idx", "torch", "total"], "bamba/modeling_bamba.py:HybridMambaAttentionDynamicCache": ["Any", "False", "LongTensor", "ModelConfig", "ModelMambaAttentionDynamicCache", "None", "Optional", "Tensor", "_", "__init__", "append", "batch_size", "beam_idx", "cache_kwargs", "cat", "class", "config", "conv_kernel_size", "conv_states", "def", "device", "dict", "dim", "dtype", "else", "float16", "for", "get_seq_length", "has_previous_state", "hidden_size", "i", "if", "in", "index_select", "int", "is_compileable", "key_cache", "key_states", "layer_idx", "layers_block_type", "len", "mamba", "mamba_d_conv", "mamba_d_head", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "not", "num_hidden_layers", "range", "reorder_cache", "return", "self", "shape", "ssm_state_size", "ssm_states", "str", "tensor", "to", "torch", "transformer_layers", "tuple", "update", "value_cache", "value_states", "zeros"], "bamba/modeling_bamba.py:BambaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "bamba/modeling_bamba.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "bamba/modeling_bamba.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "bamba/modeling_bamba.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "bamba/modeling_bamba.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cat", "cos", "def", "dim", "k", "k_embed", "k_pass", "k_rot", "position_ids", "q", "q_embed", "q_pass", "q_rot", "return", "rotary_dim", "rotate_half", "shape", "sin", "torch", "unsqueeze", "unsqueeze_dim"], "bamba/modeling_bamba.py:BambaAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "bamba/modeling_bamba.py:BambaRMSNormGated": ["ModelRMSNormGated", "Module", "None", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "float32", "forward", "functional", "gate", "hidden_size", "hidden_states", "if", "input_dtype", "is", "keepdim", "mean", "nn", "not", "ones", "pow", "return", "rsqrt", "self", "silu", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "bamba/modeling_bamba.py:pad_tensor_by_size": ["Model", "Model_shape", "Model_size", "Model_tensor_by_size", "Tensor", "constant", "def", "else", "functional", "if", "input_tensor", "int", "len", "mode", "nn", "return", "shape", "torch", "value"], "bamba/modeling_bamba.py:reshape_into_chunks": ["Model", "Model_into_chunks", "chunk_size", "def", "else", "if", "input_tensor", "len", "pad_size", "pad_tensor_by_size", "return", "shape"], "bamba/modeling_bamba.py:segment_sum": ["Model_sum", "None", "bool", "chunk_size", "cumsum", "def", "device", "diagonal", "dim", "dtype", "expand", "inf", "input_tensor", "mask", "masked_fill", "ones", "return", "size", "tensor_segsum", "torch", "tril"], "bamba/modeling_bamba.py:apply_mask_to_padding_states": ["Model_mask_to_padding_states", "None", "and", "attention_mask", "def", "dtype", "hidden_states", "if", "is", "not", "return", "shape", "to"], "bamba/modeling_bamba.py:BambaMixer": ["A", "ACT2FN", "AILab", "A_cumsum", "A_log", "B", "B_decay", "C", "C_reshaped", "C_times_states", "Conv1d", "D", "D_residual", "Dao", "Falling", "False", "G", "GPU", "G_intermediate", "HybridMambaAttentionDynamicCache", "IntTensor", "L", "Linear", "LongTensor", "M", "M_intermediate", "Model", "ModelConfig", "ModelMixer", "ModelRMSNormGated", "Module", "None", "NotImplementedError", "Optional", "Parameter", "Please", "Tensor", "The", "To", "True", "Y_diag", "Y_off", "_", "__init__", "a", "act", "activation", "and", "apply_mask_to_padding_states", "arange", "attention_mask", "available", "back", "batch_size", "be", "because", "bias", "bmm", "cache_device", "cache_params", "cache_position", "cat", "causal", "causal_conv1d", "causal_conv1d_fn", "causal_conv1d_update", "chunk_size", "clamp", "class", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_dim", "conv_kernel_size", "conv_states", "copy_", "cuda", "cuda_kernels_forward", "cumsum", "dA", "dB", "dBx", "decay_chunk", "decay_states", "def", "device", "dim", "dims", "dt", "dt_bias", "dt_limit", "dt_limit_kwargs", "dt_softplus", "dtype", "else", "eps", "exp", "expand", "fast", "float", "float32", "follow", "for", "forward", "functional", "gate", "github", "groups", "groups_time_state_size", "has_previous_state", "head_dim", "headdim", "hidden_act", "hidden_size", "hidden_states", "hidden_states_B_C", "hidden_states_B_C_transposed", "hidden_states_reshaped", "https", "if", "implementation", "in", "in_channels", "in_proj", "inf", "input_states", "install", "int", "intermediate_size", "is", "is_fast_path_available", "kernel_size", "kwargs", "layer_idx", "layer_norm_epsilon", "log", "logger", "mamba", "mamba_chunk_scan_combined", "mamba_chunk_size", "mamba_conv_bias", "mamba_d_conv", "mamba_d_head", "mamba_d_state", "mamba_expand", "mamba_n_groups", "mamba_n_heads", "mamba_proj_bias", "mamba_split_conv1d_scan_combined", "mamba_ssm", "model", "n_groups", "naive", "new_states", "ngroups", "nn", "norm", "norm_before_gate", "not", "num_heads", "of", "on", "one", "ones", "out", "out_channels", "out_proj", "outproj_bias", "outproj_weight", "output_size", "pad", "pad_size", "pad_tensor_by_size", "padding", "path", "permute", "previous_states", "projected_states", "projection_size", "raise", "repeat_interleave", "requires", "reshape", "reshape_into_chunks", "return", "return_final_states", "rms_norm_eps", "rmsnorm_eps", "rmsnorm_weight", "roll", "running", "scan_output", "segment_sum", "selective_state_update", "self", "seq_idx", "seq_len", "shape", "shifts", "silu", "softplus", "spaces", "split", "squeeze", "ssm_state", "ssm_state_size", "ssm_states", "ssm_states_reshaped", "state", "state_decay_out", "state_decay_out_permuted", "states", "sum", "super", "support", "swish", "t", "the", "time_step_limit", "time_step_max", "time_step_min", "to", "torch", "torch_forward", "training", "transpose", "type", "use_bias", "use_conv_bias", "use_precomputed_states", "used", "variance_epsilon", "view", "warning_once", "weight", "when", "will", "x", "y", "z", "zeros_like"], "bamba/modeling_bamba.py:BambaMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "bamba/modeling_bamba.py:BambaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "bamba/modeling_bamba.py:BambaDecoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "HybridMambaAttentionDynamicCache", "Invalid", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFlashAttentionKwargs", "ModelMLP", "ModelMixer", "ModelRMSNorm", "None", "Optional", "Tensor", "Unpack", "ValueError", "__init__", "attention", "attention_mask", "bool", "cache_params", "cache_position", "class", "config", "def", "deprecate_kwarg", "elif", "else", "eps", "feed_forward", "ffn_layer_class", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_type", "mamba", "new_name", "num_experts", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "pre_ff_layernorm", "raise", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "str", "super", "torch", "tuple", "use_cache", "version"], "bamba/modeling_bamba.py:BambaPreTrainedModel": ["A_log", "D", "ModelConfig", "ModelDecoderLayer", "ModelMixer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "arange", "base_model_prefix", "class", "config", "data", "def", "dt_bias", "fill_", "if", "isinstance", "log", "model", "module", "num_heads", "past_key_values", "self", "super", "supports_gradient_checkpointing", "torch"], "bamba/modeling_bamba.py:BambaModel": ["AttentionMaskConverter", "BaseModelOutputWithPast", "Embedding", "False", "FloatTensor", "HybridMambaAttentionDynamicCache", "LongTensor", "Model", "ModelConfig", "ModelDecoderLayer", "ModelFlashAttentionKwargs", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_unmask_unattended", "_update_causal_mask", "_update_mamba_mask", "a", "all", "all_hidden_states", "all_self_attns", "an", "and", "append", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "cache", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "decoder_layers", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "exactly", "expand", "fill_value", "final_layernorm", "finfo", "flash_attention_2", "for", "forward", "full", "get_seq_length", "gradient", "gradient_checkpointing", "has_previous_state", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "initialized", "input_ids", "input_tensor", "inputs_embeds", "int", "is", "is_training", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_mask", "layer_outputs", "layer_type", "layers", "layers_block_type", "logger", "mamba", "mamba_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "next_cache", "nn", "no", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_attention_mask", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "provided", "raise", "range", "requires", "reshape", "return", "returned", "rms_norm_eps", "rotary_emb", "sdpa", "self", "sequence_length", "shape", "so", "specify", "staticmethod", "super", "target_length", "to", "torch", "training", "triu", "type", "unsqueeze", "use_cache", "vocab_size", "warning_once", "was", "will", "with", "xpu"], "bamba/modeling_bamba.py:BambaForCausalLM": ["BaseModelOutputWithPast", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "HybridMambaAttentionDynamicCache", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "and", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "contiguous", "cumsum", "def", "device", "dim", "dtype", "elif", "else", "empty_past_kv", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "key", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "logsumexp", "long", "loss", "loss_function", "masked_fill_", "mean", "model", "model_inputs", "nn", "not", "num_logits_to_keep", "or", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "pow", "prepare_inputs_for_generation", "r", "return", "self", "shape", "slice", "slice_indices", "super", "to", "torch", "update", "use_cache", "value", "vocab_size", "weight", "z_loss", "z_loss_coefficient"], "olmo2/modeling_olmo2.py:Olmo2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "olmo2/modeling_olmo2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "olmo2/modeling_olmo2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "olmo2/modeling_olmo2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "k", "k_embed", "k_type", "position_ids", "q", "q_embed", "q_type", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "olmo2/modeling_olmo2.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "olmo2/modeling_olmo2.py:Olmo2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelRMSNorm", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "rms_norm_eps", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "olmo2/modeling_olmo2.py:Olmo2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "olmo2/modeling_olmo2.py:Olmo2DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "post_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "olmo2/modeling_olmo2.py:Olmo2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "olmo2/modeling_olmo2.py:Olmo2PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "olmo2/modeling_olmo2.py:Olmo2Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "olmo2/modeling_olmo2.py:Olmo2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "blip_2/modeling_blip_2.py:Blip2ForConditionalGenerationModelOutput": ["Any", "FloatTensor", "ModelForConditionalGenerationModelOutput", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "in", "k", "keys", "language_model_outputs", "logits", "loss", "not", "qformer_outputs", "r", "return", "self", "to_tuple", "torch", "tuple", "vision_outputs"], "blip_2/modeling_blip_2.py:Blip2ImageTextMatchingModelOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelImageTextMatchingModelOutput", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "blip_2/modeling_blip_2.py:Blip2TextModelOutput": ["FloatTensor", "ModelOutput", "ModelTextModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "blip_2/modeling_blip_2.py:Blip2VisionModelOutput": ["FloatTensor", "ModelOutput", "ModelVisionModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "blip_2/modeling_blip_2.py:Blip2VisionEmbeddings": ["Conv2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "dtype", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embedding", "randn", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "view", "weight", "width"], "blip_2/modeling_blip_2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "torch", "training", "transpose", "value"], "blip_2/modeling_blip_2.py:Blip2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_attn_implementation", "_shape", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bsz", "by", "cat", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key_states", "kwargs", "mixed_qkv", "must", "nn", "not", "num_attention_heads", "num_heads", "permute", "projection", "q_bias", "qkv", "qkv_bias", "query_states", "raise", "requires_grad", "reshape", "return", "scale", "scaling", "self", "seq_len", "size", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_bias", "value_states", "view", "zeros", "zeros_like"], "blip_2/modeling_blip_2.py:Blip2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "blip_2/modeling_blip_2.py:Blip2EncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "blip_2/modeling_blip_2.py:Blip2PreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelForConditionalGeneration", "ModelForImageTextRetrieval", "ModelModel", "ModelPreTrainedModel", "ModelQFormerMultiHeadAttention", "ModelTextEmbeddings", "ModelTextModelWithProjection", "ModelVisionEmbeddings", "ModelVisionModelWithProjection", "None", "OPTDecoderLayer", "PreTrainedModel", "T5Block", "True", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "factor", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "past_key_values", "position_embedding", "query_tokens", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight", "zero_"], "blip_2/modeling_blip_2.py:Blip2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "blip_2/modeling_blip_2.py:Blip2VisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "attentions", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "get_input_embeddings", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "main_input_name", "nn", "pixel_values", "pooled_output", "pooler_output", "post_init", "post_layernorm", "raise", "return", "self", "specify", "super", "to", "torch", "tuple"], "blip_2/modeling_blip_2.py:Blip2QFormerMultiHeadAttention": ["Dropout", "Embedding", "False", "Linear", "ModelQFormerMultiHeadAttention", "Module", "None", "Softmax", "The", "TransformersKwargs", "Unpack", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_map", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_probs_dropped", "attention_scores", "attn_gradients", "bhld", "bhlr", "bhrd", "class", "config", "context_layer", "contiguous", "d", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_size", "encoder_hidden_states", "forward", "get_attention_map", "get_attn_gradients", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "key", "key_layer", "kwargs", "long", "lrd", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "raise", "register_hook", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "save_attention", "save_attention_map", "save_attn_gradients", "self", "seq_length", "size", "sqrt", "super", "the", "to", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "view", "x"], "blip_2/modeling_blip_2.py:Blip2QFormerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelQFormerSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "blip_2/modeling_blip_2.py:Blip2QFormerAttention": ["False", "FloatTensor", "ModelQFormerAttention", "ModelQFormerMultiHeadAttention", "ModelQFormerSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_output", "attn_output", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_cross_attention", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "union", "value"], "blip_2/modeling_blip_2.py:Blip2QFormerIntermediate": ["ACT2FN", "Linear", "ModelQFormerIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "blip_2/modeling_blip_2.py:Blip2QFormerOutput": ["Dropout", "LayerNorm", "Linear", "ModelQFormerOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "blip_2/modeling_blip_2.py:Blip2QFormerLayer": ["False", "GradientCheckpointingLayer", "ModelQFormerAttention", "ModelQFormerIntermediate", "ModelQFormerLayer", "ModelQFormerOutput", "None", "TransformersKwargs", "True", "Unpack", "ValueError", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "be", "cat", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_frequency", "crossattention", "def", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "feed_forward_chunk", "feed_forward_chunk_query", "for", "forward", "given", "has_cross_attention", "head_mask", "hidden_states", "if", "intermediate", "intermediate_output", "intermediate_query", "is", "is_cross_attention", "kwargs", "layer_idx", "layer_output", "layer_output_text", "layers", "must", "output", "output_query", "query_attention_output", "query_length", "raise", "return", "self", "seq_len_dim", "shape", "super", "torch", "use_qformer_text_input"], "blip_2/modeling_blip_2.py:Blip2QFormerEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "False", "ModelQFormerEncoder", "ModelQFormerLayer", "Module", "ModuleList", "None", "TransformersKwargs", "Unpack", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "query_length", "range", "return", "self", "super"], "blip_2/modeling_blip_2.py:Blip2TextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "cat", "class", "config", "def", "device", "dim", "dtype", "else", "embeddings", "expand", "forward", "getattr", "hidden_size", "if", "input_ids", "is", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "query_embeds", "register_buffer", "return", "self", "seq_length", "size", "super", "to", "torch", "vocab_size", "weight", "word_embeddings"], "blip_2/modeling_blip_2.py:Blip2QFormerModel": ["BaseModelOutput", "BaseModelOutputWithPoolingAndCrossAttentions", "Dropout", "False", "FloatTensor", "LayerNorm", "ModelPreTrainedModel", "ModelQFormerConfig", "ModelQFormerEncoder", "ModelQFormerLayer", "ModelQFormerModel", "ModelQFormerMultiHeadAttention", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "Wrong", "_", "__init__", "_can_record_outputs", "_prune_heads", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "check_model_inputs", "class", "config", "cross_attentions", "crossattention", "def", "device", "dim", "dropout", "dtype", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "eps", "extended_attention_mask", "f", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "has_query", "head_mask", "heads", "heads_to_prune", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "in", "index", "input_ids", "input_shape", "int", "invert_attention_mask", "is", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "layer_name", "layer_norm_eps", "layernorm", "list", "mask", "nn", "not", "num_hidden_layers", "ones", "or", "pooled_output", "pooler_output", "post_init", "prune_heads", "query_embeds", "query_length", "r", "raise", "return", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "super", "to", "torch", "tuple", "value", "weight", "word_embeddings"], "blip_2/modeling_blip_2.py:Blip2Model": ["AutoModelForCausalLM", "AutoModelForSeq2SeqLM", "BaseModelOutputWithPooling", "BaseModelOutputWithPoolingAndCrossAttentions", "CausalLMOutputWithPast", "CrossEntropyLoss", "Currently", "Deprecation", "False", "FloatTensor", "In", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGenerationModelOutput", "ModelModel", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "Module", "None", "Optional", "Parameter", "Seq2SeqLMOutput", "Tensor", "To", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_from_config", "_keep_in_fp32_modules", "_supports_flash_attn", "_tie_weights", "_tied_weights_keys", "a", "all", "attention_mask", "auto_docstring", "behavior", "bool", "but", "can_return_tuple", "change", "class", "config", "contiguous", "decoder", "decoder_attention_mask", "decoder_input_ids", "def", "default", "device", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_hidden_states", "expand", "expand_as", "f", "filter_out_non_signature_kwargs", "for", "forward", "from_config", "get_decoder", "get_encoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_placeholder_mask", "get_qformer_features", "get_text_features", "hidden_size", "if", "image_attention_mask", "image_embeds", "image_token_id", "in", "input_ids", "inputs_embeds", "instead", "interpolate_pos_encoding", "is", "it", "k", "kwargs", "labels", "language_model", "language_model_inputs", "language_model_outputs", "language_projection", "last_hidden_state", "legacy_output", "logits", "long", "loss", "loss_fct", "main_input_name", "masked_scatter", "mean", "method", "model", "new", "new_embeddings", "nn", "not", "notice", "now", "num_query_tokens", "object", "of", "ones", "ones_like", "opt", "output", "outputs", "pixel_values", "pooler_output", "post_init", "qformer", "qformer_config", "qformer_outputs", "query_embeds", "query_output", "query_outputs", "query_tokens", "r", "reduction", "return", "return_dict", "returns", "self", "set", "set_input_embeddings", "set_output_embeddings", "shape", "shared", "shift_labels", "shift_logits", "size", "special_image_mask", "starting", "super", "tensor", "text_config", "text_outputs", "the", "this", "to", "torch", "tuple", "unsqueeze", "use_decoder_only_language_model", "v4", "value", "view", "vision_config", "vision_model", "vision_outputs", "vocab_size", "warn", "warnings", "will", "zeros"], "blip_2/modeling_blip_2.py:Blip2TextModelWithProjection": ["False", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelQFormerModel", "ModelTextEmbeddings", "ModelTextModelOutput", "ModelTextModelWithProjection", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_keep_in_fp32_modules", "_supports_flash_attn", "attention_mask", "attentions", "auto_docstring", "can_return_tuple", "class", "config", "def", "dim", "dtype", "embeddings", "forward", "functional", "get_input_embeddings", "hidden_size", "hidden_states", "image_text_hidden_size", "input_ids", "kwargs", "last_hidden_state", "nn", "normalize", "num_query_tokens", "pooled_output", "position_ids", "post_init", "qformer", "qformer_config", "query_embeds", "query_length", "query_tokens", "r", "return", "self", "set_input_embeddings", "super", "supports_gradient_checkpointing", "text_embeds", "text_outputs", "text_projection", "to", "torch", "tuple", "value", "weight", "word_embeddings", "zeros"], "blip_2/modeling_blip_2.py:Blip2VisionModelWithProjection": ["False", "FloatTensor", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "ModelVisionModelOutput", "ModelVisionModelWithProjection", "Module", "None", "Optional", "Parameter", "TransformersKwargs", "Union", "Unpack", "__init__", "_from_config", "_keep_in_fp32_modules", "_supports_flash_attn", "attentions", "auto_docstring", "can_return_tuple", "class", "config", "def", "device", "dim", "dtype", "embeddings", "embeds", "encoder_attention_mask", "encoder_hidden_states", "expand", "forward", "functional", "get_input_embeddings", "hidden_size", "hidden_states", "image_attention_mask", "image_embeds", "image_text_hidden_size", "kwargs", "last_hidden_state", "long", "main_input_name", "nn", "normalize", "num_query_tokens", "ones", "patch_embedding", "pixel_values", "pooled_output", "post_init", "qformer", "qformer_config", "query_embeds", "query_outputs", "query_tokens", "r", "return", "self", "shape", "size", "super", "to", "torch", "tuple", "vision_config", "vision_model", "vision_outputs", "vision_projection", "weight", "zeros"], "blip_2/modeling_blip_2.py:Blip2ForConditionalGeneration": ["AutoModelForCausalLM", "AutoModelForSeq2SeqLM", "CrossEntropyLoss", "False", "FloatTensor", "GPU", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelForConditionalGenerationModelOutput", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "Module", "None", "Optional", "Parameter", "Please", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_from_config", "_hf_hook", "_keep_in_fp32_modules", "_preprocess_accelerate", "_supports_flash_attn", "_tie_weights", "_tied_weights_keys", "a", "accelerate", "all", "and", "are", "attention_mask", "auto_docstring", "batch_size", "behavior", "blob", "blog", "bool", "bos_token_id", "can_return_tuple", "class", "com", "config", "contains", "contiguous", "creating", "cuda", "decoder", "decoder_attention_mask", "decoder_input_ids", "def", "details", "device", "device_count", "device_map", "dictionary", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_hidden_states", "environment", "expand", "expand_as", "f", "for", "forward", "from_config", "generate", "generate_kwargs", "get_decoder", "get_encoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_placeholder_mask", "github", "hasattr", "hf_device_map", "hidden_size", "https", "huggingface", "if", "image_attention_mask", "image_embeds", "image_token_id", "image_token_index", "image_tokens", "in", "input_ids", "inputs", "inputs_embeds", "interpolate_pos_encoding", "io_same_device", "is", "is_encoder_decoder", "k", "kwargs", "labels", "language_model", "language_model_inputs", "language_model_outputs", "language_projection", "large", "last_hidden_state", "lead", "len", "logger", "logits", "long", "loss", "loss_fct", "main", "main_input_name", "masked_scatter", "may", "md", "mean", "models", "more", "multi", "new_embeddings", "nn", "no_grad", "not", "num_query_tokens", "on", "ones", "ones_like", "outputs", "pass", "pixel_values", "post_init", "qformer", "qformer_config", "qformer_outputs", "query_embeds", "query_output", "query_outputs", "query_tokens", "r", "reduction", "refer", "remove", "repeat", "return", "return_dict", "running", "script", "self", "set_input_embeddings", "set_output_embeddings", "shape", "shared", "shift_labels", "shift_logits", "size", "special_image_mask", "start_tokens", "super", "tensor", "text_config", "that", "the", "this", "to", "torch", "tuple", "unexpected", "unsqueeze", "use_decoder_only_language_model", "using", "value", "view", "vision_config", "vision_model", "vision_outputs", "vocab_size", "warning", "when", "you", "your", "zeros"], "blip_2/modeling_blip_2.py:Blip2ForImageTextRetrieval": ["False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForImageTextRetrieval", "ModelImageTextMatchingModelOutput", "ModelPreTrainedModel", "ModelQFormerModel", "ModelTextEmbeddings", "ModelVisionModel", "None", "Optional", "Parameter", "Union", "_", "__init__", "_from_config", "_keep_in_fp32_modules", "_supports_flash_attn", "attention_mask", "auto_docstring", "bool", "cat", "class", "config", "def", "device", "dim", "dtype", "else", "embeddings", "encoder_attention_mask", "encoder_hidden_states", "expand", "forward", "functional", "get_input_embeddings", "hidden_size", "if", "image_attention_mask", "image_embeds", "image_text_hidden_size", "image_token_index", "input_ids", "is", "itm_head", "last_hidden_state", "logits_per_image", "logits_per_text", "long", "main_input_name", "matmul", "max", "mean", "nn", "normalize", "not", "num_query_tokens", "ones", "output", "output_attentions", "output_hidden_states", "pixel_values", "post_init", "qformer", "qformer_config", "query_attention_mask", "query_embeds", "query_length", "query_outputs", "query_tokens", "question_embeds", "r", "return", "return_dict", "self", "set_input_embeddings", "shape", "size", "super", "t", "text_embeds", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "use_image_text_matching_head", "use_return_dict", "value", "vision_config", "vision_model", "vision_model_output", "vision_outputs", "vision_projection", "weight", "word_embeddings", "zeros"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TGenerationOutput": ["FloatTensor", "IntTensor", "ModelGenerationOutput", "ModelOutput", "None", "Optional", "class", "r", "sequences", "torch", "tuple", "unit_sequences", "waveform", "waveform_lengths"], "seamless_m4t/modeling_seamless_m4t.py:shift_tokens_right": ["Model_tokens_right", "Modeled_input_ids", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_token_id", "def", "defined", "has", "if", "input_ids", "int", "is", "masked_fill_", "model", "new_zeros", "pad_token_id", "raise", "return", "self", "shape", "to", "torch"], "seamless_m4t/modeling_seamless_m4t.py:_compute_new_attention_mask": ["Tensor", "_compute_new_attention_mask", "arange", "batch_size", "bool_mask", "def", "device", "expand", "hidden_states", "indices", "mask", "mask_seq_len", "masked_fill", "new_ones", "return", "seq_lens", "shape", "torch", "unsqueeze"], "seamless_m4t/modeling_seamless_m4t.py:format_speech_generation_kwargs": ["Model_speech_generation_kwargs", "def", "elif", "else", "for", "generation_config", "if", "in", "items", "key", "kwargs", "kwargs_speech", "kwargs_text", "len", "not", "return", "speech_", "startswith", "text_", "value"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelConformerPositionalConvEmbedding", "ModelConformerSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "speech_encoder_hidden_act", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerRotaryPositionalEmbedding": ["ModelConformerRotaryPositionalEmbedding", "Module", "None", "__init__", "and", "arange", "base", "cached_rotary_positional_embedding", "cached_sequence_length", "cat", "class", "config", "cos", "cos_embeddings", "def", "dim", "dtype", "einsum", "embeddings", "float", "forward", "freqs", "hidden_size", "hidden_states", "i", "if", "ij", "int64", "inv_freq", "is", "j", "nn", "not", "register_buffer", "return", "rotary_embedding_base", "self", "sequence_length", "shape", "sin", "sin_embeddings", "speech_encoder_attention_heads", "stack", "super", "time_stamps", "torch", "type_as"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerRelPositionalEmbedding": ["ModelConformerRelPositionalEmbedding", "Module", "None", "Tensor", "__init__", "arange", "cat", "class", "config", "cos", "d_model", "def", "device", "dim", "div_term", "dtype", "end_idx", "exp", "expand", "extend_pe", "flip", "float", "forward", "hidden_size", "hidden_states", "if", "int64", "is", "log", "math", "max_len", "max_source_positions", "nn", "not", "or", "pe", "pe_negative", "pe_positive", "position", "relative_position_embeddings", "return", "self", "sin", "size", "start_idx", "super", "tensor", "to", "torch", "unsqueeze", "x", "zeros"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerSamePadLayer": ["ModelConformerSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelConformerFeatureProjection", "Module", "__init__", "class", "config", "def", "dropout", "eps", "feature_projection_input_dim", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "speech_encoder_dropout", "super"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelConformerFeedForward", "Module", "None", "__init__", "act_fn", "class", "config", "def", "dropout", "else", "forward", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "is", "isinstance", "nn", "not", "output_dense", "output_dropout", "return", "self", "speech_encoder_dropout", "speech_encoder_hidden_act", "speech_encoder_intermediate_size", "str", "super"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerConvolutionModule": ["ACT2FN", "BatchNorm1d", "Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelConformerConvolutionModule", "Module", "None", "SAME", "ValueError", "__init__", "a", "activation", "attention_mask", "batch_norm", "be", "bias", "bool", "class", "config", "conv_depthwise_kernel_size", "def", "depthwise_conv", "dim", "dropout", "for", "forward", "glu", "groups", "hidden_size", "hidden_states", "if", "is", "kernel_size", "layer_norm", "masked_fill", "nn", "not", "number", "odd", "padding", "pointwise_conv1", "pointwise_conv2", "raise", "return", "same", "self", "should", "speech_encoder_dropout", "speech_encoder_hidden_act", "stride", "super", "transpose", "unsqueeze"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerSelfAttention": ["Dropout", "False", "Linear", "ModelConformerSelfAttention", "Module", "None", "Optional", "Parameter", "Tensor", "True", "ValueError", "__init__", "_apply_relative_embeddings", "_apply_rotary_embedding", "attention_mask", "batch_size", "be", "bias", "bool", "cat", "class", "config", "cos", "def", "defined", "device", "dim", "dropout", "dtype", "else", "forward", "has", "head_size", "hidden_size", "hidden_states", "if", "is", "key", "linear_k", "linear_out", "linear_pos", "linear_q", "linear_v", "math", "matmul", "ndim", "nn", "not", "num_heads", "output_attentions", "p", "pos_bias_u", "pos_bias_v", "position_embeddings_type", "probs", "proj_relative_position_embeddings", "q_with_bias_u", "q_with_bias_v", "query", "query_key_states", "raise", "relative", "relative_position_embeddings", "reshape", "return", "rotary", "rotated_states", "rotated_states_begin", "rotated_states_end", "scores", "scores_ac", "scores_bd", "scores_bd_padded", "scores_bd_padded_shape", "self", "sequence_length", "shape", "sin", "size", "softmax", "speech_encoder_attention_heads", "speech_encoder_dropout", "sqrt", "super", "to", "torch", "transpose", "tuple", "use_position_embeddings", "value", "value_states", "view", "view_as", "when", "zero_pad", "zeros"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelConformerConvolutionModule", "ModelConformerEncoderLayer", "ModelConformerFeedForward", "ModelConformerSelfAttention", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weigts", "bool", "class", "config", "conv_attention_mask", "conv_module", "def", "dropout", "embed_dim", "ffn1", "ffn1_layer_norm", "ffn2", "ffn2_layer_norm", "final_layer_norm", "forward", "hidden_size", "hidden_states", "nn", "output_attentions", "relative_position_embeddings", "residual", "return", "self", "self_attn", "self_attn_dropout", "self_attn_layer_norm", "speech_encoder_dropout", "super", "torch"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelConformerEncoder", "ModelConformerEncoderLayer", "ModelConformerRelPositionalEmbedding", "ModelConformerRotaryPositionalEmbedding", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "conv_attention_mask", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "embed_positions", "enumerate", "eps", "expand", "finfo", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layers", "masked_fill", "min", "nn", "not", "or", "output_attentions", "output_hidden_states", "position_embeddings_type", "rand", "range", "relative", "relative_position_embeddings", "return", "return_dict", "rotary", "self", "shape", "skip_the_layer", "speech_encoder_dropout", "speech_encoder_layerdrop", "speech_encoder_layers", "super", "synced_gpus", "to", "torch", "training", "tuple", "unsqueeze", "v"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerAdapterLayer": ["Conv1d", "Dropout", "False", "GLU", "LayerNorm", "ModelConformerAdapterLayer", "ModelConformerFeedForward", "ModelConformerSelfAttention", "Module", "None", "Optional", "Tensor", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_prepare_4d_attention_mask", "act_fn", "activation", "adaptor_dropout", "adaptor_kernel_size", "adaptor_stride", "attention_mask", "attn_weights", "bool", "class", "config", "def", "device", "dim", "dropout", "dtype", "embed_dim", "ffn", "ffn_layer_norm", "floor", "forward", "hidden_size", "hidden_states", "if", "int", "is", "kernel_size", "nn", "not", "output_attentions", "pad", "padding", "relu", "residual", "residual_conv", "residual_layer_norm", "return", "self", "self_attn", "self_attn_conv", "self_attn_dropout", "self_attn_layer_norm", "seq_lens", "size", "stride", "sub_sampled_lengths", "sum", "super", "to", "torch", "transpose", "use_position_embeddings"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TConformerAdapter": ["ModelConformerAdapter", "ModelConformerAdapterLayer", "Module", "ModuleList", "_", "__init__", "attention_mask", "class", "config", "def", "for", "forward", "hidden_states", "in", "layer", "layers", "nn", "num_adapter_layers", "range", "return", "self", "super"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TSinusoidalPositionalEmbedding": ["False", "ModelSinusoidalPositionalEmbedding", "Module", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "cat", "class", "contiguous", "cos", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "cumsum", "def", "detach", "device", "dim", "dtype", "else", "emb", "emb_weights", "embedding_dim", "exp", "expand", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "persistent", "position_ids", "register_buffer", "return", "self", "seq_len", "sequence_length", "shape", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TAttention": ["Attention", "Cache", "EncoderDecoderCache", "False", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cache_position", "caching", "call", "class", "config", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "during", "else", "embed_dim", "encoder_hidden_states", "errors", "f", "float", "forward", "functional", "get", "got", "head_dim", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "keys", "layer_idx", "layers", "lead", "logger", "make", "mask", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "proj_shape", "provide", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "should", "size", "softmax", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "weights", "when", "will", "without"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TFeedForwardNetwork": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelFeedForwardNetwork", "Module", "Tensor", "__init__", "act", "activation_dropout", "activation_function", "and", "class", "config", "def", "dropout", "dtype", "fc1", "fc2", "ffn_dim", "forward", "hidden_size", "hidden_states", "if", "int", "int8", "isinstance", "nn", "return", "self", "super", "to", "torch", "uint8", "weight"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelFeedForwardNetwork", "None", "Tensor", "__init__", "activation_dropout", "attention_dropout", "attention_mask", "attn_dropout", "attn_weights", "bool", "class", "config", "def", "dropout", "else", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "ffn", "ffn_dim", "ffn_dropout", "ffn_layer_norm", "forward", "hidden_size", "hidden_states", "if", "is", "nn", "num_heads", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TDecoderLayer": ["ACT2FN", "Cache", "Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelFeedForwardNetwork", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_dropout", "bool", "cache_position", "class", "config", "cross_attention", "cross_attention_layer_norm", "cross_attn_weights", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_hidden_states", "ffn", "ffn_dim", "ffn_dropout", "ffn_layer_norm", "forward", "hidden_size", "hidden_states", "if", "is", "is_decoder", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "use_cache", "version"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TPreTrainedModel": ["BatchNorm1d", "Conv1d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelConformerEncoderLayer", "ModelConformerFeatureProjection", "ModelConformerPositionalConvEmbedding", "ModelConformerSelfAttention", "ModelDecoderLayer", "ModelEncoderLayer", "ModelPreTrainedModel", "Module", "None", "Optional", "PreTrainedModel", "Tensor", "True", "_compute_sub_sample_lengths_from_attention_mask", "_init_weights", "_no_split_modules", "a", "adaptor_kernel_size", "adaptor_stride", "attention_mask", "b", "base_model_prefix", "beam_indices", "beam_indices_mask", "bias", "class", "clone", "compute_last_hidden_states_per_sample", "concat", "config", "constant_", "conv", "data", "def", "dim", "elif", "expand", "fill_", "floor", "for", "gather", "groups", "hasattr", "hidden_states", "if", "in", "in_channels", "in_features", "init", "initializer_range", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "last_hidden_states", "long", "math", "max", "max_beam_length", "mean", "module", "nn", "normal_", "not", "pad", "padding_idx", "pos_bias_u", "pos_bias_v", "projection", "return", "self", "seq_lens", "shape", "size", "sqrt", "std", "stride", "sum", "supports_gradient_checkpointing", "torch", "tuple", "uniform_", "unsqueeze", "weight", "xavier_uniform_", "zero_"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TSpeechEncoder": ["LayerNorm", "ModelConfig", "ModelConformerAdapter", "ModelConformerEncoder", "ModelConformerFeatureProjection", "ModelConformerFeedForward", "ModelPreTrainedModel", "ModelSpeechEncoder", "None", "Optional", "Tensor", "Union", "ValueError", "Wav2Vec2BaseModelOutput", "__init__", "act_fn", "adapter", "add_adapter", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "else", "encoder", "encoder_outputs", "expanded_hidden_states", "feature_projection", "forward", "hidden_size", "hidden_states", "if", "inner_layer_norm", "input_features", "intermediate_ffn", "is", "kwargs", "last_hidden_state", "main_input_name", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "raise", "relu", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TEncoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Pass", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prepare_4d_attention_mask", "all_attentions", "and", "append", "at", "attention_mask", "attentions", "bool", "both", "cannot", "class", "config", "def", "device", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder", "encoder_attention_heads", "encoder_ffn_dim", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "for", "forward", "functional", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "input", "input_ids", "input_shape", "inputs_embeds", "instead", "is", "is_t2u_encoder", "kwargs", "last_hidden_state", "layer_norm", "layer_outputs", "layerdrop", "layers", "math", "max_position_embeddings", "max_source_positions", "model", "nn", "not", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "shape", "specify", "sqrt", "super", "text_to_units", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "weight"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModelSinusoidalPositionalEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "append", "at", "attention_mask", "attentions", "auto_docstring", "be", "bool", "both", "cache_position", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "decoder_attention_heads", "decoder_ffn_dim", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "dtype", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "embedding_dim", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "last_hidden_state", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "logger", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "num_embeddings", "of", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "positions", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TTextToUnitModel": ["BaseModelOutput", "Cache", "Embedding", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelPreTrainedModel", "ModelTextToUnitModel", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "True", "Union", "__init__", "and", "attention_mask", "attentions", "bool", "cache_position", "class", "config", "cross_attentions", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_tokens_decoder", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_t2u_encoder", "isinstance", "last_hidden_state", "len", "nn", "not", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_cache", "use_return_dict"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TTextToUnitForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelPreTrainedModel", "ModelTextToUnitForConditionalGeneration", "ModelTextToUnitModel", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "__setattr__", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "bias", "bool", "cache_position", "changed", "class", "config", "copy", "cross_attentions", "custom_args", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "deepcopy", "def", "device", "else", "embed_tokens", "embed_tokens_decoder", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "get_decoder", "get_encoder", "get_input_embeddings", "get_output_embeddings", "getattr", "hidden_size", "if", "in", "input_ids", "inputs_embeds", "is", "items", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "model", "nn", "not", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "return", "return_dict", "self", "set_input_embeddings", "shift_tokens_right", "since", "speech_encoder", "startswith", "super", "t2u_", "t2u_decoder_start_token_id", "t2u_pad_token_id", "t2u_vocab_size", "text_decoder", "text_encoder", "tie_word_embeddings", "to", "to_dict", "torch", "tuple", "use_cache", "use_return_dict", "val", "value", "view", "vocab_size", "vocoder", "warning", "weight"], "seamless_m4t/modeling_seamless_m4t.py:HifiGanResidualBlock": ["Conv1d", "ModelGanResidualBlock", "Module", "ModuleList", "_", "__init__", "apply_weight_norm", "channels", "class", "conv1", "conv2", "convs1", "convs2", "def", "dilation", "for", "forward", "functional", "get_padding", "hasattr", "hidden_states", "i", "if", "in", "kernel_size", "layer", "leaky_relu", "leaky_relu_slope", "len", "nn", "padding", "parametrizations", "range", "remove_weight_norm", "residual", "return", "self", "stride", "super", "utils", "weight_norm", "zip"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TVariancePredictor": ["Conv1d", "Dropout", "LayerNorm", "Linear", "ModelVariancePredictor", "Module", "ReLU", "Tensor", "__init__", "activation_function", "class", "config", "conv1", "conv2", "def", "dim", "dropout_module", "embed_dim", "forward", "hidden_states", "kernel_size", "ln1", "ln2", "nn", "p", "padding", "proj", "return", "self", "squeeze", "super", "transpose", "unit_embed_dim", "var_pred_dropout", "variance_predictor_kernel_size"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4THifiGan": ["Conv1d", "ConvTranspose1d", "FloatTensor", "HifiGanResidualBlock", "ModelConfig", "ModelHifiGan", "Module", "ModuleList", "__init__", "append", "channels", "class", "config", "conv_post", "conv_pre", "def", "dilation", "enumerate", "for", "forward", "functional", "hidden_states", "i", "in", "input_embeds", "j", "kernel_size", "lang_embed_dim", "leaky_relu", "leaky_relu_slope", "len", "model_in_dim", "nn", "num_kernels", "num_upsamples", "padding", "r", "range", "res_state", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "self", "spkr_embed_dim", "squeeze", "stride", "super", "tanh", "torch", "unit_embed_dim", "upsample_initial_channel", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "waveform", "zip"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TCodeHifiGan": ["Conv1d", "ConvTranspose1d", "Embedding", "LayerNorm", "Linear", "LongTensor", "ModelCodeHifiGan", "ModelConfig", "ModelHifiGan", "ModelVariancePredictor", "Module", "None", "PreTrainedModel", "Tensor", "True", "Union", "__init__", "_conv_out_length", "_get_dur_output_lengths", "_get_output_hifigan_lengths", "_init_weights", "_no_split_modules", "_transpose_conv_out_length", "and", "apply_weight_norm", "batch_first", "bias", "cat", "clamp", "class", "config", "conv_post", "conv_pre", "cumsum", "cumulative_dur_out", "data", "def", "dil", "dilation", "dim", "div", "dur_out", "dur_predictor", "duration", "elif", "else", "enumerate", "expm1", "fill_", "floor", "for", "forward", "gather", "hasattr", "hidden_state", "hidden_states", "hifi_gan", "i", "if", "in", "index", "initializer_range", "input_embeds", "input_ids", "input_length", "input_lengths", "int", "is", "isinstance", "kernel_size", "lang", "lang_embed_dim", "lang_id", "language_embedding", "layer", "len", "lengths", "log_dur_pred", "logger", "long", "main_input_name", "mean", "min", "module", "nn", "normal_", "not", "pad", "pad_sequence", "pad_token_id", "padding_idx", "parametrizations", "post_init", "range", "remove_weight_norm", "repeat", "repeat_interleave", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "return", "rnn", "round", "rounding_mode", "self", "shape", "size", "speaker_embedding", "spkr", "spkr_embed_dim", "spkr_id", "squeeze", "std", "stride", "sum", "super", "t2u_pad_token_id", "torch", "training", "transpose", "tuple", "unit_embed_dim", "unit_embedding", "unit_hifi_gan_vocab_size", "unit_lengths", "unsqueeze", "upsample_kernel_sizes", "upsample_rate", "upsample_rates", "upsampler", "utils", "view", "vocoder_num_langs", "vocoder_num_spkrs", "warning", "weight", "weight_norm", "zero_", "zip"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TForTextToText": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelForTextToText", "ModelPreTrainedModel", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "Union", "ValueError", "__", "__init__", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "changed", "class", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "device", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "forward", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "logits_processor", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pop", "post_init", "prefix_allowed_tokens_fn", "provided", "r", "raise", "replace", "return", "return_dict", "self", "set_input_embeddings", "shared", "shift_tokens_right", "since", "speech_encoder", "stopping_criteria", "super", "synced_gpus", "t2u_model", "tensor", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_tgt_lang_id", "tgt_lang", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "warning", "weight"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TForSpeechToText": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelDecoder", "ModelForSpeechToText", "ModelPreTrainedModel", "ModelSpeechEncoder", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "Union", "ValueError", "__", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "argument", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "changed", "class", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "device", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "forward", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_features", "input_ids", "inputs", "inputs_embeds", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "logits_processor", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pop", "post_init", "prefix_allowed_tokens_fn", "provided", "r", "raise", "replace", "return", "return_dict", "self", "seq_lens", "set_input_embeddings", "shared", "shift_tokens_right", "since", "speech_encoder", "stopping_criteria", "sub_sampled_lengths", "super", "synced_gpus", "t2u_model", "tensor", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_tgt_lang_id", "tgt_lang", "tie_word_embeddings", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "warning", "weight"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TForTextToSpeech": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "If", "It", "Linear", "LongTensor", "ModelCodeHifiGan", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelForTextToSpeech", "ModelForTextToText", "ModelGenerationOutput", "ModelPreTrainedModel", "ModelTextToUnitForConditionalGeneration", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "__", "__init__", "_compute_new_attention_mask", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "and", "arange", "argmax", "argument", "as", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "changed", "class", "clone", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "detach", "device", "doesn", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "for", "format_speech_generation_kwargs", "forward", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "getattr", "hidden_size", "hidden_states", "idx_most_probable_sequences_per_batch", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "key", "kwargs", "kwargs_speech", "kwargs_text", "labels", "lang_code_to_id", "lang_id", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "method", "model", "must", "nn", "no_grad", "not", "num_return_sequences", "output", "output_attentions", "output_hidden_states", "output_scores", "output_unit_ids", "outputs", "pad_token_id", "past_key_values", "post_init", "provided", "r", "raise", "replace", "return", "return_dict", "return_dict_in_generate", "return_intermediate_token_ids", "same", "self", "seq_lens", "sequences", "sequences_scores", "set_input_embeddings", "shape", "shared", "shift_tokens_right", "since", "specify", "speech", "speech_encoder", "spkr_id", "str", "sum", "super", "t", "t2u_decoder_input_ids", "t2u_eos_token_id", "t2u_input_embeds", "t2u_lang_code_to_id", "t2u_model", "t2u_model_attention_mask", "t2u_pad_token_id", "t2u_tgt_lang_id", "tensor", "text", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_generation_output", "text_tgt_lang_id", "tgt_lang", "the", "tie_word_embeddings", "to", "torch", "translated", "tuple", "unit", "unit_ids", "unit_sequences", "use", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "vocoder_lang_code_to_id", "vocoder_offset", "vocoder_tgt_lang_id", "want", "warning", "waveform", "waveform_lengths", "weight", "where", "you"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TForSpeechToSpeech": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "If", "It", "Linear", "LongTensor", "ModelCodeHifiGan", "ModelDecoder", "ModelForSpeechToSpeech", "ModelForSpeechToText", "ModelGenerationOutput", "ModelPreTrainedModel", "ModelSpeechEncoder", "ModelTextToUnitForConditionalGeneration", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "__", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_keys_to_ignore_on_load_missing", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "and", "arange", "argmax", "argument", "as", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "changed", "class", "clone", "config", "cross_attentions", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "detach", "device", "doesn", "elif", "else", "embed_tokens", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "for", "format_speech_generation_kwargs", "forward", "generate", "generation_config", "get", "get_decoder", "get_encoder", "get_input_embeddings", "getattr", "hidden_size", "hidden_states", "idx_most_probable_sequences_per_batch", "if", "in", "input_features", "input_ids", "inputs_embeds", "int", "is", "isinstance", "key", "kwargs", "kwargs_speech", "kwargs_text", "labels", "lang_code_to_id", "lang_id", "last_hidden_state", "len", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "method", "must", "nn", "no_grad", "not", "num_return_sequences", "output", "output_attentions", "output_hidden_states", "output_scores", "output_unit_ids", "outputs", "pad_token_id", "past_key_values", "post_init", "provided", "r", "raise", "replace", "return", "return_dict", "return_dict_in_generate", "return_intermediate_token_ids", "same", "self", "seq_lens", "sequences", "sequences_scores", "set_input_embeddings", "shape", "shared", "shift_tokens_right", "since", "specify", "speech", "speech_encoder", "spkr_id", "str", "sub_sampled_lengths", "sum", "super", "t", "t2u_decoder_input_ids", "t2u_eos_token_id", "t2u_input_embeds", "t2u_lang_code_to_id", "t2u_model", "t2u_model_attention_mask", "t2u_pad_token_id", "t2u_tgt_lang_id", "tensor", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_generation_output", "text_tgt_lang_id", "tgt_lang", "the", "tie_word_embeddings", "to", "torch", "translated", "tuple", "unit_ids", "unit_sequences", "use", "use_cache", "use_return_dict", "value", "view", "vocab_size", "vocoder", "vocoder_lang_code_to_id", "vocoder_offset", "vocoder_tgt_lang_id", "want", "warning", "waveform", "waveform_lengths", "weight", "where", "you"], "seamless_m4t/modeling_seamless_m4t.py:SeamlessM4TModel": ["BaseModelOutput", "Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "If", "It", "Linear", "LongTensor", "Make", "ModelCodeHifiGan", "ModelDecoder", "ModelEncoder", "ModelForSpeechToText", "ModelForTextToText", "ModelGenerationOutput", "ModelModel", "ModelPreTrainedModel", "ModelSpeechEncoder", "ModelTextToUnitForConditionalGeneration", "Model_COMMON_CUSTOM_ARGS", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "This", "True", "Union", "ValueError", "You", "__", "__init__", "_compute_new_attention_mask", "_compute_sub_sample_lengths_from_attention_mask", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "a", "all", "and", "arange", "are", "argmax", "argument", "as", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "been", "bias", "bool", "both", "but", "calls", "changed", "class", "clone", "config", "cross_attentions", "current_modality", "custom_args", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "decoder_start_token_id", "def", "depending", "detach", "device", "elif", "else", "embed_tokens", "empty", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "exclusive", "f", "for", "format_speech_generation_kwargs", "forward", "generate", "generate_speech", "generation_config", "get", "get_encoder", "get_input_embeddings", "getattr", "given", "has", "hidden_size", "hidden_states", "idx_most_probable_sequences_per_batch", "if", "ignored", "in", "input", "input_features", "input_ids", "inputs_embeds", "int", "is", "isinstance", "key", "kwargs", "kwargs_speech", "kwargs_text", "labels", "lang_code_to_id", "lang_id", "last_hidden_state", "least", "len", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "main_input_name", "masked_lm_loss", "method", "modality", "must", "mutually", "nn", "no_grad", "non", "not", "num_return_sequences", "of", "on", "one", "or", "output", "output_attentions", "output_hidden_states", "output_scores", "output_unit_ids", "outputs", "pad_token_id", "past_key_values", "post_init", "priority", "provided", "r", "raise", "replace", "return", "return_dict", "return_dict_in_generate", "return_intermediate_token_ids", "same", "self", "seq_lens", "sequences", "sequences_scores", "set_input_embeddings", "set_modality", "shape", "shared", "shift_tokens_right", "since", "specify", "speech", "speech_encoder", "spkr_id", "str", "sub_sampled_lengths", "sum", "super", "sure", "t2u_decoder_input_ids", "t2u_eos_token_id", "t2u_input_embeds", "t2u_lang_code_to_id", "t2u_model", "t2u_model_attention_mask", "t2u_pad_token_id", "t2u_tgt_lang_id", "tensor", "text", "text_decoder", "text_decoder_input_ids", "text_decoder_lang_to_code_id", "text_encoder", "text_generation_output", "text_tgt_lang_id", "tgt_lang", "that", "the", "them", "through", "tie_word_embeddings", "to", "torch", "translated", "tuple", "unit_ids", "unit_sequences", "use", "use_cache", "use_return_dict", "used", "valid", "value", "view", "vocab_size", "vocoder", "vocoder_lang_code_to_id", "vocoder_offset", "vocoder_tgt_lang_id", "want", "warning", "waveform", "waveform_lengths", "weight", "where", "will", "you"], "instructblip/modeling_instructblip.py:InstructBlipForConditionalGenerationModelOutput": ["Any", "FloatTensor", "ModelForConditionalGenerationModelOutput", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "in", "k", "keys", "language_model_outputs", "logits", "loss", "not", "qformer_outputs", "r", "return", "self", "to_tuple", "torch", "tuple", "vision_outputs"], "instructblip/modeling_instructblip.py:InstructBlipVisionEmbeddings": ["Conv2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "dtype", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embedding", "randn", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "view", "weight", "width"], "instructblip/modeling_instructblip.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "torch", "training", "transpose", "value"], "instructblip/modeling_instructblip.py:InstructBlipAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Parameter", "Tensor", "ValueError", "__init__", "_attn_implementation", "_shape", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bsz", "by", "cat", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "head_mask", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key_states", "kwargs", "mixed_qkv", "must", "nn", "not", "num_attention_heads", "num_heads", "permute", "projection", "q_bias", "qkv", "qkv_bias", "query_states", "raise", "requires_grad", "reshape", "return", "scale", "scaling", "self", "seq_len", "size", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_bias", "value_states", "view", "zeros", "zeros_like"], "instructblip/modeling_instructblip.py:InstructBlipMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "instructblip/modeling_instructblip.py:InstructBlipEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelMLP", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "instructblip/modeling_instructblip.py:InstructBlipPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "ModelAttention", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "ModelQFormerEmbeddings", "ModelQFormerMultiHeadAttention", "ModelQFormerSelfOutput", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "blip", "class", "class_embedding", "config", "data", "def", "elif", "factor", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "position_embedding", "query_tokens", "self", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight", "zero_"], "instructblip/modeling_instructblip.py:InstructBlipEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch", "tuple"], "instructblip/modeling_instructblip.py:InstructBlipVisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelAttention", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "None", "Optional", "TransformersKwargs", "Union", "Unpack", "ValueError", "You", "__init__", "_can_record_outputs", "attentions", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "get_input_embeddings", "have", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "main_input_name", "nn", "pixel_values", "pooled_output", "pooler_output", "post_init", "post_layernorm", "raise", "return", "self", "specify", "super", "to", "torch", "tuple"], "instructblip/modeling_instructblip.py:InstructBlipQFormerMultiHeadAttention": ["Dropout", "Embedding", "False", "Linear", "ModelQFormerMultiHeadAttention", "Module", "None", "Softmax", "The", "TransformersKwargs", "Unpack", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_map", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_probs_dropped", "attention_scores", "attention_scores_dtype", "attn_gradients", "bhld", "bhlr", "bhrd", "class", "config", "context_layer", "contiguous", "d", "def", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_size", "encoder_hidden_states", "forward", "get_attention_map", "get_attn_gradients", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "key", "key_layer", "kwargs", "long", "lrd", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "raise", "register_hook", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "save_attention", "save_attention_map", "save_attn_gradients", "self", "seq_length", "size", "sqrt", "super", "the", "to", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "view", "x"], "instructblip/modeling_instructblip.py:InstructBlipQFormerSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelQFormerSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "instructblip/modeling_instructblip.py:InstructBlipQFormerAttention": ["False", "FloatTensor", "ModelQFormerAttention", "ModelQFormerMultiHeadAttention", "ModelQFormerSelfOutput", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_output", "attn_output", "class", "config", "def", "dense", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "is_cross_attention", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "set", "super", "torch", "union", "value"], "instructblip/modeling_instructblip.py:InstructBlipQFormerIntermediate": ["ACT2FN", "Linear", "ModelQFormerIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "instructblip/modeling_instructblip.py:InstructBlipQFormerOutput": ["Dropout", "LayerNorm", "Linear", "ModelQFormerOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "instructblip/modeling_instructblip.py:InstructBlipQFormerLayer": ["False", "GradientCheckpointingLayer", "ModelQFormerAttention", "ModelQFormerIntermediate", "ModelQFormerLayer", "ModelQFormerOutput", "None", "TransformersKwargs", "True", "Unpack", "ValueError", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "be", "cat", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_frequency", "crossattention", "def", "device", "dim", "else", "encoder_attention_mask", "encoder_hidden_states", "feed_forward_chunk", "feed_forward_chunk_query", "for", "forward", "given", "has_cross_attention", "head_mask", "hidden_states", "if", "intermediate", "intermediate_output", "intermediate_query", "is", "is_cross_attention", "kwargs", "layer_idx", "layer_output", "layer_output_text", "layers", "must", "output", "output_query", "query_attention_output", "query_length", "raise", "return", "self", "seq_len_dim", "shape", "super", "to", "torch"], "instructblip/modeling_instructblip.py:InstructBlipQFormerEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "False", "ModelQFormerEncoder", "ModelQFormerLayer", "Module", "ModuleList", "None", "TransformersKwargs", "Unpack", "__init__", "attention_mask", "can_return_tuple", "class", "config", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_idx", "layer_module", "nn", "not", "num_hidden_layers", "query_length", "range", "return", "self", "super"], "instructblip/modeling_instructblip.py:InstructBlipQFormerEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "ModelQFormerEmbeddings", "Module", "None", "__init__", "absolute", "arange", "cat", "class", "clone", "config", "def", "device", "dim", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "is", "layer_norm_eps", "layernorm", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "query_embeds", "register_buffer", "return", "self", "seq_length", "size", "super", "to", "torch", "vocab_size", "weight", "word_embeddings"], "instructblip/modeling_instructblip.py:InstructBlipQFormerModel": ["BaseModelOutput", "BaseModelOutputWithPoolingAndCrossAttentions", "False", "FloatTensor", "LongTensor", "ModelPreTrainedModel", "ModelQFormerConfig", "ModelQFormerEmbeddings", "ModelQFormerEncoder", "ModelQFormerLayer", "ModelQFormerModel", "ModelQFormerMultiHeadAttention", "None", "Optional", "OutputRecorder", "Tensor", "TransformersKwargs", "Union", "Unpack", "ValueError", "Wrong", "You", "_", "__init__", "_can_record_outputs", "_prune_heads", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "and", "attention", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "check_model_inputs", "class", "config", "cross_attentions", "crossattention", "def", "device", "dim", "dtype", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "extended_attention_mask", "f", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "has_query", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "index", "input_ids", "input_shape", "int", "invert_attention_mask", "is", "isinstance", "items", "kwargs", "last_hidden_state", "layer", "layer_name", "list", "mask", "not", "num_hidden_layers", "ones", "or", "pooled_output", "pooler_output", "position_ids", "post_init", "prune_heads", "query_embeds", "query_length", "r", "raise", "return", "self", "seq_length", "sequence_output", "set_input_embeddings", "shape", "size", "specify", "super", "to", "torch", "tuple", "value", "when", "word_embeddings"], "instructblip/modeling_instructblip.py:InstructBlipModel": ["AutoModel", "False", "FlashAttentionKwargs", "FloatTensor", "GPU", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGenerationModelOutput", "ModelModel", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "None", "Optional", "Parameter", "Please", "Tensor", "The", "True", "Union", "Unpack", "__init__", "_hf_hook", "_keep_in_fp32_modules", "_no_split_modules", "_preprocess_accelerate", "_tie_weights", "a", "accelerate", "all", "and", "are", "attention_mask", "auto_docstring", "behavior", "blob", "blog", "bool", "can_return_tuple", "cat", "class", "com", "config", "contains", "creating", "cuda", "decoder", "decoder_attention_mask", "decoder_input_ids", "def", "details", "device", "device_count", "device_map", "dictionary", "dim", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_hidden_states", "environment", "expand", "expand_as", "extend", "for", "forward", "from_config", "get_input_embeddings", "get_placeholder_mask", "github", "hasattr", "hf_device_map", "hidden_size", "https", "huggingface", "if", "image_attention_mask", "image_embeds", "image_token_id", "in", "input_ids", "inputs_embeds", "interpolate_pos_encoding", "io_same_device", "is", "kwargs", "language_model", "language_model_inputs", "language_model_outputs", "language_projection", "large", "lead", "len", "logger", "long", "main", "main_input_name", "masked_scatter", "may", "md", "models", "more", "multi", "nn", "not", "num_query_tokens", "on", "ones", "ones_like", "outputs", "pass", "pixel_values", "post_init", "qformer", "qformer_attention_mask", "qformer_config", "qformer_input_ids", "qformer_outputs", "query_attention_mask", "query_embeds", "query_output", "query_outputs", "query_tokens", "r", "refer", "remove", "return", "running", "script", "self", "set_input_embeddings", "shape", "shared", "size", "special_image_mask", "super", "tensor", "text_config", "that", "the", "this", "to", "torch", "tuple", "unexpected", "unsqueeze", "use_decoder_only_language_model", "using", "value", "vision_config", "vision_model", "vision_outputs", "warning", "when", "you", "your", "zeros"], "instructblip/modeling_instructblip.py:InstructBlipForConditionalGeneration": ["AutoModelForCausalLM", "AutoModelForSeq2SeqLM", "False", "FloatTensor", "GPU", "GenerationMixin", "Linear", "LongTensor", "ModelConfig", "ModelForConditionalGeneration", "ModelForConditionalGenerationModelOutput", "ModelPreTrainedModel", "ModelQFormerModel", "ModelVisionModel", "Module", "None", "Optional", "Parameter", "Please", "The", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_can_compile_fullgraph", "_from_config", "_hf_hook", "_keep_in_fp32_modules", "_no_split_modules", "_preprocess_accelerate", "_tie_weights", "a", "accelerate", "all", "and", "are", "attention_mask", "auto_docstring", "batch_size", "behavior", "blob", "blog", "bool", "bos_token_id", "can_return_tuple", "cat", "class", "com", "config", "contains", "creating", "cuda", "decoder", "decoder_attention_mask", "decoder_input_ids", "def", "details", "device", "device_count", "device_map", "dictionary", "dim", "dtype", "else", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_hidden_states", "environment", "expand", "expand_as", "extend", "for", "forward", "from_config", "generate", "generate_kwargs", "get_decoder", "get_encoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "get_placeholder_mask", "github", "hasattr", "hf_device_map", "hidden_size", "https", "huggingface", "if", "image_attention_mask", "image_embeds", "image_token_id", "image_token_index", "image_tokens", "in", "input_ids", "inputs", "inputs_embeds", "interpolate_pos_encoding", "io_same_device", "is", "is_encoder_decoder", "kwargs", "labels", "language_model", "language_model_inputs", "language_model_outputs", "language_projection", "large", "lead", "len", "logger", "logits", "long", "loss", "loss_function", "main", "main_input_name", "masked_scatter", "may", "md", "models", "more", "multi", "new_embeddings", "nn", "no_grad", "not", "num_query_tokens", "on", "ones", "ones_like", "outputs", "pass", "pixel_values", "post_init", "qformer", "qformer_attention_mask", "qformer_config", "qformer_input_ids", "qformer_outputs", "query_attention_mask", "query_embeds", "query_output", "query_outputs", "query_tokens", "r", "refer", "remove", "repeat", "return", "return_dict", "running", "script", "self", "set_input_embeddings", "set_output_embeddings", "shape", "shared", "size", "special_image_mask", "start_tokens", "super", "tensor", "text_config", "that", "the", "this", "to", "torch", "tuple", "unexpected", "unsqueeze", "use_decoder_only_language_model", "using", "value", "vision_config", "vision_model", "vision_outputs", "vocab_size", "warning", "when", "you", "your", "zeros"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "_norm", "class", "def", "dim", "eps", "extra_repr", "f", "float", "forward", "int", "keepdim", "mean", "nn", "output", "pow", "return", "rsqrt", "self", "shape", "super", "torch", "tuple", "type_as", "weight", "x", "zeros"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_activation", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "vaultgemma/modeling_vaultgemma.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "vaultgemma/modeling_vaultgemma.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "vaultgemma/modeling_vaultgemma.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "vaultgemma/modeling_vaultgemma.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_dim", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softcap", "softmax", "tanh", "to", "torch", "training", "transpose", "tuple", "value", "value_states"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_logit_softcapping", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_pre_attn_scalar", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "softcap", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "pre_feedforward_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaPreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "RMSNorm", "True", "__class__", "__name__", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "in", "model", "module", "past_key_values", "self", "super", "supports_gradient_checkpointing", "weight", "zero_"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "checkpointing", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "dtype", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layers", "logger", "mask_kwargs", "must", "nn", "norm", "normalizer", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "tensor", "torch", "training", "unsqueeze", "use_cache", "vocab_size", "warning_once", "with"], "vaultgemma/modeling_vaultgemma.py:VaultGemmaForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "final_logit_softcapping", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "tanh", "torch", "use_cache", "vocab_size", "weight"], "mpnet/modeling_mpnet.py:MPNetPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLMHead", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "mpnet/modeling_mpnet.py:MPNetEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "Model", "Module", "None", "__init__", "arange", "class", "config", "create_position_ids_from_input_ids", "create_position_ids_from_inputs_embeds", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "kwargs", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "padding_idx", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "sequence_length", "size", "super", "torch", "unsqueeze", "vocab_size", "word_embeddings"], "mpnet/modeling_mpnet.py:MPNetSelfAttention": ["Dropout", "False", "Linear", "Model", "Module", "None", "The", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "c", "class", "config", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "functional", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "k", "kwargs", "math", "matmul", "multiple", "new_c_shape", "nn", "not", "num_attention_heads", "number", "o", "of", "output_attentions", "outputs", "permute", "position_bias", "q", "raise", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "v", "view"], "mpnet/modeling_mpnet.py:MPNetAttention": ["Dropout", "False", "LayerNorm", "Model", "ModelSelfAttention", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "attn", "class", "config", "def", "dim", "dropout", "eps", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "index", "k", "kwargs", "layer_norm_eps", "len", "nn", "num_attention_heads", "o", "output_attentions", "outputs", "position_bias", "prune_heads", "prune_linear_layer", "pruned_heads", "q", "return", "self", "self_outputs", "set", "super", "union", "v"], "mpnet/modeling_mpnet.py:MPNetIntermediate": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "mpnet/modeling_mpnet.py:MPNetOutput": ["Dropout", "LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "mpnet/modeling_mpnet.py:MPNetLayer": ["False", "Model", "ModelAttention", "ModelIntermediate", "ModelOutput", "Module", "None", "__init__", "attention", "attention_mask", "attention_output", "class", "config", "def", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "nn", "output", "output_attentions", "outputs", "position_bias", "return", "self", "self_attention_outputs", "super"], "mpnet/modeling_mpnet.py:MPNetEncoder": ["BaseModelOutput", "Embedding", "False", "Model", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "abs", "all_attentions", "all_hidden_states", "arange", "attention_mask", "attentions", "bool", "bsz", "class", "compute_position_bias", "config", "context_position", "contiguous", "def", "device", "dtype", "else", "enumerate", "expand", "float", "for", "forward", "full_like", "head_mask", "hidden_states", "i", "if", "in", "is", "is_small", "klen", "kwargs", "last_hidden_state", "layer", "layer_module", "layer_outputs", "log", "long", "math", "max_distance", "max_exact", "memory_position", "min", "n", "n_heads", "nn", "not", "num_attention_heads", "num_buckets", "num_hidden_layers", "output_attentions", "output_hidden_states", "permute", "position_bias", "position_ids", "qlen", "range", "relative_attention_bias", "relative_attention_num_buckets", "relative_position", "relative_position_bucket", "ret", "return", "return_dict", "rp_bucket", "self", "size", "staticmethod", "super", "to", "torch", "tuple", "unsqueeze", "v", "val_if_large", "values", "where", "x"], "mpnet/modeling_mpnet.py:MPNetPooler": ["Linear", "Model", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "mpnet/modeling_mpnet.py:MPNetModel": ["BaseModelOutputWithPooling", "FloatTensor", "LongTensor", "Model", "ModelEmbeddings", "ModelEncoder", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "auto_docstring", "bool", "both", "cannot", "class", "config", "def", "device", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "items", "kwargs", "last_hidden_state", "layer", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings"], "mpnet/modeling_mpnet.py:MPNetForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelLMHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_tied_weights_keys", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "class", "config", "decoder", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "lm_head", "logits", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "not", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "torch", "tuple", "use_return_dict", "view", "vocab_size"], "mpnet/modeling_mpnet.py:MPNetLMHead": ["False", "LayerNorm", "Linear", "Model", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "dense", "eps", "features", "forward", "gelu", "hidden_size", "kwargs", "layer_norm", "layer_norm_eps", "nn", "return", "self", "super", "torch", "vocab_size", "x", "zeros"], "mpnet/modeling_mpnet.py:MPNetForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dtype", "elif", "else", "forward", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_output", "single_label_classification", "squeeze", "super", "torch", "tuple", "use_return_dict", "view"], "mpnet/modeling_mpnet.py:MPNetForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_position_ids", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "shape", "size", "super", "torch", "tuple", "use_return_dict", "view"], "mpnet/modeling_mpnet.py:MPNetForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "add_pooling_layer", "attention_mask", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "dropout", "else", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict", "view"], "mpnet/modeling_mpnet.py:MPNetClassificationHead": ["Dropout", "Linear", "Model", "Module", "__init__", "class", "config", "def", "dense", "dropout", "features", "forward", "hidden_dropout_prob", "hidden_size", "kwargs", "nn", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "mpnet/modeling_mpnet.py:MPNetForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "add_pooling_layer", "and", "attention_mask", "attentions", "auto_docstring", "bool", "clamp", "class", "config", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "position_ids", "post_init", "qa_outputs", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_return_dict"], "mpnet/modeling_mpnet.py:create_position_ids_from_input_ids": ["Model_position_ids_from_input_ids", "cumsum", "def", "dim", "incremental_indices", "input_ids", "int", "long", "mask", "ne", "padding_idx", "return", "torch", "type_as"], "jamba/modeling_jamba.py:load_balancing_loss_func": ["Model_balancing_loss_func", "None", "Optional", "Tensor", "Union", "_", "attention_mask", "batch_size", "cat", "compute_device", "concatenated_router_logits", "def", "device", "device_index", "dim", "else", "expand", "expert_attention_mask", "expert_mask", "float", "for", "functional", "if", "in", "index", "int", "is", "isinstance", "layer_router", "mean", "nn", "not", "num_experts", "num_hidden_layers", "one_hot", "or", "overall_loss", "r", "rank", "reshape", "return", "router_logits", "router_per_expert_attention_mask", "router_prob_per_expert", "routing_weights", "selected_experts", "sequence_length", "shape", "softmax", "sum", "to", "tokens_per_expert", "top_k", "topk", "torch", "tuple", "unsqueeze"], "jamba/modeling_jamba.py:JambaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "jamba/modeling_jamba.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "jamba/modeling_jamba.py:HybridMambaAttentionDynamicCache": ["Any", "False", "LongTensor", "ModelMambaAttentionDynamicCache", "None", "Optional", "Tensor", "_", "__init__", "append", "batch_size", "beam_idx", "cache_kwargs", "cat", "class", "config", "conv_kernel_size", "conv_states", "def", "device", "dict", "dim", "dtype", "else", "float16", "for", "get_seq_length", "has_previous_state", "hidden_size", "i", "if", "in", "index_select", "int", "intermediate_size", "is_compileable", "key_cache", "key_states", "layer_idx", "layers_block_type", "len", "mamba", "mamba_d_conv", "mamba_d_state", "mamba_expand", "not", "num_hidden_layers", "range", "reorder_cache", "return", "self", "shape", "ssm_state_size", "ssm_states", "str", "tensor", "to", "torch", "transformer_layers", "tuple", "update", "value_cache", "value_states", "zeros"], "jamba/modeling_jamba.py:JambaAttention": ["False", "HybridMambaAttentionDynamicCache", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "ValueError", "_", "__class__", "__init__", "__name__", "a", "and", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "but", "by", "cache_position", "caching", "call", "causal_mask", "class", "config", "contiguous", "creating", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "dtype", "during", "errors", "f", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "layer_idx", "lead", "logger", "make", "math", "matmul", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "of", "output_attentions", "p", "passing", "past_key_value", "past_key_values", "position_ids", "provide", "q_len", "q_proj", "query_states", "raise", "recommended", "repeat_kv", "reshape", "return", "self", "shape", "should", "size", "softmax", "sqrt", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "jamba/modeling_jamba.py:JambaFlashAttention2": ["False", "HybridMambaAttentionDynamicCache", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "Tensor", "The", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "args", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "back", "be", "bool", "bsz", "cache_position", "cast", "casted", "class", "config", "contiguous", "cpu", "def", "device", "device_type", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "f", "fact", "flash_attn_supports_top_left_mask", "float32", "forward", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "hasattr", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "in", "input", "input_dtype", "is", "is_autocast_enabled", "is_causal", "k_proj", "key_states", "kwargs", "layer", "layer_idx", "layers", "logger", "might", "mps", "norm", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "or", "output_attentions", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "related", "repeat_kv", "reshape", "return", "seems", "self", "silently", "size", "sliding_window", "states", "super", "target_dtype", "the", "this", "to", "torch", "training", "transpose", "type", "upcasted", "update", "use_cache", "use_top_left_mask", "v_proj", "value_states", "view", "warning_once", "weight", "will", "you"], "jamba/modeling_jamba.py:JambaSdpaAttention": ["Falling", "False", "HybridMambaAttentionDynamicCache", "LongTensor", "ModelAttention", "ModelModel", "ModelSdpaAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "_", "and", "argument", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_mask", "attn_output", "back", "be", "bool", "bsz", "but", "cache_position", "can", "causal_mask", "class", "contiguous", "cuda", "def", "deprecate_kwarg", "device", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "head_dim", "hidden_size", "hidden_states", "if", "implementation", "is", "is_causal", "k_proj", "key_states", "layer_idx", "loading", "logger", "manual", "model", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "onwards", "output_attentions", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "removed", "repeat_kv", "required", "return", "scaled_dot_product_attention", "self", "shape", "size", "specifying", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "using", "v5", "v_proj", "value_states", "version", "view", "warning", "warning_once", "when", "will"], "jamba/modeling_jamba.py:JambaMambaMixer": ["A", "ACT2FN", "AILab", "A_log", "B", "C", "CUDA", "Conv1d", "D", "Dao", "False", "Fast", "HybridMambaAttentionDynamicCache", "If", "Linear", "LongTensor", "Make", "Mamba", "ModelConfig", "ModelMambaMixer", "ModelRMSNorm", "Module", "None", "Optional", "Parameter", "Tensor", "The", "To", "True", "ValueError", "_", "__init__", "a", "act", "activation", "and", "append", "arange", "are", "attention_mask", "available", "b_layernorm", "batch_size", "because", "bias", "c_layernorm", "cache_params", "causal", "causal_conv1d_fn", "causal_conv1d_update", "chunk", "class", "clone", "com", "config", "contextualized_states", "contiguous", "conv1d", "conv_kernel_size", "conv_state", "conv_states", "conv_weights", "copy_", "cuda", "cuda_kernels_forward", "data", "def", "deltaB_u", "delta_softplus", "device", "dim", "dims", "discrete_A", "discrete_B", "discrete_time_step", "dt_layernorm", "dt_proj", "dt_softplus", "dtype", "else", "eps", "exp", "expand", "fast", "float", "follow", "for", "forward", "functional", "gate", "github", "groups", "has_previous_state", "hidden_act", "hidden_size", "hidden_states", "https", "i", "if", "implementation", "in", "in_channels", "in_proj", "input_states", "install", "installed", "intermediate_size", "is", "is_fast_path_available", "isinstance", "kernel_size", "kernels", "layer_idx", "log", "logger", "mamba", "mamba_conv_bias", "mamba_d_conv", "mamba_d_state", "mamba_dt_rank", "mamba_expand", "mamba_inner_fn", "mamba_proj_bias", "matmul", "model", "module", "naive", "nn", "no_grad", "not", "of", "on", "one", "ones", "or", "out_channels", "out_proj", "pad", "padding", "path", "projected_states", "raise", "range", "return", "return_last_state", "rms_norm_eps", "roll", "scan_output", "scan_outputs", "selective_scan_fn", "selective_state_update", "self", "seq_len", "set", "shape", "shifts", "size", "slow_forward", "softplus", "spaces", "split", "squeeze", "ssm_parameters", "ssm_state", "ssm_state_size", "ssm_states", "stack", "state", "sum", "super", "sure", "that", "the", "they", "time_proj_bias", "time_step", "time_step_rank", "to", "torch", "training", "transpose", "type", "unsqueeze", "use", "use_bias", "use_cache", "use_conv_bias", "use_fast_kernels", "use_mamba_kernels", "use_precomputed_states", "view", "want", "warning_once", "weight", "with", "x_proj", "you", "zeros", "zeros_like"], "jamba/modeling_jamba.py:JambaMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "jamba/modeling_jamba.py:JambaSparseMoeBlock": ["F", "False", "Linear", "ModelConfig", "ModelMLP", "ModelSparseMoeBlock", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "batch_size", "bias", "class", "config", "continue", "current_hidden_states", "current_state", "def", "device", "dim", "dtype", "expert_idx", "expert_layer", "expert_mask", "experts", "ffn_dim", "final_hidden_states", "float", "for", "forward", "functional", "hidden_dim", "hidden_size", "hidden_states", "idx", "if", "in", "index_add_", "intermediate_size", "nn", "num_classes", "num_experts", "num_experts_per_tok", "one_hot", "permute", "range", "reshape", "return", "router", "router_logits", "routing_weights", "selected_experts", "self", "sequence_length", "shape", "softmax", "super", "to", "top_k", "top_x", "topk", "torch", "tuple", "view", "where", "zeros"], "jamba/modeling_jamba.py:JambaAttentionDecoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "HybridMambaAttentionDynamicCache", "LongTensor", "ModelAttentionDecoderLayer", "ModelConfig", "ModelMLP", "ModelRMSNorm", "ModelSparseMoeBlock", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "feed_forward", "ff_outputs", "ffn_layer_class", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "isinstance", "layer_idx", "layers_num_experts", "new_name", "num_experts", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_ids", "pre_ff_layernorm", "present_key_value", "residual", "return", "rms_norm_eps", "router_logits", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "jamba/modeling_jamba.py:JambaMambaDecoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "HybridMambaAttentionDynamicCache", "LongTensor", "ModelConfig", "ModelMLP", "ModelMambaDecoderLayer", "ModelMambaMixer", "ModelRMSNorm", "ModelSparseMoeBlock", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "cache_params", "cache_position", "class", "config", "def", "deprecate_kwarg", "else", "eps", "feed_forward", "ff_outputs", "ffn_layer_class", "forward", "hidden_size", "hidden_states", "if", "input_layernorm", "int", "isinstance", "layer_idx", "layers_num_experts", "mamba", "new_name", "num_experts", "output_attentions", "output_router_logits", "outputs", "past_key_value", "past_key_values", "position_ids", "pre_ff_layernorm", "residual", "return", "rms_norm_eps", "router_logits", "self", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "jamba/modeling_jamba.py:JambaPreTrainedModel": ["A", "A_log", "Conv1d", "D", "Embedding", "Linear", "ModelAttentionDecoderLayer", "ModelConfig", "ModelMambaDecoderLayer", "ModelMambaMixer", "ModelPreTrainedModel", "ModelRMSNorm", "None", "PreTrainedModel", "True", "_init_weights", "_is_stateful", "_no_split_modules", "_skip_keys_device_placement", "_supports_flash_attn", "_supports_sdpa", "arange", "base_model_prefix", "bias", "class", "config", "contiguous", "copy_", "data", "def", "elif", "expand", "fill_", "if", "initializer_range", "intermediate_size", "is", "isinstance", "log", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "ssm_state_size", "std", "supports_gradient_checkpointing", "torch", "weight", "zero_"], "jamba/modeling_jamba.py:JambaModel": ["ALL_DECODER_LAYER_TYPES", "AttentionMaskConverter", "Embedding", "False", "FloatTensor", "HybridMambaAttentionDynamicCache", "LongTensor", "Model", "ModelConfig", "ModelMambaDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModuleList", "MoeModelOutputWithPast", "None", "Optional", "Setting", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_unmask_unattended", "_update_causal_mask", "_update_mamba_mask", "a", "all", "all_hidden_states", "all_router_logits", "all_self_attns", "an", "and", "append", "arange", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cache", "cache_position", "can_return_tuple", "causal_mask", "checkpointing", "class", "clone", "config", "cuda", "decoder_layer", "decoder_layers", "def", "device", "diagonal", "dim", "dtype", "else", "embed_tokens", "eps", "eq", "exactly", "expand", "fill_value", "final_layernorm", "finfo", "flash_attention_2", "for", "forward", "full", "gradient", "gradient_checkpointing", "has_previous_state", "hidden_size", "hidden_states", "i", "if", "in", "incompatible", "initialized", "input_ids", "input_tensor", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_class", "layer_idx", "layer_mask", "layer_outputs", "layers", "layers_block_type", "logger", "mamba_mask", "mask_length", "masked_fill", "min", "min_dtype", "must", "next_cache", "nn", "no", "not", "npu", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "output_router_logits", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "position_ids", "post_init", "provided", "raise", "range", "requires", "reshape", "return", "returned", "rms_norm_eps", "router_logits", "sdpa", "self", "sequence_length", "shape", "so", "specify", "super", "target_length", "to", "torch", "training", "triu", "type", "unsqueeze", "use_cache", "vocab_size", "warning_once", "was", "will", "with", "xpu"], "jamba/modeling_jamba.py:JambaForCausalLM": ["False", "FloatTensor", "GenerationMixin", "HybridMambaAttentionDynamicCache", "Linear", "LongTensor", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "MoeCausalLMOutputWithPast", "MoeModelOutputWithPast", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "aux_loss", "bias", "bool", "cache_position", "can_return_tuple", "class", "config", "contiguous", "cumsum", "def", "device", "dtype", "elif", "else", "empty_past_kv", "for", "forward", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "items", "key", "kwargs", "labels", "last_hidden_state", "lm_head", "load_balancing_loss_func", "logits", "logits_to_keep", "long", "loss", "loss_function", "masked_fill_", "model", "model_inputs", "nn", "not", "num_experts", "num_experts_per_tok", "num_logits_to_keep", "or", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "past_key_values", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "router_aux_loss_coef", "router_logits", "self", "shape", "slice", "slice_indices", "super", "to", "torch", "update", "use_cache", "value", "vocab_size", "weight"], "jamba/modeling_jamba.py:JambaForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class"], "aimv2/modeling_aimv2.py:Aimv2Output": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "aimv2/modeling_aimv2.py:Aimv2RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "aimv2/modeling_aimv2.py:Aimv2MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "aimv2/modeling_aimv2.py:Aimv2VisionEmbeddings": ["Conv2d", "Embedding", "False", "ModelRMSNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Tensor", "_", "__init__", "arange", "build_2d_sincos_position_embedding", "class", "concat", "config", "cos", "cpu", "def", "device", "dim", "dtype", "else", "embed_dim", "expand", "flatten", "float32", "forward", "grid_h", "grid_w", "height", "hidden_size", "hidden_states", "if", "image_size", "indexing", "int", "is_native", "kernel_size", "meshgrid", "nn", "not", "num_channels", "num_patches", "omega", "out_h", "out_w", "patch_embed", "patch_size", "persistent", "pixel_values", "pos_dim", "pos_embed", "position_embedding", "position_ids", "register_buffer", "return", "rms_norm", "rms_norm_eps", "self", "sin", "size", "staticmethod", "stride", "super", "temperature", "torch", "transpose", "width", "xy"], "aimv2/modeling_aimv2.py:Aimv2TextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "aimv2/modeling_aimv2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "aimv2/modeling_aimv2.py:Aimv2Attention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bias", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "qkv_bias", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "aimv2/modeling_aimv2.py:Aimv2EncoderLayer": ["GradientCheckpointingLayer", "ModelAttention", "ModelEncoderLayer", "ModelMLP", "ModelRMSNorm", "ModelVisionConfig", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention", "attention_mask", "attn_output", "class", "config", "def", "ffn", "forward", "hidden_size", "hidden_states", "kwargs", "mlp_output", "norm_hidden_states", "return", "rms_norm1", "rms_norm2", "rms_norm_eps", "self", "super", "torch"], "aimv2/modeling_aimv2.py:Aimv2Encoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "aimv2/modeling_aimv2.py:Aimv2AttentionPoolingHead": ["F", "Linear", "ModelAttentionPoolingHead", "ModelVisionConfig", "Module", "Parameter", "Tensor", "True", "__init__", "attn_output", "batch_size", "bias", "class", "cls_token", "config", "def", "dim", "expand", "forward", "hidden_dim", "hidden_size", "hidden_states", "k_proj", "key", "mean", "nn", "num_attention_heads", "num_heads", "output", "output_proj", "permute", "qkv_bias", "query", "reshape", "return", "scaled_dot_product_attention", "self", "seq_len", "shape", "super", "torch", "transpose", "v_proj", "value", "zeros"], "aimv2/modeling_aimv2.py:Aimv2PreTrainedModel": ["Model", "ModelAttentionPoolingHead", "ModelConfig", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "Parameter", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "class", "cls_token", "config", "data", "def", "elif", "fill_", "hasattr", "if", "initializer_range", "isinstance", "log", "logit_scale", "math", "mean", "module", "nn", "normal_", "self", "std", "super", "supports_gradient_checkpointing"], "aimv2/modeling_aimv2.py:Aimv2VisionModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "ModelAttention", "ModelAttentionPoolingHead", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_can_record_outputs", "attention_mask", "attentions", "auto_docstring", "check_model_inputs", "class", "config", "def", "deprecate_kwarg", "else", "embeddings", "encoder", "encoder_outputs", "forward", "get_input_embeddings", "head", "hidden_size", "hidden_states", "if", "inputs_embeds", "kwargs", "last_hidden_state", "main_input_name", "nn", "patch_embed", "pixel_values", "pooler_output", "post_init", "r", "return", "rms_norm", "rms_norm_eps", "self", "super", "torch", "use_head", "v4", "version"], "aimv2/modeling_aimv2.py:Aimv2TextModel": ["BaseModelOutputWithPooling", "ModelAttention", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelRMSNorm", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_can_record_outputs", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "batch_size", "cache_position", "check_model_inputs", "class", "config", "create_causal_mask", "def", "device", "dim", "dtype", "embeddings", "encoder", "encoder_outputs", "eos_token_id", "expand", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "input_embeds", "input_ids", "inputs_embeds", "int", "is", "kwargs", "last_hidden_state", "long", "main_input_name", "nn", "not", "past_key_values", "pooled_output", "pooler_output", "position_ids", "post_init", "return", "rms_norm", "rms_norm_eps", "self", "seq_len", "set_input_embeddings", "shape", "super", "to", "token_embedding", "torch", "unsqueeze", "value"], "aimv2/modeling_aimv2.py:_get_vector_norm": ["Tensor", "True", "_get_vector_norm", "def", "dim", "keepdim", "normed_tensor", "pow", "return", "square_tensor", "sum", "sum_tensor", "tensor", "torch"], "aimv2/modeling_aimv2.py:Aimv2Model": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelEncoderLayer", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelTextModel", "ModelVisionEmbeddings", "ModelVisionModel", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "_from_config", "_get_vector_norm", "_no_split_modules", "_supports_flash_attn", "attention_mask", "auto_docstring", "bias", "bool", "can_return_tuple", "clamp", "class", "config", "def", "device", "exp", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "kwargs", "log", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "math", "max_log_logit_scale", "max_logit_scale", "nn", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "return", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "resnet/modeling_resnet.py:ResNetConvLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "Identity", "ModelConvLayer", "Module", "None", "Tensor", "__init__", "activation", "bias", "class", "convolution", "def", "else", "forward", "hidden_state", "if", "in_channels", "input", "int", "is", "kernel_size", "nn", "normalization", "not", "out_channels", "padding", "relu", "return", "self", "str", "stride", "super"], "resnet/modeling_resnet.py:ResNetEmbeddings": ["Make", "MaxPool2d", "ModelConfig", "ModelConvLayer", "ModelEmbeddings", "Module", "Tensor", "ValueError", "__init__", "activation", "channel", "class", "config", "configuration", "def", "dimension", "embedder", "embedding", "embedding_size", "forward", "hidden_act", "if", "in", "kernel_size", "match", "nn", "num_channels", "of", "one", "padding", "pixel", "pixel_values", "pooler", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "values", "with"], "resnet/modeling_resnet.py:ResNetShortCut": ["BatchNorm2d", "Conv2d", "False", "ModelShortCut", "Module", "Tensor", "__init__", "bias", "class", "convolution", "def", "forward", "hidden_state", "in_channels", "input", "int", "kernel_size", "nn", "normalization", "out_channels", "return", "self", "stride", "super"], "resnet/modeling_resnet.py:ResNetBasicLayer": ["ACT2FN", "Identity", "ModelBasicLayer", "ModelConvLayer", "ModelShortCut", "Modelidual", "Module", "None", "Sequential", "__init__", "activation", "class", "def", "else", "forward", "hidden_state", "if", "in_channels", "int", "layer", "nn", "or", "out_channels", "relu", "return", "self", "shortcut", "should_apply_shortcut", "str", "stride", "super"], "resnet/modeling_resnet.py:ResNetBottleNeckLayer": ["ACT2FN", "False", "Identity", "ModelBottleNeckLayer", "ModelConvLayer", "ModelShortCut", "Modelidual", "Module", "None", "Sequential", "__init__", "activation", "bool", "class", "def", "downsample_in_bottleneck", "else", "forward", "hidden_state", "if", "in_channels", "int", "kernel_size", "layer", "nn", "not", "or", "out_channels", "reduces_channels", "reduction", "relu", "return", "self", "shortcut", "should_apply_shortcut", "str", "stride", "super"], "resnet/modeling_resnet.py:ResNetStage": ["ModelBasicLayer", "ModelBottleNeckLayer", "ModelConfig", "ModelStage", "Module", "Sequential", "Tensor", "_", "__init__", "activation", "bottleneck", "class", "config", "def", "depth", "downsample_in_bottleneck", "else", "first_layer", "for", "forward", "hidden_act", "hidden_state", "if", "in", "in_channels", "input", "int", "layer", "layer_type", "layers", "nn", "out_channels", "range", "return", "self", "stride", "super"], "resnet/modeling_resnet.py:ResNetEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "append", "bool", "class", "config", "def", "depth", "depths", "downsample_in_first_stage", "else", "embedding_size", "for", "forward", "hidden_sizes", "hidden_state", "hidden_states", "if", "in", "in_channels", "in_out_channels", "is", "last_hidden_state", "nn", "not", "out_channels", "output_hidden_states", "return", "return_dict", "self", "stage_module", "stages", "stride", "super", "tuple", "v", "zip"], "resnet/modeling_resnet.py:ResNetPreTrainedModel": ["BatchNorm2d", "Conv2d", "GroupNorm", "Linear", "Model", "ModelConfig", "ModelConvLayer", "ModelPreTrainedModel", "ModelShortCut", "None", "PreTrainedModel", "_", "_calculate_fan_in_and_fan_out", "_init_weights", "_no_split_modules", "a", "base_model_prefix", "bias", "bound", "class", "config", "constant_", "def", "elif", "else", "fan_in", "fan_out", "if", "init", "is", "isinstance", "kaiming_normal_", "kaiming_uniform_", "main_input_name", "math", "mode", "module", "nn", "nonlinearity", "not", "pixel_values", "relu", "self", "sqrt", "uniform_", "weight"], "resnet/modeling_resnet.py:ResNetModel": ["AdaptiveAvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "__init__", "auto_docstring", "bool", "class", "config", "def", "else", "embedder", "embedding_output", "encoder", "encoder_outputs", "forward", "hidden_states", "if", "is", "last_hidden_state", "nn", "not", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "return", "return_dict", "self", "super", "use_return_dict"], "resnet/modeling_resnet.py:ResNetForImageClassification": ["Flatten", "FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Sequential", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "use_return_dict"], "resnet/modeling_resnet.py:ResNetBackbone": ["BackboneMixin", "BackboneOutput", "False", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embedder", "embedding_output", "embedding_size", "encoder", "enumerate", "feature_maps", "for", "forward", "has_attentions", "hidden_sizes", "hidden_states", "idx", "if", "in", "is", "not", "num_featuModel", "out_featuModel", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "stage", "stage_names", "super", "use_return_dict"], "diffllama/modeling_diffllama.py:DiffLlamaMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "diffllama/modeling_diffllama.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "diffllama/modeling_diffllama.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "diffllama/modeling_diffllama.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "diffllama/modeling_diffllama.py:lambda_init_fn": ["Model_init_fn", "def", "exp", "layer_idx", "math", "return"], "diffllama/modeling_diffllama.py:DiffLlamaAttention": ["Cache", "False", "Instantiating", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Parameter", "Please", "RMSNorm", "Tensor", "True", "_", "__class__", "__init__", "__name__", "a", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_mask", "attn_output", "attn_output1", "attn_output2", "attn_weights", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "caching", "call", "cat", "causal_mask", "chunk", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dim", "dropout", "dtype", "during", "elementwise_affine", "eps", "errors", "exp", "f", "float32", "forward", "functional", "getattr", "groupnorm", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "lambda_1", "lambda_2", "lambda_full", "lambda_init", "lambda_init_fn", "lambda_k1", "lambda_k2", "lambda_q1", "lambda_q2", "lambda_std_dev", "layer_idx", "lead", "logger", "make", "math", "matmul", "max_position_embeddings", "new_name", "nn", "normal", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "p", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "recommended", "repeat", "repeat_kv", "reshape", "return", "rms_norm_eps", "rope_theta", "self", "shape", "sin", "size", "softmax", "sqrt", "sum", "super", "sure", "target_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "diffllama/modeling_diffllama.py:DiffLlamaFlashAttention2": ["Cache", "False", "In", "LongTensor", "ModelAttention", "ModelFlashAttention2", "None", "Optional", "RoPE", "StaticCache", "Tensor", "The", "Tuple", "ValueError", "We", "_", "__init__", "_flash_attention_forward", "_flash_attn_uses_top_left_mask", "_pre_quantization_dtype", "an", "and", "apply_rotary_pos_emb", "are", "args", "at", "attention", "attention_dropout", "attention_mask", "attn_implementation", "attn_output", "attn_output1", "attn_output2", "back", "be", "bool", "bsz", "cache", "cache_kwargs", "cache_position", "cast", "casted", "cat", "chunk", "class", "com", "compatible", "computed", "computing", "config", "containing", "contiguous", "cos", "cpu", "def", "deprecate_kwarg", "device", "device_type", "dim", "dropout", "dropout_rate", "dtype", "elif", "else", "embedding", "embeddings", "exp", "externally", "f", "fact", "flash_attention_2", "flash_attn_supports_top_left_mask", "float32", "forward", "from", "get_autocast_dtype", "get_autocast_gpu_dtype", "getattr", "github", "groupnorm", "hasattr", "have", "head_dim", "hidden", "hidden_states", "https", "huggingface", "if", "implementation", "in", "indexes", "input", "input_dtype", "internally", "is", "is_autocast_enabled", "is_causal", "isinstance", "issue", "k_proj", "key_states", "kwargs", "lambda_1", "lambda_2", "lambda_full", "lambda_init", "lambda_k1", "lambda_k2", "lambda_q1", "lambda_q2", "layer", "layer_idx", "layers", "logger", "make", "mandatory", "mean", "might", "model", "mps", "new_name", "norm", "not", "num_heads", "num_key_value_heads", "o_proj", "of", "open", "or", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "raise", "related", "removed", "repeat", "reshape", "return", "rotary_emb", "sdpa", "seems", "self", "silently", "sin", "size", "sliding_window", "states", "static", "sum", "super", "sure", "target_dtype", "tensor", "tensors", "the", "this", "through", "time", "to", "tokens", "torch", "training", "transformers", "transitioning", "transpose", "tuple", "type", "upcasted", "update", "use", "use_cache", "use_top_left_mask", "using", "v4", "v_proj", "value_states", "value_states1", "value_states2", "version", "view", "warning_once", "weight", "will", "with", "you"], "diffllama/modeling_diffllama.py:DiffLlamaSdpaAttention": ["Cache", "False", "LongTensor", "ModelAttention", "ModelSdpaAttention", "None", "Optional", "Tensor", "_", "and", "apply_rotary_pos_emb", "attention_dropout", "attention_mask", "attn_mask", "attn_output", "attn_output1", "attn_output2", "bool", "bsz", "cache_kwargs", "cache_position", "cat", "causal_mask", "chunk", "class", "contiguous", "cos", "cuda", "def", "deprecate_kwarg", "device", "dim", "dropout_p", "dtype", "else", "exp", "float32", "forward", "functional", "groupnorm", "head_dim", "hidden_states", "if", "is", "is_causal", "k_proj", "key_states", "kwargs", "lambda_1", "lambda_2", "lambda_full", "lambda_init", "lambda_k1", "lambda_k2", "lambda_q1", "lambda_q2", "layer_idx", "new_name", "nn", "not", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "q_len", "q_proj", "query_states", "repeat", "repeat_kv", "return", "scaled_dot_product_attention", "self", "shape", "sin", "size", "sum", "to", "torch", "training", "transpose", "tuple", "type", "update", "use_cache", "v_proj", "value_states", "version", "view"], "diffllama/modeling_diffllama.py:DiffLlamaRMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "diffllama/modeling_diffllama.py:DiffLlamaDecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "Model_ATTENTION_CLASSES", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "_attn_implementation", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "diffllama/modeling_diffllama.py:DiffLlamaPreTrainedModel": ["False", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "data", "def", "hidden_states", "if", "isinstance", "lambda_k1", "lambda_k2", "lambda_q1", "lambda_q2", "lambda_std_dev", "model", "module", "normal_", "past_key_values", "self", "super", "supports_gradient_checkpointing"], "diffllama/modeling_diffllama.py:DiffLlamaRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "diffllama/modeling_diffllama.py:DiffLlamaModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "diffllama/modeling_diffllama.py:DiffLlamaForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "diffllama/modeling_diffllama.py:DiffLlamaForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "diffllama/modeling_diffllama.py:DiffLlamaForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "diffllama/modeling_diffllama.py:DiffLlamaForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "swinv2/modeling_swinv2.py:Swinv2EncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "reshaped_hidden_states", "torch", "tuple"], "swinv2/modeling_swinv2.py:Swinv2ModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "reshaped_hidden_states", "torch", "tuple"], "swinv2/modeling_swinv2.py:Swinv2MaskedImageModelingOutput": ["FloatTensor", "FutureWarning", "ModelMaskedImageModelingOutput", "ModelOutput", "None", "Optional", "Please", "Transformers", "and", "attentions", "attribute", "be", "class", "def", "deprecated", "final", "hidden_states", "in", "instead", "is", "logits", "loss", "of", "output", "property", "r", "reconstruction", "removed", "reshaped_hidden_states", "retrieve", "return", "self", "the", "to", "torch", "tuple", "use", "version", "warn", "warnings", "will"], "swinv2/modeling_swinv2.py:Swinv2ImageClassifierOutput": ["FloatTensor", "ModelImageClassifierOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "r", "reshaped_hidden_states", "torch", "tuple"], "swinv2/modeling_swinv2.py:window_partition": ["Model_partition", "Model_size", "Models", "batch_size", "contiguous", "def", "height", "input_feature", "num_channels", "permute", "return", "shape", "view", "width"], "swinv2/modeling_swinv2.py:window_reverse": ["Model_reverse", "Model_size", "Models", "contiguous", "def", "height", "num_channels", "permute", "return", "shape", "view", "width"], "swinv2/modeling_swinv2.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "swinv2/modeling_swinv2.py:Swinv2DropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "swinv2/modeling_swinv2.py:Swinv2Embeddings": ["BoolTensor", "Dropout", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "config", "def", "dim", "dropout", "else", "embed_dim", "embeddings", "expand", "forward", "functional", "grid_size", "height", "hidden_dropout_prob", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "norm", "not", "num_channels", "num_patches", "num_positions", "output_dimensions", "patch_embeddings", "patch_grid", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "tuple", "type_as", "unsqueeze", "use_absolute_embeddings", "use_mask_token", "view", "width", "zeros"], "swinv2/modeling_swinv2.py:Swinv2PatchEmbeddings": ["Conv2d", "FloatTensor", "Iterable", "ModelPatchEmbeddings", "Module", "Optional", "Tensor", "_", "__init__", "abc", "class", "collections", "config", "def", "else", "embed_dim", "embeddings", "flatten", "forward", "functional", "grid_size", "height", "hidden_size", "if", "image_size", "int", "isinstance", "kernel_size", "maybe_pad", "nn", "num_channels", "num_patches", "output_dimensions", "pad", "pad_values", "patch_size", "pixel_values", "projection", "return", "self", "shape", "stride", "super", "torch", "transpose", "tuple", "width"], "swinv2/modeling_swinv2.py:Swinv2PatchMerging": ["False", "LayerNorm", "Linear", "ModelPatchMerging", "Module", "None", "Tensor", "__init__", "batch_size", "bias", "cat", "class", "def", "dim", "forward", "functional", "height", "if", "input_dimensions", "input_feature", "input_feature_0", "input_feature_1", "input_feature_2", "input_feature_3", "input_resolution", "int", "maybe_pad", "nn", "norm", "norm_layer", "num_channels", "or", "pad", "pad_values", "reduction", "return", "self", "shape", "should_pad", "super", "torch", "tuple", "view", "width"], "swinv2/modeling_swinv2.py:Swinv2SelfAttention": ["Dropout", "False", "FloatTensor", "Iterable", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Parameter", "ReLU", "Sequential", "Tensor", "The", "True", "ValueError", "__init__", "a", "abc", "abs", "all_head_size", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "clamp", "class", "collections", "config", "context_layer", "contiguous", "continuous_position_bias_mlp", "coords", "coords_flatten", "coords_h", "coords_w", "def", "dim", "dropout", "dtype", "elif", "else", "exp", "f", "flatten", "float", "forward", "functional", "head_mask", "heads", "hidden", "hidden_states", "if", "ij", "indexing", "inplace", "int", "int64", "is", "isinstance", "key", "key_layer", "log", "log2", "logit_scale", "mask_shape", "math", "matmul", "max", "meshgrid", "multiple", "new_context_layer_shape", "next", "nn", "normalize", "not", "num_attention_heads", "num_channels", "num_heads", "number", "of", "ones", "output_attentions", "outputs", "parameters", "permute", "persistent", "pretrained_window_size", "qkv_bias", "query", "query_layer", "raise", "register_buffer", "relative_coords", "relative_coords_h", "relative_coords_table", "relative_coords_w", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "return", "self", "shape", "sigmoid", "sign", "size", "softmax", "stack", "sum", "super", "the", "to", "torch", "transpose", "tuple", "unsqueeze", "value", "value_layer", "view", "window_size"], "swinv2/modeling_swinv2.py:Swinv2SelfOutput": ["Dropout", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "swinv2/modeling_swinv2.py:Swinv2Attention": ["False", "FloatTensor", "Iterable", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "abc", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "collections", "config", "def", "dense", "dim", "else", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "isinstance", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "pretrained_window_size", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "swinv2/modeling_swinv2.py:Swinv2Intermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "swinv2/modeling_swinv2.py:Swinv2Output": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "swinv2/modeling_swinv2.py:Swinv2Layer": ["False", "FloatTensor", "Identity", "Iterable", "LayerNorm", "ModelAttention", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "_compute_window_shift", "abc", "attention", "attention_output", "attention_outputs", "attention_windows", "attn_mask", "batch_size", "bool", "channels", "class", "collections", "config", "contiguous", "count", "def", "device", "dim", "dims", "drop_path", "drop_path_rate", "dtype", "else", "eps", "for", "forward", "functional", "get_attn_mask", "head_mask", "height", "height_pad", "height_slice", "height_slices", "hidden_states", "hidden_states_windows", "if", "img_mask", "in", "input_dimensions", "input_resolution", "int", "intermediate", "is", "isinstance", "layer_norm_eps", "layer_output", "layer_outputs", "layernorm_after", "layernorm_before", "mask_windows", "masked_fill", "maybe_pad", "nn", "not", "num_heads", "or", "output", "output_attentions", "pad", "pad_bottom", "pad_right", "pad_values", "pretrained_window_size", "r", "return", "roll", "s", "self", "shape", "shift_size", "shifted_hidden_states", "shifted_windows", "shifts", "shortcut", "size", "slice", "super", "target_shift_size", "target_window_size", "to", "torch", "tuple", "unsqueeze", "view", "w", "was_padded", "width", "width_pad", "width_slice", "width_slices", "window_partition", "window_reverse", "window_size", "zeros", "zip"], "swinv2/modeling_swinv2.py:Swinv2Stage": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelLayer", "ModelStage", "ModuleList", "None", "Optional", "Tensor", "__init__", "append", "block", "blocks", "bool", "class", "config", "def", "depth", "dim", "downsample", "drop_path", "drop_path_rate", "else", "enumerate", "for", "forward", "head_mask", "height", "height_downsampled", "hidden_states", "hidden_states_before_downsampling", "i", "if", "in", "input_dimensions", "input_resolution", "int", "is", "layer_head_mask", "layer_module", "layer_outputs", "nn", "norm_layer", "not", "num_heads", "output_attentions", "output_dimensions", "pointing", "pretrained_window_size", "range", "return", "self", "shift_size", "stage_outputs", "super", "torch", "tuple", "width", "width_downsampled", "window_size"], "swinv2/modeling_swinv2.py:Swinv2Encoder": ["False", "FloatTensor", "ModelEncoder", "ModelEncoderOutput", "ModelPatchMerging", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "and", "append", "attentions", "batch_size", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "dim", "downsample", "dpr", "drop_path", "drop_path_rate", "elif", "else", "embed_dim", "enumerate", "for", "forward", "gradient_checkpointing", "grid_size", "head_mask", "hidden_size", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "input_dimensions", "input_resolution", "int", "is", "item", "last_hidden_state", "layer_head_mask", "layer_module", "layer_outputs", "layers", "len", "linspace", "nn", "not", "num_heads", "num_layers", "output_attentions", "output_dimensions", "output_hidden_states", "output_hidden_states_before_downsampling", "permute", "pretrained_window_size", "pretrained_window_sizes", "range", "reshaped_hidden_state", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "stage", "sum", "super", "torch", "tuple", "v", "view", "x"], "swinv2/modeling_swinv2.py:Swinv2PreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelPreTrainedModel", "ModelSelfAttention", "ModelStage", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "log", "logit_scale", "main_input_name", "mask_token", "math", "mean", "module", "nn", "normal_", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "swinv2/modeling_swinv2.py:Swinv2Model": ["AdaptiveAvgPool1d", "BoolTensor", "False", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "flatten", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_dimensions", "int", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "len", "nn", "not", "num_features", "num_layers", "output", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_grid", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "tuple", "use_mask_token", "use_return_dict"], "swinv2/modeling_swinv2.py:Swinv2ForMaskedImageModeling": ["BoolTensor", "Conv2d", "False", "FloatTensor", "Model", "ModelForMaskedImageModeling", "ModelMaskedImageModelingOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "PixelShuffle", "Sequential", "True", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "batch_size", "bool", "bool_masked_pos", "class", "config", "contiguous", "decoder", "def", "else", "embed_dim", "encoder_stride", "floor", "forward", "functional", "head_mask", "height", "hidden_states", "if", "image_size", "in_channels", "int", "interpolate_pos_encoding", "is", "kernel_size", "l1_loss", "loss", "mask", "masked_im_loss", "math", "nn", "none", "not", "num_channels", "num_features", "num_layers", "out_channels", "output", "output_attentions", "output_hidden_states", "outputs", "patch_size", "pixel_values", "post_init", "r", "reconstructed_pixel_values", "reconstruction", "reconstruction_loss", "reduction", "repeat_interleave", "reshape", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_length", "sequence_output", "shape", "size", "sum", "super", "torch", "transpose", "tuple", "unsqueeze", "use_mask_token", "use_return_dict", "width"], "swinv2/modeling_swinv2.py:Swinv2ForImageClassification": ["False", "FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelImageClassifierOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_features", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "post_init", "r", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "swinv2/modeling_swinv2.py:Swinv2Backbone": ["BackboneMixin", "BackboneOutput", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "__init__", "_init_backbone", "attentions", "auto_docstring", "bool", "class", "config", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "get_input_embeddings", "head_mask", "hidden_state", "hidden_states", "i", "if", "in", "input_dimensions", "int", "is", "len", "not", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "output_hidden_states_before_downsampling", "outputs", "patch_embeddings", "patch_grid", "pixel_values", "post_init", "r", "range", "reshaped_hidden_states", "return", "return_dict", "self", "stage", "stage_names", "super", "use_return_dict", "zip"], "rt_detr_v2/modeling_rt_detr_v2.py:multi_scale_deformable_attention_v2": ["False", "Model_scale_deformable_attention_v2", "Tensor", "_", "align_corners", "append", "arange", "attention_weights", "batch_size", "bilinear", "clamp", "concat", "contiguous", "def", "default", "device", "dim", "discrete", "elif", "enumerate", "flatten", "for", "functional", "grid_sample", "height", "hidden_dim", "if", "in", "int", "int64", "level_id", "list", "method", "mode", "nn", "num_heads", "num_levels", "num_points", "num_points_list", "num_queries", "output", "padding_mode", "permute", "repeat", "reshape", "return", "sampling_coord", "sampling_coord_x", "sampling_coord_y", "sampling_grid_l_", "sampling_grids", "sampling_idx", "sampling_locations", "sampling_value_l_", "sampling_value_list", "shape", "split", "stack", "sum", "tensor", "to", "torch", "transpose", "unsqueeze", "value", "value_l_", "value_list", "value_spatial_shapes", "view", "width", "zeros"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2MultiscaleDeformableAttention": ["CUDA", "F", "False", "Last", "Linear", "Make", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "ValueError", "You", "_", "__init__", "a", "align", "and", "attention", "attention_mask", "attention_weights", "authors", "batch_size", "be", "better", "bool", "but", "by", "class", "config", "d", "d_model", "decoder_attention_heads", "decoder_method", "decoder_n_levels", "decoder_n_points", "decoder_offset_scale", "def", "dim", "dim_per_head", "dimension", "divisible", "dtype", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "float", "float32", "for", "forward", "got", "head", "hidden", "hidden_states", "if", "im2col_step", "implementation", "in", "is", "is_torchdynamo_compiling", "length", "level_start_index", "make", "masked_fill", "method", "more", "multi_scale_deformable_attention_v2", "must", "n", "n_heads", "n_levels", "n_points", "n_points_list", "n_points_scale", "nn", "not", "num_heads", "num_queries", "of", "offset", "offset_normalizer", "offset_scale", "or", "output", "output_attentions", "output_proj", "position_embeddings", "power", "raise", "range", "reference_points", "register_buffer", "return", "sampling_locations", "sampling_offsets", "self", "sequence", "sequence_length", "set", "shape", "shapes", "softmax", "spatial", "spatial_shapes", "spatial_shapes_list", "stack", "states", "sum", "super", "sure", "tensor", "the", "to", "torch", "unsqueeze", "value", "value_proj", "view", "warn", "warnings", "which", "with"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2MultiheadAttention": ["Attention", "False", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_reshape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "expand", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "k_proj", "key_states", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2DecoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_activation_function", "decoder_attention_heads", "decoder_ffn_dim", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "eps", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "layer_norm_eps", "level_start_index", "nn", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "second_residual", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_shapes", "spatial_shapes_list", "super", "torch", "training"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2PreTrainedModel": ["BatchNorm2d", "Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelForObjectDetection", "ModelHybridEncoder", "ModelModel", "ModelMultiscaleDeformableAttention", "None", "Parameter", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "abs", "and", "arange", "attention_weights", "base_model_prefix", "bbox_embed", "bias", "class", "class_embed", "config", "constant_", "cos", "data", "def", "default_dtype", "denoising_class_embed", "dtype", "elif", "enc_score_head", "fill_", "float", "for", "get_default_dtype", "grid_init", "hasattr", "i", "if", "in", "init", "initializer_bias_prior_prob", "initializer_range", "int64", "is", "isinstance", "keepdim", "layer", "layers", "learn_initial_query", "log", "main_input_name", "math", "max", "mean", "module", "n_heads", "n_levels", "n_points", "nn", "no_grad", "normal_", "not", "num_denoising", "num_labels", "or", "output_proj", "pi", "pixel_values", "prior_prob", "r", "range", "repeat", "sampling_offsets", "self", "sin", "stack", "std", "thetas", "to", "torch", "value_proj", "view", "weight", "weight_embedding", "with", "xavier_uniform_", "zero_"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2DecoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "rt_detr_v2/modeling_rt_detr_v2.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2Decoder": ["F", "Model", "ModelConfig", "ModelLayer", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "class", "class_embed", "config", "cross_attentions", "d_model", "decoder_layer", "decoder_layers", "def", "detach", "dim", "dropout", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "hidden_states", "idx", "if", "in", "inputs_embeds", "intermediate", "intermediate_hidden_states", "intermediate_logits", "intermediate_reference_points", "inverse_sigmoid", "is", "last_hidden_state", "layer_outputs", "layers", "level_start_index", "logits", "new_reference_points", "nn", "not", "num_layers", "output_attentions", "output_hidden_states", "position_embeddings", "post_init", "predicted_corners", "query_pos_head", "r", "range", "reference_points", "reference_points_input", "return", "return_dict", "self", "sigmoid", "spatial_shapes", "spatial_shapes_list", "stack", "super", "torch", "tuple", "unsqueeze", "use_return_dict", "v", "valid_ratios"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2ModelOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "denoising_meta_values", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "r", "torch", "tuple"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2FrozenBatchNorm2d": ["Model", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "rt_detr_v2/modeling_rt_detr_v2.py:replace_batch_norm": ["BatchNorm2d", "ModelFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2ConvEncoder": ["Model", "Module", "None", "Tensor", "__init__", "append", "backbone", "bool", "channels", "class", "config", "def", "feature_map", "feature_maps", "features", "float", "for", "forward", "freeze_backbone_batch_norms", "functional", "if", "in", "intermediate_channel_sizes", "interpolate", "load_backbone", "mask", "model", "nn", "no_grad", "out", "pixel_mask", "pixel_values", "replace_batch_norm", "return", "self", "shape", "size", "super", "to", "torch", "with"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2ConvNormLayer": ["ACT2CLS", "BatchNorm2d", "Conv2d", "False", "Identity", "Model", "Module", "None", "__init__", "activation", "batch_norm_eps", "bias", "class", "config", "conv", "def", "else", "forward", "hidden_state", "if", "in_channels", "is", "kernel_size", "nn", "norm", "out_channels", "padding", "return", "self", "stride", "super"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2EncoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "def", "dropout", "dtype", "embed_dim", "encoder_activation_function", "encoder_ffn_dim", "encoder_hidden_dim", "eps", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "kwargs", "layer_norm_eps", "max", "min", "nn", "normalize_before", "not", "num_attention_heads", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2RepVggBlock": ["ACT2CLS", "Identity", "Model", "ModelConfig", "ModelConvNormLayer", "Module", "None", "__init__", "activation", "activation_function", "class", "config", "conv1", "conv2", "def", "else", "encoder_hidden_dim", "forward", "hidden_channels", "hidden_expansion", "if", "int", "is", "nn", "padding", "return", "self", "super", "x", "y"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2CSPRepLayer": ["Identity", "Model", "ModelConfig", "ModelConvNormLayer", "ModelRepVggBlock", "Module", "Sequential", "_", "__init__", "activation", "activation_function", "bottlenecks", "class", "config", "conv1", "conv2", "conv3", "def", "else", "encoder_hidden_dim", "for", "forward", "hidden_channels", "hidden_expansion", "hidden_state", "hidden_state_1", "hidden_state_2", "if", "in", "in_channels", "int", "nn", "num_blocks", "out_channels", "range", "return", "self", "super"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2Encoder": ["False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Tensor", "_", "__init__", "attention_mask", "bool", "class", "config", "def", "encoder_layers", "for", "forward", "hidden_states", "in", "layer", "layers", "nn", "output_attentions", "pos_embed", "position_embeddings", "range", "return", "self", "src", "src_mask", "super", "torch"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2HybridEncoder": ["BaseModelOutput", "Embed", "F", "Model", "ModelCSPRepLayer", "ModelConfig", "ModelConvNormLayer", "ModelEncoder", "Module", "ModuleList", "None", "ValueError", "_", "__init__", "activation", "activation_function", "all_attentions", "append", "arange", "attention_mask", "attentions", "backbone_feature_map", "be", "build_2d_sincos_position_embedding", "by", "class", "concat", "config", "contiguous", "cos", "cpu", "def", "device", "dim", "dimension", "divisible", "downsample_conv", "downsample_convs", "downsampled_feature_map", "dtype", "else", "embed_dim", "embedding", "enc_ind", "encode_proj_layers", "encoder", "encoder_hidden_dim", "encoder_in_channels", "encoder_layers", "encoder_states", "enumerate", "eval_size", "feat_strides", "flatten", "float32", "for", "forward", "fpn_block", "fpn_blocks", "fpn_feature_map", "fpn_feature_maps", "fused_feature_map", "grid_h", "grid_w", "height", "hidden_states", "i", "idx", "if", "ij", "in", "in_channels", "indexing", "inputs_embeds", "interpolate", "is", "kernel_size", "last_hidden_state", "lateral_conv", "lateral_convs", "layer_outputs", "len", "level_start_index", "meshgrid", "mode", "must", "nearest", "new_fpn_feature_map", "new_pan_feature_map", "nn", "not", "num_fpn_stages", "num_pan_stages", "omega", "or", "out_channels", "out_h", "out_strides", "out_w", "output_attentions", "output_hidden_states", "pan_block", "pan_blocks", "pan_feature_maps", "permute", "pos_dim", "pos_embed", "position", "position_embeddings", "positional_encoding_temperature", "r", "raise", "range", "reshape", "return", "return_dict", "scale_factor", "self", "shape", "sin", "spatial_shapes", "src_flatten", "staticmethod", "stride", "super", "temperature", "to", "top_fpn_feature_map", "top_pan_feature_map", "torch", "torch_int", "training", "tuple", "use_return_dict", "v", "valid_ratios", "width", "zip"], "rt_detr_v2/modeling_rt_detr_v2.py:get_contrastive_denoising_training_group": ["Model_contrastive_denoising_training_group", "None", "attn_mask", "batch_size", "bool", "box_noise_scale", "boxes", "center_to_corners_format", "class_embed", "class_labels", "clip_", "corners_to_center_format", "def", "denoise_positive_idx", "denoising_meta_values", "device", "diff", "dn_num_group", "dn_num_split", "dn_positive_idx", "dtype", "else", "float", "for", "full", "i", "idx_block_end", "idx_block_start", "if", "in", "inf", "input_query_bbox", "input_query_class", "int32", "inverse_sigmoid", "known_bbox", "label_noise_ratio", "len", "mask", "max", "max_gt_num", "min", "n", "negative_gt_mask", "new_label", "nonzero", "num_classes", "num_denoising_queries", "num_ground_truths", "num_groups_denoising_queries", "num_gt", "num_queries", "pad_gt_mask", "positive_gt_mask", "rand_like", "rand_part", "rand_sign", "randint_like", "range", "return", "split", "squeeze", "t", "tarModel_size", "tarModels", "tile", "torch", "torch_int", "where", "zeros"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2Model": ["BaseModelOutput", "BatchNorm2d", "Conv2d", "Embedding", "F", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelConvEncoder", "ModelDecoder", "ModelHybridEncoder", "ModelMLPPredictionHead", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Sequential", "True", "Union", "_", "__init__", "_len_sources", "all", "anchor_image_size", "anchors", "and", "append", "arange", "attention_mask", "attentions", "auto_docstring", "backbone", "batch_norm_eps", "batch_size", "bias", "bool", "box_noise_scale", "cat", "class", "class_embed", "compile_compatible_method_lru_cache", "concat", "config", "cpu", "cross_attentions", "cumsum", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_in_channels", "decoder_input_proj", "decoder_input_proj_list", "decoder_inputs_embeds", "decoder_outputs", "def", "denoising_bbox_unact", "denoising_class", "denoising_class_embed", "denoising_meta_values", "detach", "device", "dict", "dim", "dn_outputs", "dtype", "elif", "else", "empty", "enc_bbox_head", "enc_output", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "enc_score_head", "enc_topk_bboxes", "enc_topk_logits", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_dim", "encoder_hidden_states", "encoder_input_proj", "encoder_input_proj_list", "encoder_last_hidden_state", "encoder_outputs", "end", "enumerate", "eps", "feat_strides", "features", "finfo", "flatten", "float32", "for", "forward", "freeze_backbone", "gather", "generate_anchors", "get_contrastive_denoising_training_group", "get_encoder", "grid_size", "grid_x", "grid_xy", "grid_y", "height", "hidden_states", "i", "if", "ij", "in", "in_channels", "index", "indexing", "init_reference_points", "initial_reference_points", "inputs_embeds", "int", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "is", "isinstance", "keepdim", "kernel_size", "label_noise_ratio", "labels", "last_hidden_state", "layer_norm_eps", "learn_initial_query", "len", "level", "level_start_index", "list", "log", "long", "mask", "max", "maxsize", "memory", "meshgrid", "new_zeros", "nn", "not", "num_backbone_outs", "num_channels", "num_classes", "num_denoising", "num_denoising_queries", "num_feature_levels", "num_labels", "num_layers", "num_queries", "ones", "ones_like", "or", "output_attentions", "output_hidden_states", "output_memory", "padding", "padding_idx", "param", "parameters", "pixel_mask", "pixel_values", "post_init", "prod", "proj_feats", "r", "range", "reference_points", "reference_points_unact", "repeat", "requires_grad_", "reshape", "return", "return_dict", "s", "self", "shape", "sigmoid", "source", "source_flatten", "sources", "spatial_shapes", "spatial_shapes_list", "spatial_shapes_tuple", "stack", "stride", "super", "target", "targets", "tensor", "tile", "to", "topk", "topk_ind", "torch", "training", "transpose", "tuple", "tuple_outputs", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_mask", "value", "values", "weight_embedding", "wh", "where", "width"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2MLPPredictionHead": ["Linear", "Model", "Module", "ModuleList", "__init__", "class", "config", "d_model", "def", "else", "enumerate", "for", "forward", "functional", "h", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2ObjectDetectionOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "denoising_meta_values", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "initial_reference_points", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "rt_detr_v2/modeling_rt_detr_v2.py:RTDetrV2ForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "Union", "_", "__init__", "_no_split_modules", "_set_aux_loss", "_tied_weights_keys", "a", "auto_docstring", "auxiliary_outputs", "b", "bbox_embed", "bool", "class", "class_embed", "config", "cross_attentions", "d_model", "decoder", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_layers", "def", "denoising_meta_values", "device", "dict", "else", "enc_outputs_class", "enc_outputs_coord_logits", "enc_topk_bboxes", "enc_topk_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "for", "forward", "if", "in", "init_reference_points", "initial_reference_points", "inputs_embeds", "intermediate_hidden_states", "intermediate_logits", "intermediate_predicted_corners", "intermediate_reference_points", "is", "jit", "kwargs", "labels", "last_hidden_state", "list", "logits", "loss", "loss_dict", "loss_function", "model", "nn", "not", "num_labels", "num_layers", "output", "output_attentions", "output_hidden_states", "outputs", "outputs_class", "outputs_coord", "partial", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "predicted_corners", "r", "range", "return", "return_dict", "self", "super", "torch", "training", "tuple", "unused", "use_return_dict", "zip"], "ijepa/modeling_ijepa.py:IJepaPatchEmbeddings": ["Conv2d", "Expected", "False", "Input", "Iterable", "Make", "Model", "ModelConfig", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "bool", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "flatten", "forward", "got", "height", "hidden_size", "if", "image", "image_size", "in", "interpolate_pos_encoding", "isinstance", "kernel_size", "match", "model", "nn", "not", "num_channels", "num_patches", "of", "one", "or", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "torch", "transpose", "values", "width", "with"], "ijepa/modeling_ijepa.py:IJepaEmbeddings": ["BoolTensor", "Dropout", "False", "Model", "ModelConfig", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "class", "config", "def", "dim", "dropout", "else", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "randn", "reshape", "return", "self", "seq_length", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_mask_token", "view", "width", "zeros"], "ijepa/modeling_ijepa.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "ijepa/modeling_ijepa.py:IJepaSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "ijepa/modeling_ijepa.py:IJepaSelfOutput": ["Dropout", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "ijepa/modeling_ijepa.py:IJepaAttention": ["Model", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "ijepa/modeling_ijepa.py:IJepaIntermediate": ["ACT2FN", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "ijepa/modeling_ijepa.py:IJepaOutput": ["Dropout", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "ijepa/modeling_ijepa.py:IJepaLayer": ["GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "ijepa/modeling_ijepa.py:IJepaPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "dtype", "elif", "fill_", "float32", "hidden_states", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "ijepa/modeling_ijepa.py:IJepaEncoder": ["BaseModelOutput", "False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "ijepa/modeling_ijepa.py:IJepaPooler": ["ACT2FN", "Linear", "Model", "ModelConfig", "Module", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "pooler_act", "pooler_output_size", "return", "self", "super", "torch"], "ijepa/modeling_ijepa.py:IJepaModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "BoolTensor", "False", "LayerNorm", "Model", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelPatchEmbeddings", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dict", "dtype", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "expected_dtype", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "int", "interpolate_pos_encoding", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "projection", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch", "use_mask_token", "weight"], "ijepa/modeling_ijepa.py:IJepaForImageClassification": ["BaseModelOutputWithPooling", "False", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "dim", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch"], "mbart/modeling_mbart.py:shift_tokens_right": ["Model_tokens_right", "None", "Tensor", "ValueError", "be", "clone", "config", "decoder_start_tokens", "def", "defined", "dim", "gather", "has", "if", "index_of_eos", "input_ids", "int", "is", "masked_fill_", "model", "ne", "pad_token_id", "prev_output_tokens", "raise", "return", "self", "squeeze", "sum", "to", "torch", "unsqueeze"], "mbart/modeling_mbart.py:MBartLearnedPositionalEmbedding": ["Embedding", "Model", "None", "Optional", "Tensor", "__init__", "arange", "bsz", "class", "def", "device", "dtype", "else", "embedding_dim", "expand", "forward", "if", "input_ids", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "seq_len", "shape", "super", "torch", "unsqueeze", "weight"], "mbart/modeling_mbart.py:MBartScaledWordEmbedding": ["Embedding", "Model", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "mbart/modeling_mbart.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "mbart/modeling_mbart.py:MBartAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "Model", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "mbart/modeling_mbart.py:MBartEncoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "functional", "hidden_states", "if", "layer_head_mask", "max", "min", "nn", "num_heads", "output_attentions", "p", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch", "training"], "mbart/modeling_mbart.py:MBartDecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "cache_position", "class", "config", "cross_attn_layer_head_mask", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "int", "is", "is_causal", "is_decoder", "key_value_states", "layer_head_mask", "layer_idx", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "mbart/modeling_mbart.py:MBartClassificationHead": ["Dropout", "Linear", "Model", "Module", "Tensor", "__init__", "class", "def", "dense", "dropout", "float", "forward", "hidden_states", "inner_dim", "input_dim", "int", "nn", "num_classes", "out_proj", "p", "pooler_dropout", "return", "self", "super", "tanh", "torch"], "mbart/modeling_mbart.py:MBartPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "Embedding", "False", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "None", "Optional", "PreTrainedModel", "Size", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_init_weights", "_no_split_modules", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "_update_cross_attn_mask", "_update_full_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "data", "def", "device", "diagonal", "dim", "dtype", "dummy_inputs", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "expand", "fill_", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "init_std", "input_ids", "input_shape", "input_tensor", "inputs_embeds", "int", "is", "is_causal", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "mean", "min", "min_dtype", "model", "module", "ne", "nn", "normal_", "not", "npu", "ones", "pad_token", "pad_token_id", "padding_idx", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "property", "query_length", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "std", "supports_gradient_checkpointing", "target_length", "tensor", "tgt_len", "to", "torch", "training", "triu", "type", "using_compilable_cache", "weight", "xpu", "zero_"], "mbart/modeling_mbart.py:MBartEncoder": ["BaseModelOutput", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Tensor", "The", "True", "Union", "ValueError", "You", "_", "__init__", "_backward_compatibility_gradient_checkpointing", "_update_full_mask", "all_attentions", "and", "at", "attention_mask", "attentions", "be", "bool", "both", "but", "cannot", "class", "config", "d_model", "def", "device", "dropout", "dropout_probability", "either", "elif", "else", "embed_dim", "embed_pos", "embed_positions", "embed_scale", "embed_tokens", "encoder_layer", "encoder_layerdrop", "encoder_layers", "encoder_states", "enumerate", "f", "for", "forward", "functional", "getattr", "gradient_checkpointing", "gradient_checkpointing_enable", "have", "head_mask", "hidden_states", "idx", "if", "in", "input", "input_ids", "input_shape", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_norm", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "math", "max_position_embeddings", "max_source_positions", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "shape", "should", "size", "specified", "specify", "sqrt", "super", "supports_gradient_checkpointing", "the", "time", "to", "to_drop", "torch", "training", "tuple", "use_return_dict", "v", "view", "vocab_size", "weight"], "mbart/modeling_mbart.py:MBartDecoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "Embedding", "EncoderDecoderCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "ModelConfig", "ModelLayer", "ModelLearnedPositionalEmbedding", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "__init__", "_update_causal_mask", "_update_cross_attn_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "both", "but", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "deprecated", "device", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "is_torchdynamo_compiling", "isinstance", "it", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layernorm_embedding", "layers", "len", "logger", "mask_name", "mask_seq_length", "math", "max_position_embeddings", "max_target_positions", "nn", "not", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "post_init", "r", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attention_cache", "self_attn_cache", "seq_length", "shape", "should", "size", "specified", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "weight", "will", "with", "zip"], "mbart/modeling_mbart.py:MBartModel": ["BaseModelOutput", "Cache", "FloatTensor", "LongTensor", "Model", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "None", "Optional", "Seq2SeqModelOutput", "Tensor", "Union", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_outputs", "def", "elif", "else", "embed_scale", "embed_tokens", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "isinstance", "last_hidden_state", "len", "math", "not", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "return", "return_dict", "scale_embedding", "self", "set_input_embeddings", "shared", "shift_tokens_right", "sqrt", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight"], "mbart/modeling_mbart.py:MBartForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqLMOutput", "Tensor", "The", "True", "Union", "__init__", "_keys_to_ignore_on_load_missing", "_resize_final_logits_bias", "_tied_weights_keys", "and", "argument", "attention_mask", "auto_docstring", "base_model_prefix", "bias", "bool", "cache_position", "cat", "changed", "class", "config", "cross_attentions", "cross_attn_head_mask", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "extra_bias", "final_logits_bias", "forward", "get_decoder", "get_encoder", "head_mask", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "lm_head", "lm_logits", "logger", "logits", "loss", "loss_fct", "masked_lm_loss", "mean_resizing", "model", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_embeddings", "old_num_tokens", "output", "output_attentions", "output_hidden_states", "outputs", "pad_to_multiple_of", "pad_token_id", "past_key_values", "post_init", "prepare_decoder_input_ids_from_labels", "provided", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "self", "shape", "shared", "shift_tokens_right", "since", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "warning", "weight", "zeros"], "mbart/modeling_mbart.py:MBartForSequenceClassification": ["All", "BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Passing", "Seq2SeqSequenceClassifierOutput", "Tensor", "Union", "ValueError", "__class__", "__init__", "__name__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "class", "classification_head", "classifier_dropout", "config", "cross_attentions", "cross_attn_head_mask", "currently", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "device", "dtype", "elif", "else", "embed_tokens", "embeddings", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eos", "eos_mask", "eos_token_id", "eq", "examples", "f", "for", "forward", "have", "head_mask", "hidden_states", "if", "input", "input_ids", "inputs_embeds", "int", "is", "kwargs", "labels", "len", "list", "logits", "long", "loss", "loss_fct", "model", "multi_label_classification", "must", "not", "num_labels", "number", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "same", "self", "sentence_representation", "single_label_classification", "size", "squeeze", "sum", "super", "supported", "the", "to", "tokens", "torch", "tuple", "unique_consecutive", "use_cache", "use_return_dict", "view", "weight"], "mbart/modeling_mbart.py:MBartForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Seq2SeqQuestionAnsweringModelOutput", "Tensor", "Union", "__init__", "_tied_weights_keys", "and", "attention_mask", "auto_docstring", "bool", "cache_position", "clamp", "class", "config", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "def", "dim", "else", "embed_tokens", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "end_logits", "end_loss", "end_positions", "forward", "head_mask", "hidden_size", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "list", "logits", "loss", "loss_fct", "model", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_cache", "use_return_dict", "weight"], "mbart/modeling_mbart.py:MBartDecoderWrapper": ["Model", "ModelDecoder", "ModelPreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "mbart/modeling_mbart.py:MBartForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelDecoderWrapper", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "forward", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "labels", "lm_head", "logits", "loss", "loss_fct", "model", "nn", "not", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "post_init", "r", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "beit/modeling_beit.py:BeitModelOutputWithPooling": ["BaseModelOutputWithPooling", "ModelModelOutputWithPooling", "class", "r"], "beit/modeling_beit.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "beit/modeling_beit.py:BeitDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "beit/modeling_beit.py:BeitEmbeddings": ["BoolTensor", "Dropout", "False", "Iterable", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "The", "_", "__init__", "abc", "align_corners", "always", "and", "are", "argument", "batch_size", "be", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "collections", "config", "def", "dim", "dropout", "effect", "else", "embeddings", "expand", "for", "forward", "functional", "has", "height", "hidden_dropout_prob", "hidden_size", "if", "image", "image_size", "in", "input", "int", "interpolate", "interpolate_pos_encoding", "interpolated", "is", "is_tracing", "isinstance", "jit", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "no", "not", "num_patches", "num_positions", "patch_embeddings", "patch_height", "patch_pos_embed", "patch_size", "patch_width", "permute", "pixel_values", "position_embeddings", "removed", "reshape", "return", "self", "seq_len", "shape", "size", "sqrt_num_positions", "super", "the", "to", "torch", "torch_int", "transformers", "type_as", "unsqueeze", "use_absolute_position_embeddings", "use_mask_token", "v4", "view", "w", "warn", "warnings", "width", "will", "zeros"], "beit/modeling_beit.py:BeitPatchEmbeddings": ["Conv2d", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "flatten", "forward", "height", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_height", "patch_shape", "patch_size", "patch_width", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "width", "with"], "beit/modeling_beit.py:BeitSelfAttention": ["Dropout", "False", "Linear", "ModelConfig", "ModelRelativePositionBias", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "Union", "ValueError", "_", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch_size", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dim_size", "dropout", "else", "embedding_size", "f", "forward", "functional", "has_relative_position_bias", "hasattr", "head_mask", "heads", "height", "hidden", "hidden_size", "hidden_states", "if", "int", "interpolate_pos_encoding", "is", "key", "key_layer", "math", "matmul", "multiple", "new_context_layer_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "patch_size", "permute", "query", "query_layer", "raise", "relative_position_bias", "resolution", "return", "self", "seq_length", "shape", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "tuple", "value", "value_layer", "view", "width", "window_size"], "beit/modeling_beit.py:BeitSdpaSelfAttention": ["Falling", "False", "ModelSdpaSelfAttention", "ModelSelfAttention", "None", "Optional", "Tensor", "This", "Transformers", "True", "Union", "_", "all_head_size", "argument", "attention", "attention_head_size", "attention_probs_dropout_prob", "attn_bias", "attn_implementation", "attn_mask", "back", "batch_size", "be", "bool", "but", "can", "class", "config", "context_layer", "contiguous", "def", "dim_size", "does", "dropout_p", "eager", "else", "forward", "from", "functional", "has_relative_position_bias", "head_mask", "height", "hidden_states", "if", "implementation", "int", "interpolate_pos_encoding", "is", "is_causal", "key", "key_layer", "loading", "logger", "manual", "math", "model", "new_context_layer_shape", "nn", "not", "num_attention_heads", "onwards", "or", "output_attentions", "patch_size", "permute", "query", "query_layer", "relative_position_bias", "removed", "required", "resolution", "return", "scale", "scaled_dot_product_attention", "scaling", "self", "seq_length", "shape", "size", "specifying", "sqrt", "super", "support", "the", "to", "torch", "training", "transpose", "tuple", "used", "using", "v5", "value", "value_layer", "version", "view", "warning", "warning_once", "when", "width", "will", "window_size"], "beit/modeling_beit.py:BeitSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "gamma", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "beit/modeling_beit.py:BeitAttention": ["False", "ModelAttention", "ModelConfig", "ModelSelfOutput", "Model_SELF_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_attn_implementation", "all_head_size", "attention", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "interpolate_pos_encoding", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "relative_position_bias", "resolution", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "window_size"], "beit/modeling_beit.py:BeitIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "beit/modeling_beit.py:BeitOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "beit/modeling_beit.py:BeitLayer": ["False", "GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "attention", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "float", "forward", "head_mask", "hidden_size", "hidden_states", "if", "init_values", "int", "intermediate", "interpolate_pos_encoding", "is", "lambda_1", "lambda_2", "layer_norm_eps", "layer_output", "layer_scale_init_value", "layernorm_after", "layernorm_before", "nn", "not", "ones", "output", "output_attentions", "outputs", "relative_position_bias", "requires_grad", "resolution", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple", "window_size"], "beit/modeling_beit.py:BeitRelativePositionBias": ["False", "ModelConfig", "ModelRelativePositionBias", "Module", "None", "Parameter", "Tensor", "__init__", "align_corners", "arange", "bilinear", "bool", "cat", "class", "compile_compatible_method_lru_cache", "config", "contiguous", "coords", "coords_flatten", "def", "dim_size", "dtype", "flatten", "forward", "functional", "generate_relative_position_index", "grid", "if", "ij", "indexing", "int", "interpolate", "interpolate_pos_encoding", "maxsize", "meshgrid", "mode", "new_height", "new_num_relative_distance", "new_relative_position_bias_table", "new_sub_table", "new_width", "nn", "num_attention_heads", "num_relative_distance", "old_height", "old_num_relative_distance", "old_relative_position_bias_table", "old_sub_table", "old_width", "permute", "relative_coords", "relative_position_bias", "relative_position_bias_table", "relative_position_index", "reshape", "return", "self", "size", "squeeze", "stack", "sum", "super", "torch", "torch_int", "tuple", "unsqueeze", "view", "window_area", "window_size", "zeros"], "beit/modeling_beit.py:BeitEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "ModelRelativePositionBias", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "bool", "class", "config", "cpu", "def", "device", "dim_size", "dpr", "drop_path_rate", "else", "enumerate", "for", "forward", "gradient_checkpointing", "has_relative_position_bias", "head_mask", "height", "hidden_states", "i", "if", "in", "int", "interpolate_pos_encoding", "is", "item", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "linspace", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_size", "range", "relative_position_bias", "resolution", "return", "return_dict", "self", "shape", "super", "torch", "tuple", "use_relative_position_bias", "use_shared_relative_position_bias", "v", "width", "window_size", "x"], "beit/modeling_beit.py:BeitPreTrainedModel": ["Conv2d", "ConvTranspose2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelRelativePositionBias", "None", "PreTrainedModel", "True", "_init_weights", "_keys_to_ignore_on_load_unexpected", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "lambda_1", "lambda_2", "layer_scale_init_value", "main_input_name", "mask_token", "mean", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "position_embeddings", "r", "relative_position_bias_table", "relative_position_index", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "beit/modeling_beit.py:BeitModel": ["BoolTensor", "False", "Identity", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutputWithPooling", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "_prune_heads", "add_pooling_layer", "attention", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "head_mask", "head_outputs", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "patch_embeddings", "patch_shape", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "resolution", "return", "return_dict", "self", "sequence_output", "shape", "super", "torch", "tuple", "use_mean_pooling", "use_return_dict", "window_size"], "beit/modeling_beit.py:BeitPooler": ["LayerNorm", "ModelConfig", "ModelPooler", "Module", "None", "Tensor", "__init__", "class", "config", "def", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "is", "layer_norm_eps", "layernorm", "mean", "nn", "not", "patch_tokens", "pooled_output", "return", "self", "super", "torch", "use_mean_pooling"], "beit/modeling_beit.py:BeitForMaskedImageModeling": ["BoolTensor", "CrossEntropyLoss", "False", "LayerNorm", "Linear", "MaskedLMOutput", "Model", "ModelConfig", "ModelForMaskedImageModeling", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "bool_masked_pos", "class", "config", "def", "else", "eps", "forward", "get_output_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "layer_norm_eps", "layernorm", "lm_head", "logits", "loss", "loss_fct", "masked_lm_loss", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict", "vocab_size"], "beit/modeling_beit.py:BeitForImageClassification": ["False", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "beit/modeling_beit.py:BeitConvModule": ["BatchNorm2d", "Conv2d", "False", "ModelConvModule", "Module", "None", "ReLU", "Tensor", "Union", "__init__", "activation", "bias", "bn", "bool", "class", "conv", "def", "dilation", "forward", "in_channels", "input", "int", "kernel_size", "nn", "out_channels", "output", "padding", "return", "self", "str", "super", "torch", "tuple"], "beit/modeling_beit.py:BeitPyramidPoolingBlock": ["AdaptiveAvgPool2d", "ModelConvModule", "ModelPyramidPoolingBlock", "Module", "None", "Tensor", "__init__", "add_module", "channels", "class", "def", "enumerate", "for", "forward", "hidden_state", "i", "in", "in_channels", "input", "int", "kernel_size", "layer", "layers", "nn", "pool_scale", "return", "self", "str", "super", "torch"], "beit/modeling_beit.py:BeitPyramidPoolingModule": ["ModelPyramidPoolingBlock", "ModelPyramidPoolingModule", "Module", "None", "Tensor", "__init__", "add_module", "align_corners", "append", "bilinear", "block", "blocks", "bool", "channels", "class", "def", "enumerate", "for", "forward", "functional", "i", "in", "in_channels", "int", "interpolate", "list", "mode", "nn", "pool_scale", "pool_scales", "ppm", "ppm_out", "ppm_outs", "return", "self", "size", "str", "super", "torch", "tuple", "upsampled_ppm_out", "x"], "beit/modeling_beit.py:BeitUperHead": ["Conv2d", "False", "ModelConfig", "ModelConvModule", "ModelPyramidPoolingModule", "ModelUperHead", "Module", "ModuleList", "None", "Tensor", "__init__", "align_corners", "append", "bilinear", "bottleneck", "cat", "channels", "class", "classifier", "config", "def", "dim", "encoder_hidden_states", "enumerate", "extend", "for", "forward", "fpn_bottleneck", "fpn_conv", "fpn_convs", "fpn_outs", "functional", "hidden_size", "i", "in", "in_channels", "inputs", "interpolate", "kernel_size", "l_conv", "lateral_conv", "lateral_convs", "laterals", "len", "mode", "nn", "num_labels", "output", "padding", "pool_scales", "prev_shape", "psp_forward", "psp_modules", "psp_outs", "range", "return", "self", "shape", "size", "super", "torch", "used_backbone_levels", "x"], "beit/modeling_beit.py:BeitFCNHead": ["Conv2d", "Identity", "ModelConfig", "ModelConvModule", "ModelFCNHead", "Module", "None", "Sequential", "Tensor", "Union", "__init__", "append", "auxiliary_channels", "auxiliary_concat_input", "auxiliary_num_convs", "cat", "channels", "class", "classifier", "concat_input", "config", "conv_cat", "conv_padding", "convs", "def", "dilation", "dim", "else", "encoder_hidden_states", "for", "forward", "hidden_size", "hidden_states", "i", "if", "in", "in_channels", "in_index", "int", "kernel_size", "nn", "num_convs", "num_labels", "output", "padding", "range", "return", "self", "super", "torch", "tuple"], "beit/modeling_beit.py:BeitForSemanticSegmentation": ["BatchNorm2d", "ConvTranspose2d", "CrossEntropyLoss", "False", "GELU", "Identity", "MaxPool2d", "Model", "ModelConfig", "ModelFCNHead", "ModelForSemanticSegmentation", "ModelModel", "ModelPreTrainedModel", "ModelUperHead", "None", "One", "Optional", "SemanticSegmenterOutput", "Sequential", "Tensor", "The", "True", "Union", "ValueError", "__init__", "a", "add_pooling_layer", "align_corners", "and", "architecture", "attentions", "auto_docstring", "auxiliary_head", "auxiliary_logits", "auxiliary_loss", "auxiliary_loss_weight", "backbone", "base", "batch_size", "be", "bilinear", "bool", "can", "case", "class", "compute_loss", "config", "decode_head", "def", "else", "encoder_hidden_states", "enumerate", "feature", "features", "for", "forward", "fpn1", "fpn2", "fpn3", "fpn4", "from", "functional", "greater", "head_mask", "hidden_size", "hidden_states", "i", "idx", "if", "ignore_index", "image_size", "in", "integers", "interpolate", "interpolate_pos_encoding", "is", "kernel_size", "labels", "len", "list", "logits", "loss", "loss_fct", "main_loss", "mode", "nn", "not", "num_labels", "number", "of", "one", "ops", "out_indices", "output", "output_attentions", "output_hidden_states", "outputs", "patch_resolution", "patch_size", "permute", "pixel_values", "post_init", "r", "raise", "range", "requires", "reshape", "return", "return_dict", "self", "semantic_loss_ignore_index", "shape", "should", "size", "sized", "specifying", "stride", "super", "than", "the", "to", "torch", "tuple", "upsampled_auxiliary_logits", "upsampled_logits", "use", "use_auxiliary_head", "use_return_dict", "which", "x"], "beit/modeling_beit.py:BeitBackbone": ["BackboneMixin", "BackboneOutput", "BatchNorm2d", "ConvTranspose2d", "GELU", "Identity", "MaxPool2d", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "None", "One", "Optional", "Sequential", "Tensor", "True", "ValueError", "_", "__init__", "_init_backbone", "a", "add_fpn", "architecture", "attentions", "auto_docstring", "backbone", "base", "batch_norm_eps", "batch_size", "be", "bool", "can", "case", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "eps", "feature_maps", "features", "for", "forward", "fpn1", "fpn2", "fpn3", "fpn4", "from", "get_input_embeddings", "hidden_size", "hidden_state", "hidden_states", "if", "in", "integers", "is", "kernel_size", "len", "list", "nn", "not", "num_features", "num_hidden_layers", "of", "out_features", "out_indices", "output", "output_attentions", "output_hidden_states", "outputs", "patch_embeddings", "patch_height", "patch_shape", "patch_width", "permute", "pixel_values", "post_init", "r", "raise", "range", "requires", "reshape", "reshape_hidden_states", "resolution", "return", "return_dict", "self", "shape", "sized", "specifying", "stage", "stage_names", "stride", "super", "the", "to", "tuple", "use", "use_return_dict", "which", "window_size", "zip"], "align/modeling_align.py:AlignVisionModelOutput": ["FloatTensor", "ModelOutput", "ModelVisionModelOutput", "None", "Optional", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "align/modeling_align.py:AlignTextModelOutput": ["FloatTensor", "ModelOutput", "ModelTextModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "align/modeling_align.py:AlignOutput": ["Any", "BaseModelOutputWithPooling", "BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "align/modeling_align.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "label_smoothing", "len", "logits", "nn", "return", "torch"], "align/modeling_align.py:align_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "align/modeling_align.py:round_filters": ["ModelVisionConfig", "Model_filters", "config", "def", "depth_divisor", "divisor", "if", "int", "max", "new_dim", "num_channels", "r", "return", "width_coefficient"], "align/modeling_align.py:correct_pad": ["Model", "Model_pad", "True", "Union", "adjust", "bool", "def", "else", "if", "int", "isinstance", "kernel_size", "r", "return", "tuple"], "align/modeling_align.py:AlignVisionEmbeddings": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "ZeroPad2d", "__init__", "activation", "batch_norm_eps", "batch_norm_momentum", "batchnorm", "bias", "class", "config", "convolution", "def", "eps", "features", "forward", "hidden_act", "kernel_size", "momentum", "nn", "num_channels", "out_dim", "padding", "pixel_values", "r", "return", "round_filters", "self", "stride", "super", "torch", "valid"], "align/modeling_align.py:AlignVisionDepthwiseConv2d": ["Conv2d", "ModelVisionDepthwiseConv2d", "True", "__init__", "bias", "class", "def", "depth_multiplier", "dilation", "groups", "in_channels", "kernel_size", "nn", "out_channels", "padding", "padding_mode", "self", "stride", "super", "zeros"], "align/modeling_align.py:AlignVisionExpansionLayer": ["ACT2FN", "BatchNorm2d", "Conv2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionExpansionLayer", "Module", "Tensor", "__init__", "batch_norm_eps", "bias", "class", "config", "def", "eps", "expand_act", "expand_bn", "expand_conv", "forward", "hidden_act", "hidden_states", "in_channels", "in_dim", "int", "kernel_size", "nn", "num_features", "out_channels", "out_dim", "padding", "r", "return", "same", "self", "stride", "super", "torch"], "align/modeling_align.py:AlignVisionDepthwiseLayer": ["ACT2FN", "BatchNorm2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionDepthwiseConv2d", "ModelVisionDepthwiseLayer", "Module", "Tensor", "ZeroPad2d", "__init__", "adjust", "adjust_padding", "batch_norm_eps", "batch_norm_momentum", "bias", "bool", "class", "config", "conv_pad", "correct_pad", "def", "depthwise_act", "depthwise_conv", "depthwise_conv_pad", "depthwise_norm", "else", "eps", "forward", "hidden_act", "hidden_states", "if", "in_dim", "int", "kernel_size", "momentum", "nn", "num_features", "padding", "r", "return", "same", "self", "stride", "super", "torch", "valid"], "align/modeling_align.py:AlignVisionSqueezeExciteLayer": ["ACT2FN", "AdaptiveAvgPool2d", "Conv2d", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionSqueezeExciteLayer", "Module", "Sigmoid", "Tensor", "__init__", "act_expand", "act_reduce", "bool", "class", "config", "def", "dim", "dim_se", "else", "expand", "expand_dim", "forward", "hidden_act", "hidden_states", "if", "in_channels", "in_dim", "inputs", "int", "kernel_size", "max", "mul", "nn", "out_channels", "output_size", "padding", "r", "reduce", "return", "same", "self", "squeeze", "squeeze_expansion_ratio", "super", "torch"], "align/modeling_align.py:AlignVisionFinalBlockLayer": ["BatchNorm2d", "Conv2d", "Dropout", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionFinalBlockLayer", "Module", "Tensor", "__init__", "and", "apply_dropout", "batch_norm_eps", "batch_norm_momentum", "bias", "bool", "class", "config", "def", "drop_rate", "dropout", "embeddings", "eps", "float", "forward", "hidden_states", "id_skip", "if", "in_channels", "in_dim", "int", "kernel_size", "momentum", "nn", "not", "num_features", "out_channels", "out_dim", "p", "padding", "project_bn", "project_conv", "r", "return", "same", "self", "stride", "super", "torch"], "align/modeling_align.py:AlignVisionBlock": ["FloatTensor", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionDepthwiseLayer", "ModelVisionExpansionLayer", "ModelVisionFinalBlockLayer", "ModelVisionSqueezeExciteLayer", "Module", "Tensor", "__init__", "adjust_padding", "bool", "class", "config", "def", "depthwise_conv", "drop_rate", "else", "embeddings", "expand", "expand_dim", "expand_in_dim", "expand_ratio", "expansion", "float", "forward", "hidden_states", "id_skip", "if", "in_dim", "int", "kernel_size", "nn", "out_dim", "projection", "r", "return", "self", "squeeze_excite", "stride", "super", "torch"], "align/modeling_align.py:AlignVisionEncoder": ["BaseModelOutputWithNoAttention", "BaseModelOutputWithPoolingAndNoAttention", "False", "FloatTensor", "ModelVisionBlock", "ModelVisionConfig", "ModelVisionEncoder", "Module", "ModuleList", "None", "Optional", "True", "__init__", "adjust_padding", "all_hidden_states", "append", "block", "blocks", "bool", "ceil", "class", "config", "curr_block_num", "def", "depth_coefficient", "depthwise_padding", "drop_connect_rate", "drop_rate", "else", "expand_ratio", "expand_ratios", "for", "forward", "hidden_states", "i", "id_skip", "if", "in", "in_channels", "in_dim", "int", "is", "j", "kernel_size", "kernel_sizes", "last_hidden_state", "len", "math", "n", "nn", "not", "num_base_blocks", "num_block_repeats", "num_blocks", "out_channels", "out_dim", "output_hidden_states", "r", "range", "repeats", "return", "return_dict", "round_filters", "round_repeats", "self", "stride", "strides", "sum", "super", "torch", "tuple", "v"], "align/modeling_align.py:AlignTextEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "align/modeling_align.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "view"], "align/modeling_align.py:AlignTextSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "Dropout", "False", "FloatTensor", "Linear", "ModelTextSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_dropout", "attention_head_size", "attention_interface", "attention_mask", "attention_probs_dropout_prob", "attn_output", "attn_weights", "bool", "class", "config", "contiguous", "def", "dropout", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "key", "key_states", "kwargs", "multiple", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "query", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_states", "view"], "align/modeling_align.py:AlignTextSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "align/modeling_align.py:AlignTextAttention": ["False", "FloatTensor", "ModelTextAttention", "ModelTextSelfAttention", "ModelTextSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "kwargs", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "align/modeling_align.py:AlignTextIntermediate": ["ACT2FN", "Linear", "ModelTextIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "align/modeling_align.py:AlignTextOutput": ["Dropout", "LayerNorm", "Linear", "ModelTextOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "align/modeling_align.py:AlignTextLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "ModelTextAttention", "ModelTextIntermediate", "ModelTextLayer", "ModelTextOutput", "None", "Optional", "Tensor", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "kwargs", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "torch", "tuple"], "align/modeling_align.py:AlignTextEncoder": ["BaseModelOutput", "False", "FloatTensor", "ModelTextEncoder", "ModelTextLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "kwargs", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple"], "align/modeling_align.py:AlignTextPooler": ["Linear", "ModelTextPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "align/modeling_align.py:AlignPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "Module", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "init", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "temperature", "temperature_init_value", "text_projection", "weight", "xavier_uniform_", "zero_"], "align/modeling_align.py:AlignTextModel": ["BaseModelOutputWithPooling", "FloatTensor", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextEncoder", "ModelTextModel", "ModelTextPooler", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "__init__", "_no_split_modules", "add_pooling_layer", "and", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "can_return_tuple", "cannot", "class", "config", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "expand", "extended_attention_mask", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "hasattr", "have", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "kwargs", "last_hidden_state", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "align/modeling_align.py:AlignVisionModel": ["AvgPool2d", "BaseModelOutputWithPoolingAndNoAttention", "False", "FloatTensor", "MaxPool2d", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "ModelVisionModel", "Module", "None", "Optional", "True", "Union", "ValueError", "You", "__init__", "auto_docstring", "be", "bool", "can_return_tuple", "ceil_mode", "class", "config", "convolution", "def", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "f", "forward", "get_input_embeddings", "got", "have", "hidden_dim", "hidden_states", "if", "is", "last_hidden_state", "main_input_name", "max", "mean", "must", "nn", "not", "of", "one", "output_hidden_states", "pixel_values", "pooled_output", "pooler", "pooler_output", "pooling", "pooling_type", "post_init", "r", "raise", "reshape", "return", "return_dict", "self", "shape", "specify", "super", "supports_gradient_checkpointing", "to", "torch", "tuple", "use_return_dict", "vision_model"], "align/modeling_align.py:AlignModel": ["FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelVisionConfig", "ModelVisionModel", "Model_loss", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "Union", "__init__", "attention_mask", "auto_docstring", "be", "bool", "but", "can_return_tuple", "class", "config", "def", "dim", "else", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "head_mask", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "inputs_embeds", "is", "isinstance", "keepdim", "last_hidden_state", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "output_attentions", "output_hidden_states", "p", "pixel_values", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "return", "return_dict", "return_loss", "self", "super", "t", "temperature", "temperature_init_value", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "token_type_ids", "torch", "tuple", "type", "use_return_dict", "vision_config", "vision_model", "vision_model_output", "vision_outputs"], "video_llava/modeling_video_llava.py:VideoLlavaModelOutputWithPast": ["Cache", "FloatTensor", "ModelModelOutputWithPast", "ModelOutput", "Model_hidden_states", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "last_hidden_state", "past_key_values", "r", "torch", "tuple"], "video_llava/modeling_video_llava.py:VideoLlavaCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "Model_hidden_states", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "video_llava/modeling_video_llava.py:VideoLlavaMultiModalProjector": ["ACT2FN", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "bias", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "image_features", "int", "isinstance", "len", "linear_1", "linear_2", "multimodal_projector_bias", "nn", "num_feature_layers", "projector_hidden_act", "return", "self", "super", "text_config", "vision_config", "vision_feature_layer"], "video_llava/modeling_video_llava.py:VideoLlavaPreTrainedModel": ["Conv2d", "Embedding", "Linear", "ModelConfig", "ModelPreTrainedModel", "ModelVisionAttention", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "else", "hasattr", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "past_key_values", "self", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "video_llava/modeling_video_llava.py:VideoLlavaModel": ["AutoModel", "Cache", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "Model_features", "Model_hidden_states", "Model_outputs", "Model_token_id", "Model_tower", "Models", "None", "Optional", "Tensor", "True", "Unexpected", "Union", "Unpack", "ValueError", "You", "_", "__init__", "_checkpoint_conversion_mapping", "all", "and", "attention_mask", "attentions", "auto_docstring", "batch_size_vid", "bool", "cache_position", "can_return_tuple", "cat", "channels", "class", "config", "decoder", "def", "default", "device", "dim", "do", "dtype", "else", "exactly", "expand_as", "f", "feature", "features", "for", "forward", "from_config", "full", "get_Model_features", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "height", "hidden_states", "hs", "hs_pool", "if", "image", "image_features", "image_hidden_states", "image_outputs", "image_token_id", "image_tower", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "language_model", "last_hidden_state", "layer_idx", "list", "long", "masked_scatter", "match", "model", "multi_modal_projector", "must", "n_Model_tokens", "n_image_tokens", "not", "num_frames", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "pixel_values", "pixel_values_Models", "pixel_values_images", "position_ids", "post_init", "r", "raise", "reshape", "return", "return_dict", "select", "self", "set_decoder", "set_input_embeddings", "shape", "special_Model_mask", "special_image_mask", "specify", "str", "strategy", "sum", "super", "tensor", "text_config", "to", "tokens", "torch", "tuple", "unsqueeze", "use_cache", "use_return_dict", "value", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vocab_size", "width"], "video_llava/modeling_video_llava.py:VideoLlavaForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Model_hidden_states", "Model_tower", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_prepare_4d_causal_attention_mask_with_cache_position", "_tied_weights_keys", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "causal_mask", "class", "clone", "config", "decoder", "def", "device", "diagonal", "dim", "dtype", "else", "expand", "fill_value", "finfo", "forward", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_hidden_states", "image_tower", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "mask_length", "masked_fill", "min", "min_dtype", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "past_key_values", "pixel_values_Models", "pixel_values_images", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "reshape", "return", "return_dict", "self", "sequence_length", "set_decoder", "set_input_embeddings", "shape", "slice", "slice_indices", "staticmethod", "str", "super", "target_length", "text_config", "to", "torch", "triu", "tuple", "use_cache", "use_return_dict", "value", "vision_feature_layer", "vision_feature_select_strategy", "vocab_size", "weight"], "x_clip/modeling_x_clip.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "x_clip/modeling_x_clip.py:x_clip_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "x_clip/modeling_x_clip.py:XCLIPOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "in", "k", "keys", "logits_per_text", "logits_per_video", "loss", "mit_output", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "video_embeds", "vision_model_output"], "x_clip/modeling_x_clip.py:XCLIPVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "Model", "ModelVisionConfig", "Module", "Parameter", "Tensor", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "dtype", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "x_clip/modeling_x_clip.py:XCLIPTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelTextConfig", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "x_clip/modeling_x_clip.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "x_clip/modeling_x_clip.py:XCLIPAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "x_clip/modeling_x_clip.py:XCLIPMLP": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "x_clip/modeling_x_clip.py:XCLIPEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "x_clip/modeling_x_clip.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "x_clip/modeling_x_clip.py:XCLIPDropPath": ["Model", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "x_clip/modeling_x_clip.py:XCLIPVisionEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "Identity", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "batch_size", "batch_time", "bool", "cat", "causal_attention_mask", "class", "config", "def", "dim", "drop_path", "drop_path_rate", "else", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "message_attn", "message_fc", "message_ln", "mlp", "msg_token", "nn", "num_frames", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "seq_length", "size", "super", "torch", "tuple", "view"], "x_clip/modeling_x_clip.py:XCLIPPreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "ModelModel", "ModelMultiframeIntegrationTransformer", "ModelTextEmbeddings", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "class_embedding", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "position_embedding", "prompts_visual_projection", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_embedding", "v_proj", "vision_embed_dim", "visual_projection", "weight", "zero_"], "x_clip/modeling_x_clip.py:XCLIPEncoder": ["BaseModelOutput", "False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "x_clip/modeling_x_clip.py:XCLIPTextTransformer": ["BaseModelOutputWithPooling", "LayerNorm", "Model", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "class", "config", "def", "device", "dim", "dtype", "either", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "final_layer_norm", "forward", "have", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "raise", "return", "return_dict", "self", "shape", "size", "specify", "super", "to", "torch", "tuple", "use_return_dict", "view"], "x_clip/modeling_x_clip.py:XCLIPTextModel": ["BaseModelOutputWithPooling", "Model", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "attention_mask", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "tuple", "value"], "x_clip/modeling_x_clip.py:XCLIPVisionEncoder": ["BaseModelOutput", "False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "v"], "x_clip/modeling_x_clip.py:XCLIPVisionTransformer": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "Model", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionEncoder", "Module", "None", "Optional", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layernorm", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "x_clip/modeling_x_clip.py:XCLIPVisionModel": ["BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionTransformer", "Module", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "vision_model"], "x_clip/modeling_x_clip.py:XCLIPMultiframeIntegrationTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "Model", "ModelEncoder", "ModelVisionConfig", "Module", "None", "Optional", "Parameter", "Union", "__init__", "attentions", "bool", "class", "config", "def", "dim", "dtype", "empty", "encoder", "encoder_outputs", "forward", "hidden_size", "hidden_states", "if", "inputs_embeds", "keepdim", "last_hidden_state", "mean", "nn", "not", "num_frames", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_embedding", "residual", "return", "return_dict", "self", "super", "torch", "tuple", "type"], "x_clip/modeling_x_clip.py:XCLIPCrossAttention": ["Dropout", "False", "Linear", "Model", "Module", "Tensor", "__init__", "_shape", "attn", "attn_drop", "batch_size", "bias", "class", "config", "contiguous", "def", "dim", "forward", "head_dim", "hidden_size", "int", "k_proj", "key_seq_len", "keys", "nn", "num_heads", "permute", "proj", "proj_drop", "projection_dim", "prompt_attention_dropout", "prompt_num_attention_heads", "prompt_projection_dropout", "q_proj", "queries", "query_seq_len", "reshape", "return", "scale", "self", "seq_len", "shape", "softmax", "super", "tensor", "torch", "transpose", "v_proj", "values", "view", "x"], "x_clip/modeling_x_clip.py:PromptGeneratorLayer": ["ACT2FN", "Dropout", "LayerNorm", "Linear", "ModelCrossAttention", "ModelGeneratorLayer", "Model_attention_dropout", "Model_hidden_act", "Module", "Sequential", "__init__", "class", "config", "cross_attn", "def", "embed_dim", "eps", "forward", "layer_norm_eps", "mlp", "nn", "norm1", "norm3", "projection_dim", "return", "self", "super", "text_config", "visual", "x"], "x_clip/modeling_x_clip.py:XCLIPPromptGenerator": ["LayerNorm", "Model", "Module", "ModuleList", "Parameter", "PromptGeneratorLayer", "_", "__init__", "alpha", "class", "config", "decoder", "def", "embed_dim", "eps", "for", "forward", "in", "layer", "layer_norm_eps", "layernorm", "nn", "ones", "projection_dim", "prompt_alpha", "prompt_layers", "range", "return", "self", "super", "text", "torch", "vision_config", "visual"], "x_clip/modeling_x_clip.py:XCLIPModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelMultiframeIntegrationTransformer", "ModelOutput", "ModelPreTrainedModel", "ModelPromptGenerator", "ModelTextConfig", "ModelTextTransformer", "ModelVisionConfig", "ModelVisionTransformer", "Model_loss", "None", "Optional", "Parameter", "T", "Tensor", "True", "TypeError", "Union", "__init__", "_attn_implementation", "attention_mask", "auto_docstring", "batch_size", "bd", "be", "bias", "bk", "bkd", "bool", "but", "class", "cls_features", "config", "copy", "def", "dim", "einsum", "else", "eps", "exp", "expand", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_text_features", "get_video_features", "height", "hidden_size", "if", "img_features", "input_ids", "intermediate_size", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "layer_norm_eps", "logit_scale", "logit_scale_init_value", "logits_per_text", "logits_per_video", "loss", "mean", "mit", "mit_config", "mit_hidden_size", "mit_intermediate_size", "mit_num_attention_heads", "mit_num_hidden_layers", "mit_output", "mit_outputs", "nn", "norm", "not", "num_attention_heads", "num_channels", "num_frames", "num_hidden_layers", "of", "output", "output_attentions", "output_hidden_states", "p", "pixel_values", "pooler_output", "position_ids", "post_init", "projection_dim", "prompts_generator", "prompts_visual_layernorm", "prompts_visual_projection", "r", "raise", "randn", "reshape", "return", "return_dict", "return_loss", "self", "shape", "super", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "unsqueeze", "use_return_dict", "video_embeds", "view", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection", "width"], "levit/modeling_levit.py:LevitForImageClassificationWithTeacherOutput": ["FloatTensor", "ModelForImageClassificationWithTeacherOutput", "ModelOutput", "None", "Optional", "class", "cls_logits", "distillation_logits", "hidden_states", "logits", "r", "torch", "tuple"], "levit/modeling_levit.py:LevitConvEmbeddings": ["BatchNorm2d", "Conv2d", "False", "ModelConvEmbeddings", "Module", "__init__", "batch_norm", "bias", "bn_weight_init", "class", "convolution", "def", "dilation", "embeddings", "forward", "groups", "in_channels", "kernel_size", "nn", "out_channels", "padding", "return", "self", "stride", "super"], "levit/modeling_levit.py:LevitPatchEmbeddings": ["Hardswish", "Make", "ModelConvEmbeddings", "ModelPatchEmbeddings", "Module", "ValueError", "__init__", "activation_layer_1", "activation_layer_2", "activation_layer_3", "channel", "class", "config", "configuration", "def", "dimension", "embedding_layer_1", "embedding_layer_2", "embedding_layer_3", "embedding_layer_4", "embeddings", "flatten", "forward", "hidden_sizes", "if", "in", "kernel_size", "match", "nn", "num_channels", "of", "one", "padding", "pixel", "pixel_values", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "transpose", "values", "with"], "levit/modeling_levit.py:MLPLayerWithBN": ["BatchNorm1d", "False", "Linear", "Model", "Module", "__init__", "batch_norm", "bias", "bn_weight_init", "class", "def", "flatten", "forward", "hidden_state", "in_features", "input_dim", "linear", "nn", "out_features", "output_dim", "reshape_as", "return", "self", "super"], "levit/modeling_levit.py:LevitSubsample": ["ModelSubsample", "Module", "_", "__init__", "batch_size", "channels", "class", "def", "forward", "hidden_state", "nn", "reshape", "resolution", "return", "self", "shape", "stride", "super", "view"], "levit/modeling_levit.py:LevitAttention": ["False", "Hardswish", "LongTensor", "MLPLayerWithBN", "ModelAttention", "Module", "Parameter", "True", "_", "__init__", "abs", "activation", "and", "append", "attention", "attention_bias_cache", "attention_bias_idxs", "attention_biases", "attention_offsets", "attention_ratio", "batch_size", "bn_weight_init", "class", "def", "device", "device_key", "dim", "else", "for", "forward", "get_attention_biases", "hidden_sizes", "hidden_state", "if", "in", "indices", "itertools", "key", "key_dim", "len", "len_points", "list", "mode", "nn", "no_grad", "not", "num_attention_heads", "offset", "out_dim_keys_values", "out_dim_projection", "p1", "p2", "permute", "persistent", "points", "product", "projection", "queries_keys_values", "query", "range", "register_buffer", "reshape", "resolution", "return", "scale", "self", "seq_length", "shape", "softmax", "split", "str", "super", "torch", "train", "training", "transpose", "value", "view", "zeros"], "levit/modeling_levit.py:LevitAttentionSubsample": ["False", "Hardswish", "LongTensor", "MLPLayerWithBN", "ModelAttentionSubsample", "ModelSubsample", "Module", "Parameter", "True", "_", "__init__", "abs", "activation", "and", "append", "attention", "attention_bias_cache", "attention_bias_idxs", "attention_biases", "attention_offsets", "attention_ratio", "batch_size", "class", "def", "device", "device_key", "dim", "else", "for", "forward", "get_attention_biases", "hidden_state", "if", "in", "indices", "input_dim", "itertools", "key", "key_dim", "keys_values", "len", "len_points", "len_points_", "list", "mode", "nn", "no_grad", "not", "num_attention_heads", "offset", "out_dim_keys_values", "out_dim_projection", "output_dim", "p1", "p2", "permute", "persistent", "points", "points_", "product", "projection", "queries", "queries_subsample", "query", "range", "register_buffer", "reshape", "resolution_in", "resolution_out", "return", "scale", "self", "seq_length", "shape", "size", "softmax", "split", "str", "stride", "super", "torch", "train", "training", "transpose", "value", "view", "zeros"], "levit/modeling_levit.py:LevitMLPLayer": ["Hardswish", "MLPLayerWithBN", "ModelMLPLayer", "Module", "__init__", "activation", "class", "def", "forward", "hidden_dim", "hidden_state", "input_dim", "linear_down", "linear_up", "nn", "return", "self", "super"], "levit/modeling_levit.py:LevitResidualLayer": ["ModelResidualLayer", "Module", "__init__", "and", "class", "def", "detach", "device", "div", "drop_rate", "else", "forward", "ge_", "hidden_state", "if", "module", "nn", "rand", "return", "rnd", "self", "size", "super", "torch", "training"], "levit/modeling_levit.py:LevitStage": ["ModelAttention", "ModelAttentionSubsample", "ModelMLPLayer", "ModelResidualLayer", "ModelStage", "Module", "ModuleList", "Subsample", "_", "__init__", "append", "attention_ratio", "class", "config", "def", "depths", "down_ops", "drop_path_rate", "for", "forward", "get_resolution", "hidden_dim", "hidden_sizes", "hidden_state", "idx", "if", "in", "key_dim", "layer", "layers", "mlp_ratio", "nn", "num_attention_heads", "range", "resolution_in", "resolution_out", "return", "self", "stride", "super"], "levit/modeling_levit.py:LevitEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "True", "__init__", "all_hidden_states", "append", "attention_ratio", "class", "config", "def", "depths", "down_ops", "else", "for", "forward", "get_resolution", "hidden_sizes", "hidden_state", "hidden_states", "if", "image_size", "in", "is", "key_dim", "last_hidden_state", "len", "mlp_ratio", "nn", "not", "num_attention_heads", "output_hidden_states", "patch_size", "range", "resolution", "return", "return_dict", "self", "stage", "stage_idx", "stages", "super", "tuple", "v"], "levit/modeling_levit.py:LevitClassificationLayer": ["BatchNorm1d", "Linear", "ModelClassificationLayer", "Module", "__init__", "batch_norm", "class", "def", "forward", "hidden_state", "input_dim", "linear", "logits", "nn", "output_dim", "return", "self", "super"], "levit/modeling_levit.py:LevitPreTrainedModel": ["BatchNorm1d", "BatchNorm2d", "Conv2d", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelResidualLayer", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "levit/modeling_levit.py:LevitModel": ["BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Union", "ValueError", "You", "__init__", "auto_docstring", "bool", "class", "config", "def", "dim", "else", "embeddings", "encoder", "encoder_outputs", "forward", "have", "hidden_states", "if", "is", "last_hidden_state", "mean", "not", "output_hidden_states", "patch_embeddings", "pixel_values", "pooled_output", "pooler_output", "post_init", "raise", "return", "return_dict", "self", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "levit/modeling_levit.py:LevitForImageClassification": ["FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "LongTensor", "Model", "ModelClassificationLayer", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "config", "def", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "levit/modeling_levit.py:LevitForImageClassificationWithTeacher": ["FloatTensor", "Identity", "Model", "ModelClassificationLayer", "ModelForImageClassificationWithTeacher", "ModelForImageClassificationWithTeacherOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Union", "__init__", "auto_docstring", "bool", "class", "classifier", "classifier_distill", "cls_logits", "config", "def", "distill_logits", "distillation_logits", "else", "forward", "hidden_sizes", "hidden_states", "if", "is", "logits", "mean", "nn", "not", "num_labels", "output", "output_hidden_states", "outputs", "pixel_values", "post_init", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "smollm3/modeling_smollm3.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "smollm3/modeling_smollm3.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "smollm3/modeling_smollm3.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "smollm3/modeling_smollm3.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "smollm3/modeling_smollm3.py:SmolLM3Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "and", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "no_rope_layers", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "use_rope", "use_sliding_window", "v_proj", "value_states", "version", "view"], "smollm3/modeling_smollm3.py:SmolLM3RMSNorm": ["ModelRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "smollm3/modeling_smollm3.py:SmolLM3MLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "mlp_bias", "nn", "return", "self", "super", "up_proj", "x"], "smollm3/modeling_smollm3.py:SmolLM3DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "smollm3/modeling_smollm3.py:SmolLM3PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "smollm3/modeling_smollm3.py:SmolLM3RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "smollm3/modeling_smollm3.py:SmolLM3Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_types", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "smollm3/modeling_smollm3.py:SmolLM3ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "smollm3/modeling_smollm3.py:SmolLM3ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "smollm3/modeling_smollm3.py:SmolLM3ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "smollm3/modeling_smollm3.py:SmolLM3ForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "clipseg/modeling_clipseg.py:contrastive_loss": ["Model_loss", "Tensor", "arange", "cross_entropy", "def", "device", "functional", "len", "logits", "nn", "return", "torch"], "clipseg/modeling_clipseg.py:clipseg_loss": ["Model_loss", "Tensor", "caption_loss", "contrastive_loss", "def", "image_loss", "return", "similarity", "t", "torch"], "clipseg/modeling_clipseg.py:CLIPSegOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "clipseg/modeling_clipseg.py:CLIPSegDecoderOutput": ["FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "r", "torch", "tuple"], "clipseg/modeling_clipseg.py:CLIPSegImageSegmentationOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "class", "conditional_embeddings", "decoder_output", "def", "else", "for", "getattr", "if", "in", "k", "keys", "logits", "loss", "not", "pooled_output", "r", "return", "self", "to_tuple", "torch", "tuple", "vision_model_output"], "clipseg/modeling_clipseg.py:CLIPSegVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "Input", "Model", "ModelVisionConfig", "Module", "Parameter", "Tensor", "True", "ValueError", "_", "__init__", "align_corners", "and", "arange", "batch_size", "bias", "bicubic", "cat", "class", "class_embedding", "class_embeds", "class_pos_embed", "config", "def", "dim", "doesn", "else", "embed_dim", "embeddings", "expand", "f", "flatten", "forward", "functional", "height", "hidden_size", "if", "image", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "match", "mode", "model", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "or", "out_channels", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "raise", "randn", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "t", "torch", "torch_int", "transpose", "unsqueeze", "view", "weight", "width"], "clipseg/modeling_clipseg.py:CLIPSegTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "Model", "ModelTextConfig", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "clipseg/modeling_clipseg.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "clipseg/modeling_clipseg.py:CLIPSegAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model", "ModelTextConfig", "ModelVisionConfig", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bool", "by", "causal_attention_mask", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "elif", "else", "embed_dim", "f", "flash_attention_2", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is", "is_causal", "k_proj", "keys", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "output_attentions", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "clipseg/modeling_clipseg.py:CLIPSegMLP": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "clipseg/modeling_clipseg.py:CLIPSegEncoderLayer": ["False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "clipseg/modeling_clipseg.py:CLIPSegPreTrainedModel": ["LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "ModelModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_init_weights", "and", "base_model_prefix", "bias", "class", "class_embedding", "clip", "config", "data", "def", "elif", "embed_dim", "factor", "fc1", "fc2", "fc_std", "fill_", "hidden_size", "if", "in_proj_std", "init", "initializer_factor", "initializer_range", "is", "isinstance", "k_proj", "mean", "module", "nn", "normal_", "not", "num_hidden_layers", "out_proj", "out_proj_std", "patch_embedding", "position_embedding", "q_proj", "self", "std", "supports_gradient_checkpointing", "text_embed_dim", "text_projection", "token_embedding", "v_proj", "vision_embed_dim", "visual_projection", "weight", "zero_"], "clipseg/modeling_clipseg.py:CLIPSegEncoder": ["BaseModelOutput", "False", "Model", "ModelConfig", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "Union", "_", "__init__", "all_attentions", "attention_mask", "attentions", "bool", "can_return_tuple", "causal_attention_mask", "class", "config", "def", "else", "encoder_layer", "encoder_states", "enumerate", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "r", "range", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "clipseg/modeling_clipseg.py:CLIPSegTextTransformer": ["BaseModelOutputWithPooling", "LayerNorm", "Model", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Tensor", "Union", "ValueError", "You", "__init__", "_create_4d_causal_attention_mask", "_prepare_4d_attention_mask", "arange", "argmax", "attention_mask", "attentions", "auto_docstring", "bool", "causal_attention_mask", "class", "config", "def", "device", "dim", "dtype", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eos_token_id", "eps", "final_layer_norm", "forward", "have", "hidden_size", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pooled_output", "pooler_output", "position_ids", "raise", "return", "return_dict", "self", "shape", "size", "specify", "super", "to", "torch", "tuple", "use_return_dict", "view"], "clipseg/modeling_clipseg.py:CLIPSegTextModel": ["BaseModelOutputWithPooling", "Model", "ModelEncoderLayer", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "Union", "__init__", "_no_split_modules", "attention_mask", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "nn", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "return", "return_dict", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "tuple", "value"], "clipseg/modeling_clipseg.py:CLIPSegVisionTransformer": ["BaseModelOutputWithPooling", "FloatTensor", "LayerNorm", "Model", "ModelEncoder", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "None", "Optional", "True", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "is", "last_hidden_state", "layer_norm_eps", "nn", "not", "output_attentions", "output_hidden_states", "pixel_values", "pooled_output", "pooler_output", "post_layernorm", "pre_layrnorm", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "clipseg/modeling_clipseg.py:CLIPSegVisionModel": ["BaseModelOutputWithPooling", "FloatTensor", "Model", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionTransformer", "Module", "None", "Optional", "True", "Union", "__init__", "auto_docstring", "bool", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "main_input_name", "nn", "output_attentions", "output_hidden_states", "patch_embedding", "pixel_values", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "vision_model"], "clipseg/modeling_clipseg.py:CLIPSegModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelConfig", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextTransformer", "ModelVisionConfig", "ModelVisionTransformer", "Model_loss", "None", "Optional", "Parameter", "Tensor", "True", "TypeError", "Union", "__init__", "_attn_implementation", "attention_mask", "auto_docstring", "be", "bias", "bool", "but", "class", "config", "def", "dim", "else", "exp", "expected", "f", "filter_out_non_signature_kwargs", "forward", "get_image_features", "get_text_features", "hidden_size", "if", "image_embeds", "image_features", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "logit_scale", "logit_scale_init_value", "logits_per_image", "logits_per_text", "loss", "matmul", "nn", "norm", "not", "of", "output", "output_attentions", "output_hidden_states", "p", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "r", "raise", "return", "return_dict", "return_loss", "self", "super", "t", "tensor", "text_config", "text_embed_dim", "text_embeds", "text_features", "text_model", "text_model_output", "text_outputs", "text_projection", "to", "torch", "tuple", "type", "use_return_dict", "vision_config", "vision_embed_dim", "vision_model", "vision_model_output", "vision_outputs", "visual_projection"], "clipseg/modeling_clipseg.py:CLIPSegDecoderLayer": ["False", "FloatTensor", "LayerNorm", "Model", "ModelAttention", "ModelConfig", "ModelMLP", "Module", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "causal_attention_mask", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "if", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "output_attentions", "outputs", "residual", "return", "self", "self_attn", "super", "torch", "tuple"], "clipseg/modeling_clipseg.py:CLIPSegDecoder": ["Conv2d", "ConvTranspose2d", "Linear", "Model", "ModelConfig", "ModelLayer", "ModelOutput", "ModelPreTrainedModel", "ModuleList", "None", "Optional", "ReLU", "Sequential", "Tensor", "True", "_", "__init__", "activation", "activations", "all_attentions", "all_hidden_states", "attention_mask", "attentions", "batch_size", "bool", "causal_attention_mask", "class", "conditional_embeddings", "conditional_layer", "config", "copy", "decoder_config", "decoder_intermediate_size", "decoder_num_attention_heads", "deepcopy", "def", "depth", "else", "enumerate", "extract_layers", "film_add", "film_mul", "for", "forward", "hidden_act", "hidden_size", "hidden_states", "i", "if", "in", "int", "intermediate_size", "is", "kernel_size", "layer", "layer_outputs", "layers", "len", "logits", "math", "nn", "not", "num_attention_heads", "output", "output_attentions", "output_hidden_states", "padding", "patch_size", "permute", "projection_dim", "range", "reduce", "reduce_dim", "reduces", "relu", "return", "return_dict", "self", "shape", "size", "sqrt", "squeeze", "stride", "super", "torch", "transposed_convolution", "transposed_kernels", "tuple", "use_complex_transposed_convolution", "v", "view", "vision_config", "zip"], "clipseg/modeling_clipseg.py:CLIPSegForImageSegmentation": ["BCEWithLogitsLoss", "BaseModelOutputWithPooling", "FloatTensor", "Invalid", "LongTensor", "Make", "Model", "ModelConfig", "ModelDecoder", "ModelImageSegmentationOutput", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "ValueError", "__init__", "activations", "are", "as", "attention_mask", "attentions", "auto_docstring", "batch", "batch_size", "be", "bool", "class", "clip", "conditional", "conditional_embeddings", "conditional_pixel_values", "config", "decoder", "decoder_output", "decoder_outputs", "def", "device", "dimension", "either", "elif", "else", "embeddings", "extract_layers", "feature", "for", "forward", "get_conditional_embeddings", "get_image_features", "get_text_features", "hidden_states", "i", "if", "images", "in", "input_ids", "int", "interpolate_pos_encoding", "is", "labels", "last_hidden_state", "len", "logits", "loss", "loss_fn", "many", "matches", "nn", "no_grad", "not", "of", "or", "output", "output_attentions", "output_hidden_states", "pass", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "projection_dim", "prompt", "provided", "query", "r", "raise", "return", "return_dict", "self", "shape", "should", "super", "sure", "texts", "that", "the", "there", "to", "torch", "tuple", "use_return_dict", "vision_model", "vision_model_output", "vision_outputs", "visual_projection", "with"], "cohere2/modeling_cohere2.py:Cohere2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "repeat_interleave", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "cohere2/modeling_cohere2.py:Cohere2LayerNorm": ["False", "ModelLayerNorm", "Module", "None", "Parameter", "True", "__init__", "bias", "class", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "cohere2/modeling_cohere2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "cohere2/modeling_cohere2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "cohere2/modeling_cohere2.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "rot_x", "stack", "torch", "x", "x1", "x2"], "cohere2/modeling_cohere2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "float", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "cohere2/modeling_cohere2.py:Cohere2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "cohere2/modeling_cohere2.py:Cohere2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "cohere2/modeling_cohere2.py:Cohere2DecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm", "ModelMLP", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "hidden_states_attention", "hidden_states_mlp", "input_layernorm", "int", "kwargs", "layer_idx", "layer_norm_eps", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "cohere2/modeling_cohere2.py:Cohere2PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "cohere2/modeling_cohere2.py:Cohere2Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_eps", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "training", "unsqueeze", "use_cache", "vocab_size"], "cohere2/modeling_cohere2.py:Cohere2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logit_scale", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "tie_word_embeddings", "torch", "use_cache", "vocab_size", "weight"], "llava_next/modeling_llava_next.py:get_anyres_image_grid_shape": ["Model_anyres_image_grid_shape", "Tensor", "TypeError", "a", "be", "def", "either", "f", "grid_pinpoints", "height", "if", "image_size", "invalid", "isinstance", "list", "lists", "ndarray", "not", "np", "of", "or", "patch_size", "raise", "return", "select_best_resolution", "should", "tensor", "tolist", "torch", "tuple", "tuples", "type", "valid", "width"], "llava_next/modeling_llava_next.py:image_size_to_num_patches": ["Model_size", "Model_size_to_num_patches", "Tensor", "TypeError", "a", "be", "best_resolution", "def", "f", "for", "grid_pinpoints", "height", "i", "if", "in", "int", "invalid", "isinstance", "j", "list", "lists", "ndarray", "not", "np", "num_patches", "of", "or", "patch_size", "raise", "range", "return", "select_best_resolution", "should", "tolist", "torch", "tuple", "tuples", "type", "value", "width", "with"], "llava_next/modeling_llava_next.py:unpad_image": ["Model_image", "Modelded_tensor", "Tensor", "TypeError", "be", "current_aspect_ratio", "current_height", "current_width", "def", "either", "else", "f", "if", "image_size", "int", "invalid", "isinstance", "list", "ndarray", "new_height", "new_width", "not", "np", "or", "original_aspect_ratio", "original_height", "original_size", "original_width", "padding", "raise", "return", "round", "scale_factor", "shape", "should", "tensor", "tolist", "torch", "tuple", "type", "valid"], "llava_next/modeling_llava_next.py:LlavaNextModelOutputWithPast": ["BaseModelOutputWithPast", "FloatTensor", "ModelModelOutputWithPast", "None", "Optional", "class", "image_hidden_states", "r", "torch"], "llava_next/modeling_llava_next.py:LlavaNextCausalLMOutputWithPast": ["Cache", "FloatTensor", "ModelCausalLMOutputWithPast", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_hidden_states", "logits", "loss", "past_key_values", "r", "torch", "tuple"], "llava_next/modeling_llava_next.py:LlavaNextMultiModalProjector": ["ACT2FN", "Linear", "ModelConfig", "ModelMultiModalProjector", "Module", "__init__", "act", "bias", "class", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "image_features", "int", "isinstance", "len", "linear_1", "linear_2", "multimodal_projector_bias", "nn", "num_feature_layers", "projector_hidden_act", "return", "self", "super", "text_config", "vision_config", "vision_feature_layer"], "llava_next/modeling_llava_next.py:LlavaNextPreTrainedModel": ["Linear", "LlamaDecoderLayer", "ModelConfig", "ModelModel", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_can_compile_fullgraph", "_init_weights", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "embed_std", "get_text_config", "getattr", "hidden_size", "if", "image_newline", "initializer_range", "is", "isinstance", "math", "mean", "module", "nn", "normal_", "not", "past_key_values", "self", "sqrt", "std", "supports_gradient_checkpointing", "text_config", "weight", "zero_"], "llava_next/modeling_llava_next.py:LlavaNextModel": ["AutoModel", "CLS", "Cache", "FlashAttentionKwargs", "FloatTensor", "Image", "LongTensor", "ModelConfig", "ModelModel", "ModelModelOutputWithPast", "ModelMultiModalProjector", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_checkpoint_conversion_mapping", "_pixel_values_list", "a", "all", "and", "append", "attention_mask", "attentions", "auto_docstring", "base_image_feature", "be", "bool", "cache_position", "can_return_tuple", "cat", "class", "config", "contiguous", "decoder", "def", "default", "device", "dim", "dimensions", "do", "does", "dtype", "elif", "else", "embed_std", "encoder", "enumerate", "exactly", "expand", "expand_as", "expect", "f", "feature", "feature_lens", "features", "flatten", "for", "forward", "from_config", "get_anyres_image_grid_shape", "get_decoder", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "grid_pinpoints", "have", "height", "hidden_size", "hidden_states", "hs_pool", "if", "image", "image_feature", "image_features", "image_grid_pinpoints", "image_hidden_states", "image_idx", "image_newline", "image_num_patches", "image_size", "image_size_to_num_patches", "image_sizes", "image_token_id", "imsize", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "language_model", "last_hidden_state", "layer_idx", "line", "list", "logger", "long", "masked_scatter", "match", "math", "may", "model", "multi_modal_projector", "must", "n_image_tokens", "new_image_features", "nn", "not", "np", "num_patch", "num_patch_height", "num_patch_width", "numel", "of", "one", "or", "output_attentions", "output_hidden_states", "outputs", "pack_image_features", "pad_token_id", "past_key_values", "patch", "patch_size", "permute", "pix_val", "pixel_values", "position_ids", "post_init", "prod", "provided", "r", "raise", "randn", "return", "return_dict", "selected_image_feature", "self", "set_decoder", "set_input_embeddings", "shape", "size", "special_image_mask", "specify", "split", "sqrt", "str", "sum", "super", "tensor", "text_config", "that", "the", "to", "tokens", "torch", "transpose", "tuple", "unpad_image", "unsqueeze", "up", "use_cache", "use_return_dict", "using", "value", "view", "vision_config", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "visual", "vocab_size", "warning_once", "width", "with", "zip"], "llava_next/modeling_llava_next.py:LlavaNextForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelCausalLMOutputWithPast", "ModelConfig", "ModelForConditionalGeneration", "ModelModel", "ModelPreTrainedModel", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "__init__", "_checkpoint_conversion_mapping", "_prepare_4d_causal_attention_mask_with_cache_position", "_tied_weights_keys", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "bias", "bool", "cache_position", "can_return_tuple", "causal_mask", "class", "clone", "config", "decoder", "def", "device", "diagonal", "dim", "dtype", "else", "expand", "fill_value", "finfo", "forward", "full", "get_decoder", "get_image_features", "get_input_embeddings", "get_output_embeddings", "hidden_size", "hidden_states", "if", "image_features", "image_hidden_states", "image_newline", "image_sizes", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "language_model", "list", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "mask_length", "masked_fill", "min", "min_dtype", "model", "model_inputs", "multi_modal_projector", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "pack_image_features", "padding_mask", "past_key_values", "pixel_values", "position_ids", "post_init", "prepare_inputs_for_generation", "property", "r", "reshape", "return", "return_dict", "self", "sequence_length", "set_decoder", "set_input_embeddings", "shape", "slice", "slice_indices", "staticmethod", "str", "super", "target_length", "text_config", "to", "torch", "triu", "tuple", "use_cache", "value", "vision_feature_layer", "vision_feature_select_strategy", "vision_tower", "vocab_size", "weight"], "cpmant/modeling_cpmant.py:CpmAntLayerNorm": ["AssertionError", "ModelConfig", "ModelLayerNorm", "Module", "Parameter", "Tensor", "True", "__init__", "class", "config", "def", "dim", "dim_norm", "dtype", "empty", "eps", "float32", "forward", "hidden_size", "hidden_states", "if", "keepdim", "mean", "nn", "old_dtype", "pow", "raise", "return", "rsqrt", "self", "size", "super", "to", "torch", "variance", "weight"], "cpmant/modeling_cpmant.py:CpmAntAttention": ["BoolTensor", "Cache", "Dropout", "False", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Softmax", "Tensor", "__init__", "attention_mask", "attention_out", "attn_weights", "batch_size", "bias", "bool", "cache_position", "class", "config", "contiguous", "def", "device", "dim", "dim_head", "dim_model", "dropout", "dropout_p", "dtype", "else", "float", "forward", "hidden_kv", "hidden_q", "hidden_size", "if", "inf", "is", "key", "layer_idx", "len_k", "len_q", "masked_fill", "math", "matmul", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "p", "past_key_values", "permute", "position_bias", "project_k", "project_q", "project_v", "query", "return", "scalar_tensor", "score", "self", "size", "softmax", "sqrt", "super", "tensor", "torch", "transpose", "update", "use_cache", "value", "view"], "cpmant/modeling_cpmant.py:CpmAntSelfAttentionBlock": ["Cache", "Dropout", "False", "ModelAttention", "ModelConfig", "ModelLayerNorm", "ModelSelfAttentionBlock", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "cache_position", "class", "config", "def", "dropout", "dropout_p", "else", "forward", "hidden_states", "if", "is", "layer_idx", "layernorm_before_attention", "nn", "not", "output_attentions", "outputs", "past_key_values", "position_bias", "return", "self", "self_attention", "super", "torch", "use_cache"], "cpmant/modeling_cpmant.py:CpmAntDenseGatedACT": ["False", "GELU", "Linear", "ModelConfig", "ModelDenseGatedACT", "Module", "Tensor", "__init__", "act", "bias", "class", "config", "def", "dim_ff", "forward", "gate_score", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "w_0", "w_1"], "cpmant/modeling_cpmant.py:CpmAntFeedForward": ["Dropout", "False", "Linear", "ModelConfig", "ModelDenseGatedACT", "ModelFeedForward", "Module", "None", "Tensor", "__init__", "bias", "class", "config", "def", "dim_ff", "dropout", "dropout_p", "else", "forward", "hidden_size", "hidden_states", "if", "is", "nn", "not", "return", "self", "super", "torch", "w_in", "w_out"], "cpmant/modeling_cpmant.py:CpmAntFFNBlock": ["Dropout", "ModelConfig", "ModelFFNBlock", "ModelFeedForward", "ModelLayerNorm", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dropout", "dropout_p", "else", "ffn", "forward", "hidden_states", "if", "is", "layernorm_before_ffn", "ln_outputs", "nn", "not", "outputs", "return", "self", "super", "torch"], "cpmant/modeling_cpmant.py:CpmAntTransformerBlock": ["Cache", "False", "ModelConfig", "ModelFFNBlock", "ModelSelfAttentionBlock", "ModelTransformerBlock", "Module", "None", "Optional", "Tensor", "__init__", "attention_mask", "attn_weights", "bool", "cache_position", "class", "config", "def", "ffn", "forward", "hidden_states", "layer_idx", "nn", "output_attentions", "past_key_values", "position_bias", "return", "self", "self_att", "super", "torch", "use_cache"], "cpmant/modeling_cpmant.py:CpmAntEncoder": ["Cache", "ModelConfig", "ModelEncoder", "ModelLayerNorm", "ModelTransformerBlock", "Module", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_hidden_states", "all_self_attns", "attention_mask", "attn_weights", "bool", "cache_position", "class", "config", "def", "else", "enumerate", "for", "forward", "hidden_states", "i", "if", "in", "layer", "layer_idx", "layer_outputs", "layers", "nn", "num_hidden_layers", "num_layers", "output_attentions", "output_hidden_states", "output_layernorm", "past_key_values", "position_bias", "range", "return", "self", "super", "torch", "use_cache"], "cpmant/modeling_cpmant.py:CpmAntIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "cpmant/modeling_cpmant.py:CpmAntSegmentPositionEmbedding": ["AssertionError", "F", "ModelConfig", "ModelSegmentPositionEmbedding", "Module", "None", "Parameter", "Tensor", "__init__", "_position_bucket", "_segment_relative_position_bucket", "abs", "absolute_position_bucket", "and", "arange", "batch", "be", "but", "class", "config", "contiguous", "def", "device", "dtype", "embedding", "embeds", "empty", "equal", "f", "float", "forward", "full_like", "got", "if", "int32", "is_small", "key_pos", "key_segment", "keylen", "log", "math", "max_distance", "max_exact", "min", "nn", "no_grad", "num_attention_heads", "num_buckets", "num_heads", "num_segments", "or", "permute", "position_bias_max_distance", "position_bias_num_buckets", "query_pos", "query_segment", "querylen", "raise", "relative_attention_bias", "relative_buckets", "relative_position", "relative_position_bucket", "relative_position_if_large", "return", "segment_types", "self", "should", "size", "super", "to", "torch", "view", "where", "with"], "cpmant/modeling_cpmant.py:CpmAntOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "cpmant/modeling_cpmant.py:CpmAntPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelLayerNorm", "ModelPreTrainedModel", "ModelSegmentPositionEmbedding", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "init_std", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "relative_attention_bias", "self", "std", "weight", "zero_"], "cpmant/modeling_cpmant.py:CpmAntModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "ModelConfig", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "ModelSegmentPositionEmbedding", "None", "Optional", "Passing", "Tensor", "Transformers", "Union", "You", "__init__", "_prepare_attention_mask", "a", "all_attentions", "all_hidden_states", "an", "and", "arange", "attention", "attention_mask", "attentions", "auto_docstring", "batch", "be", "bool", "cache_position", "cat", "class", "config", "context", "contiguous", "def", "deprecated", "device", "dim", "directional_mask_2d", "dtype", "e", "else", "embeddings", "encoder", "for", "forward", "from_legacy_cache", "full", "g", "get_input_embeddings", "get_seq_length", "hidden_size", "hidden_state", "hidden_states", "if", "in", "input_embedding", "input_ids", "instance", "instead", "int32", "is", "isinstance", "kwargs", "last_hidden_state", "length", "list", "logger", "logical_not", "mask_1d", "new_attentions", "new_hidden_states", "nn", "not", "of", "ones", "output_attentions", "output_hidden_states", "pass", "past_key_values", "past_length", "position", "position_bias", "post_init", "prompt_length", "prompt_types", "r", "range", "removed", "repeat", "return", "return_dict", "segment", "segment_embedding", "segment_states", "segment_types", "self", "seq_length", "seqlen", "set_input_embeddings", "should", "size", "span", "sum", "super", "tensor", "to", "torch", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "where", "will", "zeros"], "cpmant/modeling_cpmant.py:CpmAntForCausalLM": ["Cache", "CausalLMOutputWithPast", "CrossEntropyLoss", "False", "GenerationMixin", "Linear", "Model", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "Union", "__init__", "_reorder_cache", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "beam_idx", "bias", "bool", "cache_position", "class", "config", "def", "each", "else", "embeddings", "for", "forward", "get_input_embeddings", "hidden_size", "hidden_states", "if", "in", "input_embedding", "input_ids", "is", "key_value_layer", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logits", "loss", "loss_func", "model_output", "nn", "not", "output", "output_attentions", "output_hidden_states", "past_key_values", "post_init", "prompt_length", "prompt_types", "r", "return", "return_dict", "self", "set_input_embeddings", "size", "super", "torch", "tuple", "use_cache", "use_return_dict", "view", "vocab_size", "weight"], "sew_d/modeling_sew_d.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "sew_d/modeling_sew_d.py:make_log_bucket_position": ["Model_log_bucket_position", "abs", "abs_pos", "bucket_pos", "bucket_size", "ceil", "def", "log", "log_pos", "max_position", "mid", "relative_pos", "return", "sign", "tensor", "torch", "type_as", "where"], "sew_d/modeling_sew_d.py:build_relative_position": ["Model_relative_position", "None", "and", "arange", "bucket_size", "def", "device", "if", "k_ids", "key_size", "long", "make_log_bucket_position", "max_position", "q_ids", "query_size", "rel_pos_ids", "return", "to", "torch", "unsqueeze"], "sew_d/modeling_sew_d.py:c2p_dynamic_expand": ["Model_dynamic_expand", "Model_pos", "def", "expand", "query_layer", "relative_pos", "return", "size"], "sew_d/modeling_sew_d.py:p2c_dynamic_expand": ["Model_dynamic_expand", "c2p_pos", "def", "expand", "key_layer", "query_layer", "return", "size"], "sew_d/modeling_sew_d.py:pos_dynamic_expand": ["Model_dynamic_expand", "Model_index", "def", "expand", "key_layer", "p2c_att", "return", "size"], "sew_d/modeling_sew_d.py:get_mask": ["DropoutContext", "Model_mask", "None", "and", "bernoulli_", "bool", "def", "dropout", "else", "empty_like", "if", "input", "is", "isinstance", "local_context", "mask", "not", "return", "reuse_mask", "scale", "to", "torch"], "sew_d/modeling_sew_d.py:SEWDNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "Model", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "sew_d/modeling_sew_d.py:SEWDLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "Model", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "sew_d/modeling_sew_d.py:SEWDGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "Model", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "sew_d/modeling_sew_d.py:SEWDPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "Model", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "squeeze_factor", "stride", "super", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "sew_d/modeling_sew_d.py:SEWDSamePadLayer": ["Model", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "sew_d/modeling_sew_d.py:SEWDUpsampling": ["ACT2FN", "Linear", "Model", "Module", "__init__", "activation", "bsz", "class", "config", "def", "feat_extract_activation", "forward", "hidden_size", "hidden_states", "if", "nn", "projection", "reshape", "return", "self", "size", "squeeze_factor", "src_embed_dim", "src_len", "super", "tgt_embed_dim", "tgt_len"], "sew_d/modeling_sew_d.py:SEWDFeatureEncoder": ["False", "Model", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "sew_d/modeling_sew_d.py:SEWDFeatureExtractor": ["FutureWarning", "Model", "ModelFeatureEncoder", "The", "Transformers", "Use", "__bases__", "__class__", "__init__", "__name__", "and", "be", "been", "class", "config", "def", "depreciated", "f", "has", "in", "instead", "removed", "self", "super", "v5", "warn", "warnings", "will"], "sew_d/modeling_sew_d.py:ContextPooler": ["ACT2FN", "Linear", "ModelPooler", "Model_token", "Module", "StableDropout", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_size", "hidden_states", "nn", "output_dim", "pooled_output", "pooler_dropout", "pooler_hidden_act", "pooler_hidden_size", "property", "return", "self", "super"], "sew_d/modeling_sew_d.py:XSoftmax": ["Bool", "Cast", "Constant", "Function", "Long", "Model", "None", "Sub", "autograd", "backward", "bool", "cast_pytorch_to_onnx", "class", "ctx", "def", "dim", "dtype", "finfo", "forward", "g", "grad_output", "input", "inputGrad", "int64", "mask", "mask_cast_value", "masked_fill", "masked_fill_", "min", "op", "output", "r_mask", "return", "rmask", "save_for_backward", "saved_tensors", "self", "softmax", "softmax_backward_data", "staticmethod", "sym_help", "symbolic", "tensor", "to", "to_i", "torch", "type", "value_t"], "sew_d/modeling_sew_d.py:DropoutContext": ["Model", "ModelContext", "None", "True", "__init__", "class", "def", "mask", "reuse_mask", "scale", "self"], "sew_d/modeling_sew_d.py:XDropout": ["DropoutContext", "Function", "Graph", "Model", "None", "True", "Union", "Value", "_C", "autograd", "backward", "class", "ctx", "def", "dropout", "dropout_p", "else", "float", "forward", "g", "get_mask", "grad_output", "if", "input", "isinstance", "local_ctx", "mask", "masked_fill", "return", "save_for_backward", "saved_tensors", "scale", "staticmethod", "symbolic", "symbolic_opset12", "torch", "train"], "sew_d/modeling_sew_d.py:StableDropout": ["DropoutContext", "ModelDropout", "Module", "None", "True", "XDropout", "__init__", "and", "append", "apply", "c", "class", "clear_context", "context_stack", "count", "ctx", "def", "drop_prob", "dropout", "else", "for", "forward", "get_context", "if", "in", "init_context", "is", "len", "nn", "not", "return", "reuse_mask", "scale", "self", "super", "training", "x"], "sew_d/modeling_sew_d.py:SEWDSelfOutput": ["Dropout", "LayerNorm", "Linear", "Model", "Module", "__init__", "activation_dropout", "class", "config", "def", "dense", "dropout", "forward", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super"], "sew_d/modeling_sew_d.py:DisentangledSelfAttention": ["False", "Linear", "ModelSelfAttention", "Model_attention_bias", "Module", "None", "Relative", "StableDropout", "The", "True", "ValueError", "XSoftmax", "__init__", "_attention_head_size", "a", "activation_dropout", "all_head_size", "apply", "att_span", "attention", "attention_dropout", "attention_head_size", "attention_heads", "attention_mask", "attention_probs", "attention_scores", "be", "bias", "bmm", "bucket_size", "build_relative_position", "c2p", "c2p_att", "c2p_pos", "clamp", "class", "config", "context_layer", "contiguous", "def", "device", "dim", "dropout", "dtype", "elif", "else", "expand", "f", "float", "forward", "gather", "getattr", "heads", "hidden", "hidden_size", "hidden_states", "ids", "if", "in", "index", "is", "key_layer", "key_proj", "long", "max_position", "max_position_embeddings", "max_relative_positions", "multiple", "must", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "p2c", "p2c_att", "p2c_pos", "permute", "pos_att_type", "pos_dropout", "pos_ebd_size", "pos_key_layer", "pos_key_proj", "pos_query_layer", "pos_query_proj", "position", "position_buckets", "q", "query_layer", "query_proj", "query_states", "r_pos", "raise", "rel_att", "rel_embeddings", "relative_attention", "relative_pos", "repeat", "return", "scale", "scale_factor", "score", "self", "share_att_key", "size", "sqrt", "squeeze", "super", "tensor", "the", "to", "torch", "transpose", "transpose_for_scores", "unsqueeze", "value_layer", "value_proj", "view", "x"], "sew_d/modeling_sew_d.py:SEWDAttention": ["DisentangledSelfAttention", "False", "Model", "ModelSelfOutput", "Module", "None", "__init__", "att_matrix", "attention_mask", "attention_output", "class", "config", "def", "else", "forward", "hidden_states", "if", "is", "nn", "output", "output_attentions", "query_states", "rel_embeddings", "relative_pos", "return", "self", "self_output", "super"], "sew_d/modeling_sew_d.py:SEWDIntermediate": ["ACT2FN", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "sew_d/modeling_sew_d.py:SEWDOutput": ["Dropout", "LayerNorm", "Linear", "Model", "Module", "__init__", "activation_dropout", "class", "config", "def", "dense", "dropout", "forward", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super"], "sew_d/modeling_sew_d.py:SEWDLayer": ["False", "GradientCheckpointingLayer", "Model", "ModelAttention", "ModelIntermediate", "ModelOutput", "None", "__init__", "att_matrix", "attention", "attention_mask", "attention_output", "class", "config", "def", "else", "forward", "hidden_states", "if", "intermediate", "intermediate_output", "layer_output", "output", "output_attentions", "query_states", "rel_embeddings", "relative_pos", "return", "self", "super"], "sew_d/modeling_sew_d.py:ConvLayer": ["ACT2FN", "LayerNorm", "Model", "Model1d", "ModelLayer", "Model_act", "Model_groups", "Model_kernel_size", "Module", "None", "StableDropout", "__init__", "bool", "class", "config", "contiguous", "def", "dim", "dropout", "dtype", "else", "expand", "forward", "getattr", "groups", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_mask", "is", "kernel_size", "layer_norm_eps", "layer_norm_input", "masked_fill_", "nn", "out", "output", "output_states", "padding", "permute", "residual_states", "return", "rmask", "self", "size", "squeeze", "super", "tanh", "to", "unsqueeze"], "sew_d/modeling_sew_d.py:SEWDTransformerEncoder": ["BaseModelOutput", "ConvLayer", "Embedding", "False", "LayerNorm", "Model", "ModelLayer", "Module", "ModuleList", "None", "Sequence", "True", "_", "__init__", "all_attentions", "all_hidden_states", "and", "att_m", "attention_mask", "attentions", "bucket_size", "build_relative_position", "class", "config", "conv", "conv_kernel_size", "def", "device", "dim", "elementwise_affine", "elif", "else", "enumerate", "extended_attention_mask", "for", "forward", "get_attention_mask", "get_rel_embedding", "get_rel_pos", "getattr", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "input_mask", "is", "isinstance", "last_hidden_state", "layer", "layer_module", "layer_norm", "layer_norm_eps", "len", "lower", "max_position", "max_position_embeddings", "max_relative_positions", "next_kv", "nn", "none", "norm_rel_ebd", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "output_states", "pos_ebd_size", "position_buckets", "q", "query_states", "range", "rel_embeddings", "relative_attention", "relative_pos", "return", "return_dict", "self", "size", "split", "squeeze", "strip", "sum", "super", "tuple", "unsqueeze", "v", "weight", "x"], "sew_d/modeling_sew_d.py:SEWDEncoder": ["AvgPool1d", "BaseModelOutput", "False", "Model", "ModelPositionalConvEmbedding", "ModelTransformerEncoder", "ModelUpsampling", "Module", "None", "Optional", "Tensor", "True", "__init__", "arange", "attention_ids", "attention_mask", "attentions", "bool", "class", "config", "def", "device", "dtype", "else", "encoder", "encoder_outputs", "expand", "expand_attention_mask", "for", "forward", "functional", "gradient_checkpointing", "hidden_states", "if", "in", "input_lengths", "is", "last_hidden_state", "long", "max_encoder_length", "min", "min_length", "n_input_timesteps", "nn", "not", "ones", "output_attentions", "output_hidden_states", "output_lengths", "pad", "pool", "pooled_hidden_states", "pos_conv_embed", "position_embeddings", "repeat", "return", "return_dict", "self", "shape", "size", "squeeze_factor", "sum", "super", "tensor", "torch", "transpose", "tuple", "unsqueeze", "upsample", "v", "view"], "sew_d/modeling_sew_d.py:SEWDPreTrainedModel": ["Conv1d", "Embedding", "GatheredParameters", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelPositionalConvEmbedding", "None", "PreTrainedModel", "True", "Union", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "and", "arange", "attention_mask", "base_model_prefix", "batch_size", "bias", "bool", "class", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "d", "data", "deepspeed", "def", "device", "div", "dtype", "elif", "else", "feature_vector_length", "fill_", "flip", "floor", "for", "hasattr", "if", "in", "in_channels", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "is_deepspeed_zero3_enabled", "isinstance", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "modifier_rank", "module", "nn", "normal_", "not", "output_lengths", "padding_idx", "return", "rounding_mode", "self", "sew", "shape", "sqrt", "std", "stride", "sum", "supports_gradient_checkpointing", "to", "torch", "weight", "weight_g", "weight_v", "with", "zero", "zero_", "zeros", "zip"], "sew_d/modeling_sew_d.py:SEWDModel": ["BaseModelOutput", "Dropout", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelEncoder", "ModelFeatureEncoder", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Tensor", "True", "Union", "__init__", "_compute_mask_indices", "_get_feature_vector_attention_mask", "_mask_hidden_states", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "conv_dim", "def", "device", "dtype", "elif", "else", "encoder", "encoder_outputs", "eps", "expand", "extract_features", "feat_proj_dropout", "feature_dropout", "feature_extractor", "feature_layer_norm_eps", "feature_projection", "forward", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "layer_norm", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "or", "output_attentions", "output_hidden_states", "post_init", "project_features", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "uniform_", "use_return_dict"], "sew_d/modeling_sew_d.py:SEWDForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "sew_d/modeling_sew_d.py:SEWDForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "Union", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "vivit/modeling_vivit.py:VivitTubeletEmbeddings": ["Conv3d", "False", "Image", "ModelConfig", "ModelTubeletEmbeddings", "Module", "Tensor", "ValueError", "__init__", "and", "batch_size", "bool", "class", "config", "def", "doesn", "embed_dim", "f", "flatten", "forward", "height", "hidden_size", "if", "image", "image_size", "interpolate_pos_encoding", "kernel_size", "match", "model", "nn", "not", "num_channels", "num_frames", "num_patches", "or", "patch_size", "permute", "pixel_values", "projection", "raise", "return", "self", "shape", "size", "stride", "super", "t", "torch", "transpose", "tubelet_size", "width", "x"], "vivit/modeling_vivit.py:VivitEmbeddings": ["Dropout", "False", "ModelConfig", "ModelEmbeddings", "ModelTubeletEmbeddings", "Module", "Parameter", "Tensor", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "else", "embeddings", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_frames", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "super", "tile", "torch", "torch_int", "tubelet_size", "view", "width", "zeros"], "vivit/modeling_vivit.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "vivit/modeling_vivit.py:VivitSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "vivit/modeling_vivit.py:VivitSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "vivit/modeling_vivit.py:VivitAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "vivit/modeling_vivit.py:VivitIntermediate": ["ACT2FN", "Dropout", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "else", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "vivit/modeling_vivit.py:VivitOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "vivit/modeling_vivit.py:VivitLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "vivit/modeling_vivit.py:VivitEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "range", "return", "self", "super", "torch"], "vivit/modeling_vivit.py:VivitPooler": ["Linear", "ModelConfig", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "vivit/modeling_vivit.py:VivitPreTrainedModel": ["Conv3d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "vivit/modeling_vivit.py:VivitModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "FloatTensor", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "if", "in", "interpolate_pos_encoding", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "nn", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch"], "vivit/modeling_vivit.py:VivitForVideoClassification": ["BaseModelOutput", "False", "FloatTensor", "Identity", "ImageClassifierOutput", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForVideoClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "TransformersKwargs", "Unpack", "__init__", "add_pooling_layer", "attentions", "auto_docstring", "bool", "can_return_tuple", "class", "classifier", "config", "def", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "nn", "not", "num_labels", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch"], "biogpt/modeling_biogpt.py:BioGptLearnedPositionalEmbedding": ["Embedding", "LongTensor", "ModelLearnedPositionalEmbedding", "None", "Optional", "__init__", "attention_mask", "class", "cumsum", "def", "dim", "embedding_dim", "forward", "if", "int", "is", "long", "nn", "num_embeddings", "offset", "past_key_values_length", "position_ids", "return", "self", "super", "torch"], "biogpt/modeling_biogpt.py:BioGptScaledWordEmbedding": ["Embedding", "ModelScaledWordEmbedding", "Optional", "Tensor", "__init__", "class", "def", "embed_scale", "embedding_dim", "float", "forward", "input_ids", "int", "nn", "num_embeddings", "padding_idx", "return", "self", "super", "torch"], "biogpt/modeling_biogpt.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "biogpt/modeling_biogpt.py:BioGptAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "EncoderDecoderCache", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "ValueError", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "cache_position", "caching", "call", "class", "config", "contiguous", "creating", "cross_attention_cache", "curr_past_key_value", "current_states", "decoder", "def", "deprecate_kwarg", "divisible", "dropout", "during", "eager", "eager_attention_forward", "else", "embed_dim", "errors", "f", "float", "forward", "get", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "is_updated", "isinstance", "k_proj", "key_states", "key_value_states", "keys", "kv_input_shape", "kwargs", "layer_head_mask", "layer_idx", "layers", "lead", "logger", "make", "must", "new_name", "nn", "not", "num_heads", "out_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "provide", "q_input_shape", "q_proj", "query_states", "raise", "recommended", "reshape", "return", "scaling", "self", "self_attention_cache", "shape", "src_len", "super", "sure", "tgt_len", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "used", "v_proj", "value_states", "values", "version", "view", "warning_once", "when", "will", "without"], "biogpt/modeling_biogpt.py:BioGptDecoderLayer": ["ACT2FN", "Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "__init__", "activation_dropout", "activation_fn", "attention_mask", "attention_probs_dropout_prob", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "dropout", "embed_dim", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "int", "intermediate_size", "is_causal", "is_decoder", "kwargs", "layer_head_mask", "layer_idx", "new_name", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "position_ids", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "super", "torch", "training", "tuple", "use_cache", "version"], "biogpt/modeling_biogpt.py:BioGptPreTrainedModel": ["AttentionMaskConverter", "BlockMask", "Cache", "False", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "Optional", "PreTrainedModel", "Tensor", "True", "Union", "_attn_implementation", "_can_compile_fullgraph", "_ignore_causal_mask_sdpa", "_prepare_4d_causal_attention_mask_with_cache_position", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "_unmask_unattended", "_update_causal_mask", "and", "any", "arange", "attention_mask", "base_model_prefix", "batch_size", "cache_position", "causal_mask", "class", "clone", "config", "cuda", "def", "device", "diagonal", "dim", "dtype", "elif", "else", "expand", "fill_value", "finfo", "flash", "flex_attention", "full", "get_max_cache_shape", "get_seq_length", "if", "in", "input_tensor", "inputs_embeds", "int", "is", "is_compileable", "is_training", "isinstance", "kwargs", "make_flex_block_causal_mask", "mask_length", "masked_fill", "min", "min_dtype", "not", "npu", "ones", "padding_mask", "past_key_values", "past_key_values_length", "past_seen_tokens", "reshape", "return", "sdpa", "self", "sequence_length", "shape", "size", "staticmethod", "supports_gradient_checkpointing", "target_length", "to", "torch", "training", "triu", "type", "using_compilable_cache", "xpu"], "biogpt/modeling_biogpt.py:BioGptModel": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "DynamicCache", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelLearnedPositionalEmbedding", "ModelModel", "ModelPreTrainedModel", "ModelScaledWordEmbedding", "ModuleList", "None", "Optional", "Passing", "Setting", "Tensor", "Transformers", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_update_causal_mask", "a", "all_cross_attentions", "all_hidden_states", "all_self_attns", "an", "and", "arange", "at", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bool", "both", "cache_position", "cannot", "causal_mask", "checkpointing", "class", "config", "continue", "cross_attentions", "cumsum", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "def", "deprecated", "device", "dim", "dropout", "dropout_probability", "e", "either", "elif", "else", "embed_dim", "embed_positions", "embed_scale", "embed_tokens", "enumerate", "for", "forward", "from_legacy_cache", "functional", "g", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "i", "idx", "if", "in", "incompatible", "input", "input_ids", "input_shape", "inputs_embeds", "instance", "instead", "is", "isinstance", "kwargs", "last_hidden_state", "layer_head_mask", "layer_idx", "layer_norm", "layer_outputs", "layerdrop", "layers", "logger", "long", "mask_seq_length", "math", "max_position_embeddings", "nn", "not", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "pass", "past_key_values", "past_key_values_length", "position_ids", "positions", "post_init", "raise", "rand", "range", "removed", "return", "return_dict", "same", "scale_embedding", "self", "self_attn_cache", "seq_length", "shape", "should", "size", "specify", "sqrt", "super", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "v4", "view", "vocab_size", "warning_once", "will", "with"], "biogpt/modeling_biogpt.py:BioGptForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "Model", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_tied_weights_keys", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "class", "config", "cross_attentions", "def", "else", "forward", "get_output_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "lm_loss", "logits", "loss", "loss_function", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_projection", "outputs", "past_key_values", "position_ids", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_output", "set_output_embeddings", "super", "torch", "tuple", "use_cache", "use_return_dict", "vocab_size", "weight"], "biogpt/modeling_biogpt.py:BioGptForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "__init__", "active_labels", "active_logits", "active_loss", "and", "attention_mask", "attentions", "auto_docstring", "bool", "cache_position", "class", "classifier", "classifier_dropout", "config", "def", "dropout", "else", "forward", "hasattr", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "ignore_index", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "past_key_values", "position_ids", "post_init", "r", "return", "return_dict", "self", "super", "tensor", "token_type_ids", "torch", "transformer_outputs", "tuple", "type_as", "use_cache", "use_return_dict", "view", "where"], "biogpt/modeling_biogpt.py:BioGptForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "ModelConfig", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Results", "SequenceClassifierOutputWithPast", "Tensor", "Union", "__class__", "__init__", "__name__", "and", "arange", "attention_mask", "attentions", "auto_docstring", "batch_size", "be", "bias", "bool", "cache_position", "class", "config", "conjunction", "def", "detect", "device", "dtype", "elif", "else", "embed_tokens", "f", "forward", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "labels", "logger", "logits", "long", "loss", "loss_fct", "may", "multi_label_classification", "ne", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "score", "self", "sequence_length", "set_input_embeddings", "shape", "single_label_classification", "squeeze", "sum", "super", "to", "tokens", "torch", "transformer_outputs", "tuple", "unexpected", "use_cache", "use_return_dict", "using", "value", "view", "warning_once", "will", "with"], "yolos/modeling_yolos.py:YolosObjectDetectionOutput": ["FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "attentions", "auxiliary_outputs", "class", "dict", "hidden_states", "last_hidden_state", "list", "logits", "loss", "loss_dict", "pred_boxes", "r", "torch", "tuple"], "yolos/modeling_yolos.py:YolosEmbeddings": ["Dropout", "InterpolateInitialPositionEmbeddings", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Parameter", "Tensor", "_", "__init__", "batch_size", "cat", "class", "cls_token", "cls_tokens", "config", "def", "detection_tokens", "dim", "dropout", "embeddings", "expand", "forward", "height", "hidden_dropout_prob", "hidden_size", "interpolation", "nn", "num_channels", "num_detection_tokens", "num_patches", "patch_embeddings", "pixel_values", "position_embeddings", "return", "self", "seq_len", "shape", "size", "super", "torch", "width", "zeros"], "yolos/modeling_yolos.py:InterpolateInitialPositionEmbeddings": ["False", "Model", "ModelInitialPositionEmbeddings", "Module", "None", "Tensor", "__init__", "align_corners", "batch_size", "bicubic", "cat", "class", "cls_pos_embed", "config", "def", "det_pos_embed", "dim", "flatten", "forward", "functional", "height", "hidden_size", "image_size", "img_size", "mode", "new_patch_height", "new_patch_width", "nn", "num_detection_tokens", "patch_height", "patch_pos_embed", "patch_size", "patch_width", "pos_embed", "return", "scale_pos_embed", "self", "seq_len", "shape", "size", "super", "torch", "transpose", "view", "width"], "yolos/modeling_yolos.py:InterpolateMidPositionEmbeddings": ["False", "Model", "ModelMidPositionEmbeddings", "Module", "None", "Tensor", "__init__", "align_corners", "batch_size", "bicubic", "cat", "class", "cls_pos_embed", "config", "contiguous", "def", "depth", "det_pos_embed", "dim", "flatten", "forward", "functional", "height", "hidden_size", "image_size", "img_size", "mode", "new_patch_height", "new_patch_width", "nn", "num_detection_tokens", "patch_height", "patch_pos_embed", "patch_size", "patch_width", "pos_embed", "return", "scale_pos_embed", "self", "seq_len", "shape", "size", "super", "torch", "transpose", "view", "width"], "yolos/modeling_yolos.py:YolosPatchEmbeddings": ["Conv2d", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "flatten", "forward", "height", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "width", "with"], "yolos/modeling_yolos.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "yolos/modeling_yolos.py:YolosSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "yolos/modeling_yolos.py:YolosSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "yolos/modeling_yolos.py:YolosAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "yolos/modeling_yolos.py:YolosIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "yolos/modeling_yolos.py:YolosOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "yolos/modeling_yolos.py:YolosLayer": ["GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelConfig", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "__init__", "attention", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "return", "self", "seq_len_dim", "super", "torch"], "yolos/modeling_yolos.py:YolosEncoder": ["BaseModelOutput", "False", "InterpolateMidPositionEmbeddings", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Parameter", "Tensor", "_", "__init__", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "height", "hidden_size", "hidden_states", "i", "if", "image_size", "in", "int", "interpolated_mid_position_embeddings", "interpolation", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "mid_position_embeddings", "nn", "not", "num_detection_tokens", "num_hidden_layers", "patch_size", "range", "return", "self", "seq_length", "super", "torch", "use_mid_position_embeddings", "width", "zeros"], "yolos/modeling_yolos.py:YolosPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "ModelConfig", "ModelLayer", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "hidden_states", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "vit", "weight", "zero_"], "yolos/modeling_yolos.py:YolosModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPooler", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Unpack", "ValueError", "You", "__init__", "_prune_heads", "add_pooling_layer", "attention", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "dict", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "height", "hidden_size", "if", "in", "int", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "not", "num_hidden_layers", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "shape", "specify", "super", "to", "torch", "width"], "yolos/modeling_yolos.py:YolosPooler": ["Linear", "ModelConfig", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "yolos/modeling_yolos.py:YolosMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "yolos/modeling_yolos.py:YolosForObjectDetection": ["BaseModelOutputWithPooling", "False", "FloatTensor", "ModelConfig", "ModelForObjectDetection", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "None", "Optional", "TransformersKwargs", "Unpack", "__init__", "_set_aux_loss", "a", "add_pooling_layer", "attentions", "auto_docstring", "auxiliary_loss", "auxiliary_outputs", "b", "bbox_predictor", "can_return_tuple", "class", "class_labels_classifier", "config", "def", "device", "dict", "for", "forward", "hidden_dim", "hidden_size", "hidden_states", "if", "in", "input_dim", "intermediate", "is", "jit", "kwargs", "labels", "last_hidden_state", "list", "logits", "loss", "loss_dict", "loss_function", "not", "num_detection_tokens", "num_labels", "num_layers", "output_dim", "outputs", "outputs_class", "outputs_coord", "pixel_values", "post_init", "pred_boxes", "r", "return", "self", "sequence_output", "sigmoid", "super", "torch", "unused", "vit", "zip"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "codevector_perplexity", "hidden_states", "logits", "loss", "projected_quantized_states", "projected_states", "r", "torch", "tuple"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatSamePadLayer": ["ModelSamePadLayer", "Module", "__init__", "class", "def", "else", "forward", "hidden_states", "if", "nn", "num_conv_pos_embeddings", "num_pad_remove", "return", "self", "super"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatPositionalConvEmbedding": ["ACT2FN", "Conv1d", "GatheredParameters", "ModelPositionalConvEmbedding", "ModelSamePadLayer", "Module", "__init__", "activation", "class", "config", "conv", "deepspeed", "def", "dim", "else", "feat_extract_activation", "forward", "groups", "hasattr", "hidden_size", "hidden_states", "if", "is_deepspeed_zero3_enabled", "kernel_size", "modifier_rank", "name", "nn", "num_conv_pos_embedding_groups", "num_conv_pos_embeddings", "original0", "original1", "padding", "parametrizations", "register_external_parameter", "return", "self", "super", "transpose", "utils", "weight", "weight_g", "weight_norm", "weight_v", "with", "zero"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatNoLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "ModelNoLayerNormConvLayer", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "nn", "out_conv_dim", "return", "self", "stride", "super"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatLayerNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "LayerNorm", "ModelLayerNormConvLayer", "True", "__init__", "activation", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "elementwise_affine", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "out_conv_dim", "return", "self", "stride", "super", "transpose"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatGroupNormConvLayer": ["ACT2FN", "Conv1d", "GradientCheckpointingLayer", "GroupNorm", "ModelGroupNormConvLayer", "True", "__init__", "activation", "affine", "bias", "class", "config", "conv", "conv_bias", "conv_dim", "conv_kernel", "conv_stride", "def", "else", "feat_extract_activation", "forward", "hidden_states", "if", "in_conv_dim", "kernel_size", "layer_id", "layer_norm", "nn", "num_channels", "num_groups", "out_conv_dim", "return", "self", "stride", "super"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatFeatureEncoder": ["False", "ModelFeatureEncoder", "ModelGroupNormConvLayer", "ModelLayerNormConvLayer", "ModelNoLayerNormConvLayer", "Module", "ModuleList", "None", "True", "ValueError", "__init__", "_freeze_parameters", "_requires_grad", "and", "be", "but", "class", "config", "conv_layer", "conv_layers", "def", "elif", "else", "f", "feat_extract_norm", "for", "forward", "gradient_checkpointing", "group", "has", "hidden_states", "i", "if", "in", "input_values", "is", "layer", "layer_id", "nn", "num_feat_extract_layers", "of", "one", "param", "parameters", "raise", "range", "requires_grad", "return", "self", "super", "to", "training"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatFeatureProjection": ["Dropout", "LayerNorm", "Linear", "ModelFeatureProjection", "Module", "__init__", "class", "config", "conv_dim", "def", "dropout", "eps", "feat_proj_dropout", "forward", "hidden_size", "hidden_states", "layer_norm", "layer_norm_eps", "nn", "norm_hidden_states", "projection", "return", "self", "super"], "unispeech_sat/modeling_unispeech_sat.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatFeedForward": ["ACT2FN", "Dropout", "Linear", "ModelFeedForward", "Module", "__init__", "activation_dropout", "class", "config", "def", "else", "forward", "hidden_act", "hidden_dropout", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_dense", "intermediate_dropout", "intermediate_size", "isinstance", "nn", "output_dense", "output_dropout", "return", "self", "str", "super"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatEncoderLayer": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelFeedForward", "None", "_", "__init__", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "class", "config", "def", "dropout", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "hidden_dropout", "hidden_size", "hidden_states", "if", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatEncoder": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalConvEmbedding", "Modelon", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "tensor", "torch", "training", "tuple", "unsqueeze", "v"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatAttnAdapterLayer": ["FloatTensor", "LayerNorm", "Linear", "ModelAttnAdapterLayer", "Module", "ReLU", "__init__", "act_fn", "adapter_attn_dim", "class", "config", "def", "forward", "hidden_dim", "hidden_size", "hidden_states", "input_dim", "linear_1", "linear_2", "nn", "norm", "return", "self", "super", "torch"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatEncoderLayerStableLayerNorm": ["Dropout", "False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelAttnAdapterLayer", "ModelEncoderLayerStableLayerNorm", "ModelFeedForward", "None", "Optional", "Tensor", "_", "__init__", "adapter_attn_dim", "adapter_layer", "attention", "attention_dropout", "attention_mask", "attn_residual", "attn_weights", "bool", "class", "config", "def", "dropout", "else", "embed_dim", "eps", "feed_forward", "final_layer_norm", "forward", "getattr", "hidden_dropout", "hidden_size", "hidden_states", "if", "is", "is_decoder", "layer_norm", "layer_norm_eps", "nn", "not", "num_attention_heads", "num_heads", "output_attentions", "outputs", "return", "self", "super", "torch"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatEncoderStableLayerNorm": ["BaseModelOutput", "Dropout", "False", "LayerNorm", "ModelEncoderLayerStableLayerNorm", "ModelEncoderStableLayerNorm", "ModelPositionalConvEmbedding", "Modelon", "Module", "ModuleList", "None", "Tensor", "True", "_", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "_prepare_4d_attention_mask_for_sdpa", "_update_full_mask", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "def", "dropout", "dropout_probability", "dtype", "elif", "else", "eps", "expand_attention_mask", "flash", "flex_attention", "for", "forward", "gradient_checkpointing", "hidden_dropout", "hidden_size", "hidden_states", "if", "in", "inputs_embeds", "is", "is_causal", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "isinstance", "last_hidden_state", "layer", "layer_norm", "layer_norm_eps", "layer_outputs", "layerdrop", "layers", "make_flex_block_causal_mask", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "pos_conv_embed", "position_embeddings", "rand", "range", "repeat", "return", "return_dict", "sdpa", "self", "shape", "skip_the_layer", "super", "synced_gpus", "torch", "training", "tuple", "unsqueeze", "v"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatGumbelVectorQuantizer": ["FloatTensor", "Linear", "ModelGumbelVectorQuantizer", "Module", "None", "Parameter", "True", "ValueError", "__init__", "_compute_perplexity", "argmax", "batch_size", "be", "by", "class", "codevector_dim", "codevector_idx", "codevector_probs", "codevector_soft_dist", "codevectors", "codevectors_per_group", "concatenation", "config", "def", "dim", "divisible", "else", "exp", "f", "float", "for", "forward", "functional", "gumbel_softmax", "hard", "hidden_size", "hidden_states", "if", "log", "marginal_probs", "mask", "mean", "must", "new_zeros", "nn", "num_codevector_groups", "num_codevectors_per_group", "num_groups", "num_vars", "perplexity", "probs", "raise", "return", "scatter_", "self", "sequence_length", "shape", "softmax", "staticmethod", "sum", "super", "tau", "temperature", "torch", "training", "type_as", "unsqueeze", "view", "weight_proj"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatPreTrainedModel": ["Conv1d", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "Model", "ModelConfig", "ModelFeatureProjection", "ModelGumbelVectorQuantizer", "ModelPositionalConvEmbedding", "ModelPreTrainedModel", "Modelform_", "Modelon", "None", "PreTrainedModel", "True", "_conv_out_length", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "a", "arange", "attention_mask", "b", "base_model_prefix", "batch_size", "bias", "bool", "class", "codevectors", "config", "constant_", "conv", "conv_kernel", "conv_stride", "cumsum", "data", "def", "device", "dim", "div", "dtype", "elif", "feature_vector_length", "fill_", "flip", "floor", "for", "groups", "if", "in", "in_channels", "in_features", "init", "initializer_range", "input_length", "input_lengths", "input_values", "int", "is", "isinstance", "k", "kaiming_normal_", "kernel_size", "long", "main_input_name", "math", "mean", "module", "nn", "non_padded_lengths", "normal_", "not", "output_lengths", "projection", "return", "rounding_mode", "self", "shape", "sqrt", "std", "stride", "supports_gradient_checkpointing", "to", "torch", "weight", "weight_proj", "zero_", "zeros", "zip"], "unispeech_sat/modeling_unispeech_sat.py:_compute_mask_indices": ["False", "LongTensor", "None", "Optional", "ValueError", "_", "_compute_mask_indices", "and", "append", "arange", "array", "attention_mask", "batch_size", "be", "bigger", "bool", "broadcast_to", "but", "choice", "compute_num_masked_span", "concatenate", "def", "detach", "dtype", "dummy_mask_idx", "else", "epsilon", "f", "float", "for", "got", "has", "if", "in", "input_length", "input_lengths", "int", "int32", "is", "item", "len", "mask_length", "mask_prob", "max", "max_num_masked_span", "min_masks", "ndarray", "not", "np", "num_masked_span", "offsets", "ones", "put_along_axis", "raise", "rand", "random", "range", "replace", "reshape", "return", "sequence_length", "shape", "smaller", "spec_aug_mask", "spec_aug_mask_idx", "spec_aug_mask_idxs", "sum", "than", "to", "tolist", "torch", "tuple", "zeros"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatModel": ["FloatTensor", "LongTensor", "ModelBaseModelOutput", "ModelConfig", "ModelEncoder", "ModelEncoderStableLayerNorm", "ModelFeatureEncoder", "ModelFeatureProjection", "ModelModel", "ModelPreTrainedModel", "Modelform_", "Modelon", "None", "Optional", "Parameter", "Tensor", "True", "__init__", "_compute_mask_indices", "_get_feature_vector_attention_mask", "_mask_hidden_states", "and", "apply_spec_augment", "attention_mask", "attentions", "auto_docstring", "batch_size", "bool", "class", "config", "def", "device", "do_stable_layer_norm", "dtype", "elif", "else", "encoder", "encoder_outputs", "expand", "extract_features", "feature_extractor", "feature_projection", "forward", "getattr", "hidden_size", "hidden_states", "if", "input_values", "is", "last_hidden_state", "mask_feature_indices", "mask_feature_length", "mask_feature_min_masks", "mask_feature_prob", "mask_length", "mask_prob", "mask_time_indices", "mask_time_length", "mask_time_min_masks", "mask_time_prob", "masked_spec_embed", "min_masks", "nn", "not", "output_attentions", "output_hidden_states", "post_init", "r", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "use_return_dict"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatForPreTraining": ["Dropout", "False", "FloatTensor", "FutureWarning", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelGumbelVectorQuantizer", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Parameter", "Please", "Tensor", "The", "Transformers", "__init__", "_freeze_parameters", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cat", "class", "codevector_dim", "codevector_perplexity", "compute_contrastive_logits", "config", "cosine_similarity", "data", "def", "deprecated", "dim", "do_stable_layer_norm", "dropout", "dropout_features", "else", "eps", "equivalent", "extract_features", "feat_quantizer_dropout", "feature_extractor", "final_dropout", "float", "forward", "freeze_feature_encoder", "freeze_feature_extractor", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "int", "is", "label_embeddings_concat", "layer_norm_eps", "layer_norm_for_extract", "logits", "loss", "method", "negative_features", "nn", "not", "num_clusters", "output_attentions", "output_hidden_states", "outputs", "post_init", "predicted_features", "proj_codevector_dim", "project_hid", "project_q", "projected_quantized_states", "projected_states", "quantized_features", "quantizer", "r", "removed", "requires_grad", "return", "return_dict", "self", "set_gumbel_temperature", "speaker_proj", "staticmethod", "super", "target_features", "temperature", "the", "torch", "transformer_features", "tuple", "type_as", "use", "use_return_dict", "v5", "warn", "warnings", "will", "zero_"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatForCTC": ["By", "Cannot", "CausalLMOutput", "Dropout", "False", "FutureWarning", "Label", "Linear", "Model", "ModelForCTC", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Please", "Tensor", "The", "Transformers", "True", "ValueError", "You", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_freeze_parameters", "_get_feat_extract_output_lengths", "a", "adapter_attn_dim", "add_adapter", "and", "are", "as", "attention_mask", "attentions", "auto_docstring", "backends", "be", "blank", "bool", "class", "config", "configuration", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "default", "define", "defined", "deprecated", "dim", "does", "dropout", "dtype", "elif", "else", "enabled", "eng", "equivalent", "f", "feature_extractor", "final_dropout", "flags", "flattened_targets", "float32", "follows", "for", "force_load", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "from_pretrained", "functional", "getattr", "hasattr", "head", "hidden_size", "hidden_states", "if", "in", "info", "input_lengths", "input_values", "instantiate", "instead", "is", "labels", "labels_mask", "language", "lm_head", "load_adapter", "log_probs", "log_softmax", "logger", "logits", "long", "loss", "masked_select", "max", "method", "model", "must", "nn", "not", "of", "ones_like", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "outputs", "pad_token_id", "param", "parameters", "pass", "post_init", "r", "raise", "reduction", "removed", "requires_grad", "return", "return_dict", "s", "self", "set", "size", "str", "sum", "super", "target_lang", "target_lengths", "that", "the", "tie_weights", "to", "torch", "transpose", "trying", "tuple", "use", "use_return_dict", "v5", "values", "vocab_size", "vocabulary", "warn", "warnings", "will", "with", "your", "zero_infinity"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatForSequenceClassification": ["CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Parameter", "Please", "Sequence", "SequenceClassifierOutput", "Tensor", "The", "Transformers", "True", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "_get_feature_vector_attention_mask", "adapters", "add_adapter", "and", "attention_mask", "attentions", "auto_docstring", "be", "bool", "class", "classification", "classifier", "classifier_proj_size", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "expand_padding_mask", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "mean", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "padding_mask", "param", "parameters", "pooled_output", "post_init", "projector", "r", "raise", "removed", "repeat", "requires_grad", "return", "return_dict", "self", "shape", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "unsqueeze", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatForAudioFrameClassification": ["Audio", "CrossEntropyLoss", "False", "FutureWarning", "Linear", "Model", "ModelForAudioFrameClassification", "ModelModel", "ModelPreTrainedModel", "Modelon", "None", "Optional", "Parameter", "Please", "Tensor", "The", "TokenClassifierOutput", "Transformers", "True", "ValueError", "_HIDDEN_STATES_START_POSITION", "__init__", "_freeze_parameters", "adapters", "add_adapter", "and", "argmax", "attention_mask", "attentions", "auto_docstring", "axis", "be", "bool", "class", "classification", "classifier", "config", "def", "deprecated", "dim", "does", "else", "equivalent", "feature_extractor", "for", "forward", "frame", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hasattr", "hidden_size", "hidden_states", "if", "in", "init_weights", "input_values", "instead", "is", "labels", "layer_weights", "logits", "loss", "loss_fct", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "of", "ones", "output", "output_attentions", "output_hidden_states", "outputs", "param", "parameters", "r", "raise", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "sum", "super", "support", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will"], "unispeech_sat/modeling_unispeech_sat.py:AMSoftmaxLoss": ["CrossEntropyLoss", "Model", "Module", "Parameter", "True", "__init__", "bool", "class", "cos_theta", "def", "dim", "flatten", "forward", "functional", "hidden_states", "input_dim", "labels", "logits", "loss", "margin", "mm", "nn", "normalize", "num_labels", "one_hot", "onehot", "psi", "randn", "requires_grad", "return", "scale", "self", "super", "torch", "weight", "where"], "unispeech_sat/modeling_unispeech_sat.py:TDNNLayer": ["Detected", "Linear", "LoRA", "LoraLayer", "Model", "Module", "ReLU", "Tensor", "You", "__init__", "activation", "applied", "be", "bias", "class", "config", "conv1d", "def", "dilation", "due", "else", "exclude", "forward", "from", "functional", "hidden_states", "if", "in_conv_dim", "is_peft_available", "isinstance", "kernel", "kernel_size", "layer_id", "modules", "nn", "on", "optimization", "out_conv_dim", "return", "s", "self", "should", "super", "t", "target", "tdnn_dilation", "tdnn_dim", "tdnn_kernel", "to", "torch", "transpose", "view", "warn", "warnings", "weight", "weights", "won"], "unispeech_sat/modeling_unispeech_sat.py:UniSpeechSatForXVector": ["AMSoftmaxLoss", "False", "FutureWarning", "Linear", "LongTensor", "Model", "ModelForXVector", "ModelModel", "ModelPreTrainedModel", "Modelon", "ModuleList", "None", "Optional", "Parameter", "Please", "TDNNLayer", "Tensor", "The", "Transformers", "True", "XVectorOutput", "_HIDDEN_STATES_START_POSITION", "__init__", "_conv_out_length", "_freeze_parameters", "_get_feat_extract_output_lengths", "_get_tdnn_output_lengths", "and", "append", "attention_mask", "attentions", "auto_docstring", "be", "bool", "cat", "class", "classifier", "config", "def", "deprecated", "dim", "else", "embeddings", "enumerate", "equivalent", "feat_extract_output_lengths", "feature_extractor", "for", "forward", "freeze_base_model", "freeze_feature_encoder", "freeze_feature_extractor", "functional", "hidden_size", "hidden_states", "i", "if", "in", "init_weights", "input_length", "input_lengths", "input_values", "instead", "int", "is", "kernel_size", "labels", "layer_weights", "len", "length", "logits", "loss", "mean", "mean_features", "method", "nn", "norm_weights", "not", "num_hidden_layers", "num_labels", "num_layers", "objective", "ones", "output", "output_attentions", "output_embeddings", "output_hidden_states", "outputs", "param", "parameters", "projector", "r", "range", "removed", "requires_grad", "return", "return_dict", "self", "softmax", "stack", "statistic_pooling", "std", "std_features", "stride", "sum", "super", "tdnn", "tdnn_dim", "tdnn_kernel", "tdnn_layer", "tdnn_layers", "tdnn_output_lengths", "the", "torch", "tuple", "use", "use_return_dict", "use_weighted_layer_sum", "v5", "view", "warn", "warnings", "will", "xvector_output_dim"], "patchtst/modeling_patchtst.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "float", "functional", "head_mask", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "size", "softmax", "torch", "training", "transpose", "value", "view"], "patchtst/modeling_patchtst.py:PatchTSTAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "FlashAttentionKwargs", "Linear", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "ValueError", "__init__", "_attn_implementation", "and", "attention_interface", "attention_mask", "attn_output", "attn_weights", "be", "bias", "bool", "bsz", "by", "class", "config", "contiguous", "current_states", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "float", "forward", "got", "head_dim", "head_mask", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "kv_input_shape", "kwargs", "layer_head_mask", "must", "nn", "not", "num_heads", "out_proj", "output_attentions", "q_input_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "shape", "src_len", "super", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view"], "patchtst/modeling_patchtst.py:PatchTSTBatchNorm": ["BatchNorm1d", "ModelBatchNorm", "ModelConfig", "Module", "Tensor", "__init__", "batchnorm", "class", "config", "d_model", "def", "eps", "forward", "inputs", "nn", "norm_eps", "output", "return", "self", "super", "torch", "transpose"], "patchtst/modeling_patchtst.py:random_masking": ["False", "Mask", "Model_masking", "None", "Optional", "Tensor", "ValueError", "and", "argsort", "batch_size", "be", "between", "bool", "channel_consistent_masking", "def", "device", "dim", "else", "f", "float", "gather", "has", "ids_restore", "ids_shuffle", "if", "index", "inputs", "inputs_mask", "int", "is", "len_keep", "list", "mask", "mask_ratio", "mask_value", "masked_fill", "noise", "not", "num_channels", "num_features", "ones", "or", "raise", "rand", "ratio", "repeat", "return", "sequence_length", "shape", "to", "torch", "unmasked_channel_indices", "unsqueeze"], "patchtst/modeling_patchtst.py:forecast_masking": ["Model_mask_ratios", "Model_masking", "None", "Optional", "Tensor", "Union", "ValueError", "_", "and", "append", "batch1", "batch2", "batch_size", "be", "bool", "def", "device", "elif", "f", "for", "greater", "if", "in", "inputs", "inputs_mask", "int", "is", "isinstance", "key", "lambda", "less", "list", "mask", "mask_value", "masked_fill", "not", "num_Model_mask_patches", "num_channels", "num_features", "or", "patch_len", "patch_length", "patches", "perm", "raise", "randperm", "ratio", "repeat", "return", "sequence_length", "shape", "should", "sorted", "sum", "t_list", "temp_len", "than", "torch", "total", "total_length", "total_ratio", "unmasked_channel_indices", "unsqueeze", "x", "zeros", "zip"], "patchtst/modeling_patchtst.py:PatchTSTPatchify": ["Input", "Model", "ModelConfig", "ModelModelify", "Model_length", "Model_stride", "Module", "Sequence", "Tensor", "ValueError", "__init__", "be", "class", "config", "configuration", "context_length", "contiguous", "def", "dimension", "doesn", "f", "forward", "greater", "has", "if", "length", "match", "max", "model", "new_sequence_length", "nn", "num_Modeles", "output", "past_values", "raise", "return", "self", "sequence", "sequence_length", "sequence_start", "shape", "size", "step", "super", "t", "than", "the", "to", "torch", "transpose", "unfold"], "patchtst/modeling_patchtst.py:PatchTSTMasking": ["Invalid", "ModelConfig", "ModelMasking", "Model_input", "Module", "None", "Tensor", "ValueError", "__init__", "bool", "channel_consistent_masking", "class", "config", "def", "elif", "else", "f", "forecast", "forecast_masking", "forward", "if", "inputs", "is", "mask", "mask_ratio", "mask_type", "mask_value", "masked_input", "nn", "not", "num_forecast_mask_Modeles", "raise", "random", "random_mask_ratio", "random_masking", "return", "self", "sorted", "super", "torch", "type", "unmasked_channel_indices"], "patchtst/modeling_patchtst.py:PatchTSTEncoderLayer": ["ACT2CLS", "Dropout", "Identity", "LayerNorm", "Linear", "ModelAttention", "ModelBatchNorm", "ModelConfig", "ModelEncoderLayer", "Module", "None", "Optional", "Sequential", "Tensor", "ValueError", "_", "__init__", "a", "activation_function", "attention_dropout", "attn_output", "attn_weights", "batch_size", "batchnorm", "bias", "bool", "channel_attention", "channel_attn_weights", "class", "config", "contiguous", "d_model", "def", "dropout", "dropout_path1", "dropout_path2", "dropout_path3", "elif", "else", "embed_dim", "eps", "f", "ff", "ff_dropout", "ffn_dim", "forward", "hidden_state", "hidden_states", "if", "is", "layer", "layernorm", "nn", "norm", "norm_eps", "norm_sublayer1", "norm_sublayer2", "norm_sublayer3", "norm_type", "not", "num_attention_heads", "num_heads", "num_input_channels", "output_attentions", "outputs", "path_dropout", "pre_norm", "raise", "reshape", "return", "self", "self_attn", "sequence_length", "shape", "super", "supported", "torch", "transpose", "type", "view"], "patchtst/modeling_patchtst.py:PatchTSTPreTrainedModel": ["False", "LayerNorm", "Linear", "ModelBatchNorm", "ModelConfig", "ModelEncoder", "ModelPositionalEncoding", "ModelPreTrainedModel", "Model_length", "Model_stride", "Module", "None", "PreTrainedModel", "_init_pe", "_init_weights", "_set_gradient_checkpointing", "base_model_prefix", "batchnorm", "bias", "class", "cls_token", "config", "context_length", "data", "def", "elif", "fill_", "gradient_checkpointing", "if", "init", "init_std", "is", "isinstance", "main_input_name", "max", "mean", "model", "module", "nn", "normal_", "not", "num_Modeles", "past_values", "position_enc", "self", "std", "supports_gradient_checkpointing", "use_cls_token", "value", "weight", "zero_"], "patchtst/modeling_patchtst.py:PatchTSTEmbedding": ["Linear", "ModelConfig", "ModelEmbedding", "Model_input", "Model_length", "Module", "ModuleList", "Tensor", "The", "ValueError", "_", "__init__", "append", "as", "batch", "be", "channels", "class", "config", "d_model", "def", "defined", "dim", "else", "embeddings", "f", "for", "forward", "has", "i", "if", "in", "input", "input_embedding", "nn", "num_input_channels", "number", "of", "raise", "range", "return", "same", "self", "shape", "share_embedding", "stack", "super", "the", "to", "torch"], "patchtst/modeling_patchtst.py:PatchTSTPositionalEncoding": ["Available", "Dropout", "False", "Identity", "ModelConfig", "ModelPositionalEncoding", "Model_input", "Module", "Parameter", "Tensor", "True", "ValueError", "__init__", "_init_pe", "a", "and", "arange", "are", "cat", "class", "cls_token", "cls_tokens", "config", "cos", "d_model", "def", "dim", "div_term", "elif", "else", "encoder", "exp", "expand", "f", "forward", "hidden_state", "if", "int", "is", "log", "math", "mean", "nn", "not", "num_Modeles", "num_input_channels", "position", "position_enc", "positional", "positional_dropout", "positional_encoding_type", "raise", "randn", "random", "requires_grad", "return", "self", "shape", "sin", "sincos", "staticmethod", "std", "super", "torch", "types", "unsqueeze", "use_cls_token", "valid", "zeros"], "patchtst/modeling_patchtst.py:PatchTSTEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEmbedding", "ModelEncoder", "ModelEncoderLayer", "ModelPositionalEncoding", "ModelPreTrainedModel", "Model_input", "ModuleList", "None", "Optional", "Tensor", "__init__", "all_attentions", "attentions", "bool", "class", "config", "def", "else", "embedder", "encoder_layer", "encoder_states", "for", "forward", "gradient_checkpointing", "hidden_state", "hidden_states", "i", "if", "in", "int", "is", "last_hidden_state", "layer_outputs", "layers", "nn", "not", "num_Modeles", "num_hidden_layers", "output_attentions", "output_hidden_states", "positional_encoder", "post_init", "range", "return", "self", "super", "torch"], "patchtst/modeling_patchtst.py:PatchTSTModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "Model_input", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "loc", "mask", "r", "scale", "torch", "tuple"], "patchtst/modeling_patchtst.py:PatchTSTForPretrainingOutput": ["FloatTensor", "ModelForPretrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_output", "r", "torch", "tuple"], "patchtst/modeling_patchtst.py:PatchTSTForRegressionOutput": ["FloatTensor", "ModelForRegressionOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "r", "regression_outputs", "torch", "tuple"], "patchtst/modeling_patchtst.py:PatchTSTForPredictionOutput": ["FloatTensor", "ModelForPredictionOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loc", "loss", "prediction_outputs", "r", "scale", "torch", "tuple"], "patchtst/modeling_patchtst.py:PatchTSTForClassificationOutput": ["FloatTensor", "ModelForClassificationOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "r", "torch", "tuple"], "patchtst/modeling_patchtst.py:SamplePatchTSTOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "r", "sequences", "torch"], "patchtst/modeling_patchtst.py:nll": ["Distribution", "Model", "Tensor", "def", "distributions", "input", "log_prob", "return", "target", "torch"], "patchtst/modeling_patchtst.py:weighted_average": ["Model_average", "Model_tensor", "None", "Optional", "Tensor", "clamp", "def", "dim", "else", "if", "input_tensor", "is", "mean", "min", "not", "return", "sum", "sum_weights", "torch", "weights", "where", "zeros_like"], "patchtst/modeling_patchtst.py:PatchTSTStdScaler": ["ModelConfig", "ModelStdScaler", "Module", "Tensor", "True", "__init__", "clamp_min", "class", "config", "data", "def", "denominator", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "minimum_scale", "nn", "observed_indicator", "return", "scale", "scaling_dim", "self", "sqrt", "sum", "super", "torch", "tuple", "variance"], "patchtst/modeling_patchtst.py:PatchTSTMeanScaler": ["ModelConfig", "ModelMeanScaler", "Module", "None", "Tensor", "True", "__init__", "abs", "batch_observations", "batch_sum", "clamp", "class", "config", "data", "def", "default_scale", "dim", "else", "forward", "hasattr", "if", "is", "keepdim", "min", "minimum_scale", "nn", "not", "num_observed", "observed_indicator", "ones_like", "return", "scale", "scaled_data", "scaling_dim", "self", "squeeze", "sum", "super", "torch", "ts_sum", "tuple", "where", "zeros_like"], "patchtst/modeling_patchtst.py:PatchTSTNOPScaler": ["False", "ModelConfig", "ModelNOPScaler", "Module", "None", "Optional", "Tensor", "True", "__init__", "class", "config", "data", "def", "dim", "else", "forward", "hasattr", "if", "keepdim", "loc", "mean", "nn", "observed_indicator", "ones_like", "requires_grad", "return", "scale", "scaling_dim", "self", "super", "torch", "tuple", "zeros_like"], "patchtst/modeling_patchtst.py:PatchTSTScaler": ["ModelConfig", "ModelMeanScaler", "ModelNOPScaler", "ModelScaler", "ModelStdScaler", "Module", "Tensor", "True", "__init__", "class", "config", "data", "def", "elif", "else", "forward", "if", "is", "loc", "mean", "nn", "observed_indicator", "or", "return", "scale", "scaler", "scaling", "self", "std", "super", "torch", "tuple"], "patchtst/modeling_patchtst.py:PatchTSTModel": ["Identity", "ModelConfig", "ModelEncoder", "ModelMasking", "ModelModel", "ModelModelOutput", "ModelModelify", "ModelPreTrainedModel", "ModelScaler", "Model_input", "Modeled_values", "Modelifier", "None", "Optional", "Tensor", "Union", "__init__", "attentions", "bool", "class", "config", "def", "do_mask_input", "else", "encoder", "encoder_output", "for", "forward", "future_values", "hidden_states", "if", "in", "is", "last_hidden_state", "loc", "mask", "masked_values", "masking", "nn", "not", "num_Modeles", "ones_like", "output_attentions", "output_hidden_states", "outputs", "past_observed_mask", "past_values", "post_init", "r", "return", "return_dict", "scale", "scaled_past_values", "scaler", "self", "super", "torch", "tuple", "use_return_dict", "v"], "patchtst/modeling_patchtst.py:PatchTSTMaskPretrainHead": ["Dropout", "Identity", "Linear", "ModelConfig", "ModelMaskPretrainHead", "Model_length", "Module", "Tensor", "__init__", "class", "config", "d_model", "def", "dropout", "else", "embedding", "forward", "head_dropout", "if", "linear", "nn", "return", "self", "super", "torch", "use_cls_token"], "patchtst/modeling_patchtst.py:PatchTSTForPretraining": ["MSELoss", "ModelConfig", "ModelForPretraining", "ModelForPretrainingOutput", "ModelMaskPretrainHead", "ModelModel", "ModelPreTrainedModel", "Model_input", "None", "Optional", "Tensor", "True", "Union", "__init__", "attentions", "bool", "class", "config", "def", "dim", "do_mask_input", "else", "encoder_states", "forward", "head", "hidden_states", "if", "is", "last_hidden_state", "loss", "loss_val", "mask", "masked_loss", "mean", "model", "model_output", "nn", "none", "not", "output_attentions", "output_hidden_states", "outputs", "past_observed_mask", "past_values", "post_init", "prediction_output", "r", "reduction", "return", "return_dict", "self", "sum", "super", "torch", "tuple", "use_return_dict", "x_hat"], "patchtst/modeling_patchtst.py:PatchTSTClassificationHead": ["Dropout", "Flatten", "Identity", "Linear", "ModelClassificationHead", "ModelConfig", "Module", "Tensor", "ValueError", "__init__", "class", "config", "d_model", "def", "dim", "dropout", "elif", "else", "embedding", "f", "flatten", "forward", "head_dropout", "if", "implemented", "is", "linear", "max", "mean", "nn", "not", "num_input_channels", "num_targets", "operator", "output", "pooled_embedding", "pooling", "pooling_type", "raise", "return", "self", "start_dim", "super", "torch", "use_cls_token", "values", "yet"], "patchtst/modeling_patchtst.py:PatchTSTForClassification": ["CrossEntropyLoss", "False", "ModelClassificationHead", "ModelConfig", "ModelForClassification", "ModelForClassificationOutput", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Setting", "Tensor", "True", "Union", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "do_mask_input", "else", "forward", "head", "hidden_states", "if", "is", "last_hidden_state", "logger", "loss", "loss_val", "model", "model_output", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "parameter", "past_observed_mask", "past_values", "post_init", "prediction_logits", "r", "return", "return_dict", "self", "super", "target_values", "to", "torch", "tuple", "use_return_dict", "warning", "y_hat"], "patchtst/modeling_patchtst.py:PatchTSTPredictionHead": ["Dropout", "Flatten", "Identity", "Linear", "ModelConfig", "ModelPredictionHead", "Module", "ModuleList", "None", "Tensor", "__init__", "append", "class", "config", "d_model", "def", "dim", "distribution_output", "dropout", "dropouts", "elif", "else", "embedding", "flatten", "flattens", "for", "forward", "get_parameter_projection", "head_dim", "head_dropout", "i", "if", "in", "int", "is", "isinstance", "max", "mean", "nn", "not", "num_Modeles", "num_input_channels", "or", "output", "pooled_embedding", "pooling_type", "prediction_length", "projection", "projections", "r", "range", "return", "self", "share_projection", "stack", "start_dim", "super", "torch", "transpose", "tuple", "use_cls_token", "values", "z"], "patchtst/modeling_patchtst.py:PatchTSTForPrediction": ["False", "MSELoss", "ModelConfig", "ModelForPrediction", "ModelForPredictionOutput", "ModelModel", "ModelPreTrainedModel", "ModelPredictionHead", "Modelifier", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleModelOutput", "Setting", "StudentTOutput", "Tensor", "True", "Union", "Unknown", "ValueError", "_", "__init__", "attentions", "bool", "class", "config", "def", "dim", "distribution", "distribution_output", "do_mask_input", "elif", "else", "f", "for", "forward", "future_values", "generate", "head", "hidden_states", "if", "in", "is", "last_hidden_state", "loc", "logger", "loss", "loss_val", "mean", "model", "model_output", "mse", "negative_binomial", "nll", "nn", "no_grad", "normal", "not", "num_Modeles", "num_parallel_samples", "output", "output_attentions", "output_hidden_states", "outputs", "parameter", "past_observed_mask", "past_values", "post_init", "prediction_length", "prediction_outputs", "r", "raise", "range", "reduction", "return", "return_dict", "sample", "samples", "scale", "self", "sequences", "stack", "student_t", "super", "to", "torch", "tuple", "unsqueeze", "use_return_dict", "warning", "weighted_average", "y_hat", "y_hat_out"], "patchtst/modeling_patchtst.py:PatchTSTRegressionHead": ["Dropout", "Flatten", "Identity", "Linear", "ModelConfig", "ModelRegressionHead", "Module", "None", "Tensor", "ValueError", "__init__", "class", "config", "d_model", "def", "dim", "distribution_output", "dropout", "elif", "else", "embedding", "f", "flatten", "forward", "get_parameter_projection", "head_dim", "head_dropout", "if", "implemented", "is", "max", "mean", "nn", "not", "num_input_channels", "num_targets", "operator", "output", "output_range", "pooled_embedding", "pooling", "pooling_type", "projection", "raise", "return", "self", "sigmoid", "start_dim", "super", "torch", "use_cls_token", "values", "y_range", "yet"], "patchtst/modeling_patchtst.py:PatchTSTForRegression": ["False", "MSELoss", "ModelConfig", "ModelForRegression", "ModelForRegressionOutput", "ModelModel", "ModelPreTrainedModel", "ModelRegressionHead", "NegativeBinomialOutput", "None", "NormalOutput", "Optional", "SampleModelOutput", "Setting", "StudentTOutput", "Tensor", "True", "Union", "Unknown", "ValueError", "_", "__init__", "attentions", "auto_docstring", "bool", "class", "config", "def", "dim", "distribution", "distribution_output", "do_mask_input", "elif", "else", "f", "for", "forward", "generate", "head", "hidden_states", "if", "in", "is", "item", "last_hidden_state", "logger", "loss", "mean", "model", "model_output", "mse", "negative_binomial", "nll", "nn", "no_grad", "normal", "not", "num_parallel_samples", "num_targets", "output", "output_attentions", "output_hidden_states", "outputs", "parameter", "past_observed_mask", "past_values", "post_init", "r", "raise", "range", "reduction", "regression_outputs", "return", "return_dict", "sample", "samples", "self", "sequences", "stack", "student_t", "super", "target_values", "to", "torch", "tuple", "use_return_dict", "view", "warning", "weighted_average", "y_hat"], "siglip/modeling_siglip.py:_trunc_normal_": ["The", "_trunc_normal_", "a", "add_", "b", "be", "clamp_", "def", "distribution", "erf", "erfinv_", "from", "if", "in", "incorrect", "init", "is", "l", "math", "max", "may", "mean", "min", "more", "mul_", "nn", "norm_cdf", "of", "or", "return", "sqrt", "stacklevel", "std", "tensor", "than", "trunc_normal_", "u", "uniform_", "values", "warn", "warnings", "x"], "siglip/modeling_siglip.py:trunc_normal_tf_": ["Model_normal_tf_", "Tensor", "_Model_normal_", "a", "add_", "b", "def", "float", "mean", "mul_", "no_grad", "std", "tensor", "torch", "with"], "siglip/modeling_siglip.py:variance_scaling_": ["Model", "Model_scaling_", "ValueError", "_calculate_fan_in_and_fan_out", "bound", "def", "denom", "distribution", "elif", "else", "f", "fan_avg", "fan_in", "fan_out", "if", "invalid", "math", "mode", "no_grad", "normal", "normal_", "raise", "scale", "sqrt", "std", "tensor", "torch", "trunc_normal_tf_", "truncated_normal", "uniform", "uniform_", "with"], "siglip/modeling_siglip.py:lecun_normal_": ["Model_normal_", "def", "distribution", "fan_in", "mode", "tensor", "truncated_normal", "variance_scaling_"], "siglip/modeling_siglip.py:default_flax_embed_init": ["Model_flax_embed_init", "def", "distribution", "fan_in", "mode", "normal", "tensor", "variance_scaling_"], "siglip/modeling_siglip.py:SiglipVisionModelOutput": ["FloatTensor", "ModelOutput", "ModelVisionModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "image_embeds", "last_hidden_state", "r", "torch", "tuple"], "siglip/modeling_siglip.py:SiglipTextModelOutput": ["FloatTensor", "ModelOutput", "ModelTextModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "r", "text_embeds", "torch", "tuple"], "siglip/modeling_siglip.py:SiglipOutput": ["Any", "BaseModelOutputWithPooling", "FloatTensor", "ModelOutput", "None", "Optional", "class", "def", "else", "for", "getattr", "if", "image_embeds", "in", "k", "keys", "logits_per_image", "logits_per_text", "loss", "not", "r", "return", "self", "text_embeds", "text_model_output", "to_tuple", "torch", "tuple", "vision_model_output"], "siglip/modeling_siglip.py:SiglipVisionEmbeddings": ["Conv2d", "Embedding", "False", "FloatTensor", "ModelVisionConfig", "ModelVisionEmbeddings", "Module", "Tensor", "_", "__init__", "align_corners", "and", "arange", "bicubic", "class", "config", "def", "dim", "dtype", "else", "embed_dim", "embeddings", "expand", "flatten", "forward", "functional", "height", "hidden_size", "if", "image_size", "in_channels", "int", "interpolate", "interpolate_pos_encoding", "is_tracing", "jit", "kernel_size", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "out_channels", "padding", "patch_embedding", "patch_embeds", "patch_pos_embed", "patch_size", "permute", "persistent", "pixel_values", "position_embedding", "position_ids", "register_buffer", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "stride", "super", "target_dtype", "to", "torch", "torch_int", "transpose", "unsqueeze", "valid", "view", "weight", "width"], "siglip/modeling_siglip.py:SiglipTextEmbeddings": ["Embedding", "False", "FloatTensor", "LongTensor", "ModelTextConfig", "ModelTextEmbeddings", "Module", "None", "Optional", "Sequence", "Tensor", "ValueError", "__init__", "and", "arange", "be", "class", "config", "def", "else", "embed_dim", "embeddings", "expand", "f", "forward", "got", "hidden_size", "if", "input_ids", "inputs_embeds", "is", "length", "less", "max_position_embedding", "max_position_embeddings", "must", "nn", "not", "persistent", "position_embedding", "position_embeddings", "position_ids", "raise", "register_buffer", "return", "self", "seq_length", "sequence", "shape", "super", "than", "token_embedding", "torch", "vocab_size", "weight"], "siglip/modeling_siglip.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "siglip/modeling_siglip.py:SiglipAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelAttention", "Module", "None", "Optional", "Tensor", "ValueError", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "by", "class", "config", "contiguous", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "f", "forward", "got", "head_dim", "hidden_size", "hidden_states", "if", "is_causal", "k_proj", "keys", "kwargs", "must", "nn", "not", "num_attention_heads", "num_heads", "out_proj", "q_proj", "queries", "raise", "reshape", "return", "scale", "scaling", "self", "seq_length", "shape", "super", "torch", "training", "transpose", "tuple", "v_proj", "values", "view"], "siglip/modeling_siglip.py:SiglipMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "Tensor", "__init__", "activation_fn", "class", "config", "def", "fc1", "fc2", "forward", "hidden_act", "hidden_size", "hidden_states", "intermediate_size", "nn", "return", "self", "super", "torch"], "siglip/modeling_siglip.py:SiglipEncoderLayer": ["FloatTensor", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelEncoderLayer", "ModelMLP", "ModelTextConfig", "ModelVisionConfig", "Tensor", "TransformersKwargs", "Union", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "embed_dim", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "layer_norm1", "layer_norm2", "layer_norm_eps", "mlp", "nn", "residual", "return", "self", "self_attn", "super", "torch"], "siglip/modeling_siglip.py:SiglipPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelAttention", "ModelConfig", "ModelEncoderLayer", "ModelForImageClassification", "ModelMLP", "ModelModel", "ModelMultiheadAttentionPoolingHead", "ModelPreTrainedModel", "ModelTextEmbeddings", "ModelVisionEmbeddings", "None", "PreTrainedModel", "True", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attention", "attentions", "base_model_prefix", "bias", "class", "classifier", "config", "data", "def", "default_flax_embed_init", "elif", "else", "fc1", "fc2", "fill_", "hidden_size", "hidden_states", "if", "in_proj_bias", "in_proj_weight", "init", "initializer_factor", "is", "isinstance", "k_proj", "lecun_normal_", "log", "logit_bias", "logit_scale", "logit_scale_init", "module", "nn", "normal_", "not", "np", "out_proj", "position_embedding", "probe", "q_proj", "self", "sqrt", "std", "supports_gradient_checkpointing", "tensor", "torch", "v_proj", "vision_config", "weight", "width", "xavier_uniform_", "zero_", "zeros_"], "siglip/modeling_siglip.py:SiglipEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "auto_docstring", "class", "config", "def", "encoder_layer", "for", "forward", "gradient_checkpointing", "hidden_states", "in", "inputs_embeds", "kwargs", "last_hidden_state", "layers", "nn", "num_hidden_layers", "range", "return", "self", "super", "torch"], "siglip/modeling_siglip.py:SiglipTextTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "Linear", "ModelEncoder", "ModelTextConfig", "ModelTextEmbeddings", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_prepare_4d_attention_mask", "and", "attention_mask", "auto_docstring", "can_return_tuple", "class", "config", "def", "dtype", "elif", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "final_layer_norm", "flash", "forward", "have", "head", "hidden_size", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "pooled_output", "pooler_output", "position_ids", "projection_size", "raise", "return", "self", "size", "specify", "super", "to", "torch", "uses_flash_attention", "view"], "siglip/modeling_siglip.py:SiglipTextModel": ["BaseModelOutputWithPooling", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelTextTransformer", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attention_mask", "auto_docstring", "check_model_inputs", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "input_ids", "kwargs", "nn", "position_ids", "post_init", "r", "return", "self", "set_input_embeddings", "super", "text_model", "token_embedding", "torch", "value"], "siglip/modeling_siglip.py:SiglipVisionTransformer": ["BaseModelOutput", "BaseModelOutputWithPooling", "False", "LayerNorm", "ModelEncoder", "ModelMultiheadAttentionPoolingHead", "ModelVisionConfig", "ModelVisionEmbeddings", "ModelVisionTransformer", "Module", "None", "Optional", "TransformersKwargs", "True", "Unpack", "__init__", "auto_docstring", "bool", "can_return_tuple", "class", "config", "def", "else", "embed_dim", "embeddings", "encoder", "encoder_outputs", "eps", "forward", "hasattr", "head", "hidden_size", "hidden_states", "if", "inputs_embeds", "interpolate_pos_encoding", "kwargs", "last_hidden_state", "layer_norm_eps", "nn", "not", "pixel_values", "pooler_output", "post_layernorm", "return", "self", "super", "use_head", "vision_use_head"], "siglip/modeling_siglip.py:SiglipMultiheadAttentionPoolingHead": ["LayerNorm", "ModelMLP", "ModelMultiheadAttentionPoolingHead", "ModelVisionConfig", "Module", "MultiheadAttention", "Parameter", "True", "__init__", "attention", "batch_first", "batch_size", "class", "config", "def", "eps", "forward", "hidden_size", "hidden_state", "layer_norm_eps", "layernorm", "mlp", "nn", "num_attention_heads", "probe", "randn", "repeat", "residual", "return", "self", "shape", "super", "torch"], "siglip/modeling_siglip.py:SiglipVisionModel": ["BaseModelOutputWithPooling", "False", "ModelPreTrainedModel", "ModelVisionConfig", "ModelVisionModel", "ModelVisionTransformer", "Module", "TransformersKwargs", "Unpack", "__init__", "auto_docstring", "bool", "check_model_inputs", "class", "config", "def", "embeddings", "forward", "get_input_embeddings", "interpolate_pos_encoding", "kwargs", "main_input_name", "nn", "patch_embedding", "pixel_values", "post_init", "r", "return", "self", "super", "vision_model"], "siglip/modeling_siglip.py:SiglipModel": ["BaseModelOutputWithPooling", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelModel", "ModelOutput", "ModelPreTrainedModel", "ModelTextConfig", "ModelTextModel", "ModelVisionConfig", "ModelVisionModel", "None", "Optional", "Parameter", "Tensor", "TransformersKwargs", "True", "TypeError", "Unpack", "__init__", "_from_config", "attention_mask", "auto_docstring", "be", "bool", "but", "can_return_tuple", "class", "config", "def", "device", "dim", "exp", "expected", "eye", "f", "filter_out_non_signature_kwargs", "forward", "functional", "get_image_features", "get_text_features", "if", "image_embeds", "input_ids", "interpolate_pos_encoding", "is", "isinstance", "keepdim", "kwargs", "logit_bias", "logit_scale", "logits_per_image", "logits_per_text", "loglik", "logsigmoid", "loss", "m1_diag1", "matmul", "mean", "nll", "nn", "norm", "not", "of", "ones_like", "p", "pixel_values", "pooled_output", "pooler_output", "position_ids", "post_init", "r", "raise", "randn", "return", "return_loss", "self", "size", "sum", "super", "t", "text_config", "text_embeds", "text_model", "text_model_output", "text_outputs", "to", "torch", "type", "vision_config", "vision_model", "vision_model_output", "vision_outputs"], "siglip/modeling_siglip.py:SiglipForImageClassification": ["BaseModelOutputWithPooling", "False", "Identity", "ImageClassifierOutput", "Linear", "ModelConfig", "ModelForImageClassification", "ModelPreTrainedModel", "ModelVisionModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "_from_config", "auto_docstring", "bool", "check_model_inputs", "class", "classifier", "config", "def", "dim", "else", "forward", "hidden_size", "if", "interpolate_pos_encoding", "is", "kwargs", "labels", "last_hidden_state", "logits", "loss", "loss_function", "main_input_name", "mean", "nn", "not", "num_labels", "outputs", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch", "vision_config", "vision_model"], "qwen2/modeling_qwen2.py:Qwen2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "qwen2/modeling_qwen2.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen2/modeling_qwen2.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "qwen2/modeling_qwen2.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen2/modeling_qwen2.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen2/modeling_qwen2.py:Qwen2Attention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "sliding_attention", "sliding_window", "super", "torch", "training", "transpose", "tuple", "update", "v_proj", "value_states", "version", "view"], "qwen2/modeling_qwen2.py:Qwen2RMSNorm": ["ModelRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "qwen2/modeling_qwen2.py:Qwen2DecoderLayer": ["Cache", "False", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelMLP", "ModelRMSNorm", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "_", "__init__", "attention_mask", "attention_type", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "input_layernorm", "int", "kwargs", "layer_idx", "layer_types", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "return", "rms_norm_eps", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "qwen2/modeling_qwen2.py:Qwen2PreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen2/modeling_qwen2.py:Qwen2RotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen2/modeling_qwen2.py:Qwen2Model": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelModel", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "attention_type", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "check_model_inputs", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "else", "embed_tokens", "eps", "exactly", "for", "forward", "full_attention", "get_seq_length", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "isinstance", "kwargs", "last_hidden_state", "layer_idx", "layer_types", "layers", "mask_kwargs", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "qwen2/modeling_qwen2.py:Qwen2ForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "lm_head", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "torch", "use_cache", "vocab_size", "weight"], "qwen2/modeling_qwen2.py:Qwen2ForSequenceClassification": ["GenericForSequenceClassification", "ModelForSequenceClassification", "ModelPreTrainedModel", "class", "pass"], "qwen2/modeling_qwen2.py:Qwen2ForTokenClassification": ["GenericForTokenClassification", "ModelForTokenClassification", "ModelPreTrainedModel", "class", "pass"], "qwen2/modeling_qwen2.py:Qwen2ForQuestionAnswering": ["GenericForQuestionAnswering", "ModelForQuestionAnswering", "ModelPreTrainedModel", "base_model_prefix", "class", "transformer"], "cohere/modeling_cohere.py:CohereLayerNorm": ["False", "ModelLayerNorm", "Module", "None", "Parameter", "True", "__init__", "bias", "class", "def", "dtype", "eps", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "super", "to", "torch", "variance", "variance_epsilon", "weight"], "cohere/modeling_cohere.py:CohereRotaryEmbedding": ["False", "ModelConfig", "ModelRotaryEmbedding", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dict", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "repeat_interleave", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "cohere/modeling_cohere.py:CohereMLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "intermediate_size", "nn", "return", "self", "super", "up_proj", "x"], "cohere/modeling_cohere.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "cohere/modeling_cohere.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "cohere/modeling_cohere.py:rotate_half": ["Model_half", "def", "dim", "flatten", "return", "rot_x", "stack", "torch", "x", "x1", "x2"], "cohere/modeling_cohere.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dtype", "float", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "to", "unsqueeze", "unsqueeze_dim"], "cohere/modeling_cohere.py:CohereAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "FlashAttentionKwargs", "Linear", "LongTensor", "ModelAttention", "ModelConfig", "ModelLayerNorm", "Module", "None", "Optional", "Tensor", "True", "Unpack", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_bias", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "cache_kwargs", "cache_position", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "dropout", "eager", "eager_attention_forward", "else", "eps", "forward", "getattr", "head_dim", "hidden_shape", "hidden_size", "hidden_states", "if", "input_shape", "int", "is", "is_causal", "k_norm", "k_proj", "key_states", "kwargs", "layer_idx", "layer_norm_eps", "new_name", "nn", "not", "num_attention_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "past_key_value", "past_key_values", "position_embeddings", "q_norm", "q_proj", "query_states", "reshape", "return", "scaling", "self", "shape", "sin", "super", "torch", "training", "transpose", "tuple", "update", "use_qk_norm", "v_proj", "value_states", "version", "view"], "cohere/modeling_cohere.py:CohereDecoderLayer": ["Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm", "ModelMLP", "None", "Optional", "Tensor", "Unpack", "_", "__init__", "attention_mask", "bool", "cache_position", "class", "config", "def", "deprecate_kwarg", "eps", "forward", "hidden_size", "hidden_states", "hidden_states_attention", "hidden_states_mlp", "input_layernorm", "int", "kwargs", "layer_idx", "layer_norm_eps", "mlp", "new_name", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "residual", "return", "self", "self_attn", "super", "torch", "tuple", "use_cache", "version"], "cohere/modeling_cohere.py:CoherePreTrainedModel": ["ModelAttention", "ModelConfig", "ModelDecoderLayer", "ModelPreTrainedModel", "PreTrainedModel", "True", "_can_compile_fullgraph", "_can_record_outputs", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "class", "config", "hidden_states", "model", "past_key_values", "supports_gradient_checkpointing"], "cohere/modeling_cohere.py:CohereModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FloatTensor", "LongTensor", "ModelConfig", "ModelDecoderLayer", "ModelLayerNorm", "ModelModel", "ModelPreTrainedModel", "ModelRotaryEmbedding", "ModuleList", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "ValueError", "You", "__init__", "and", "arange", "attention_mask", "auto_docstring", "bool", "cache_position", "causal_mask", "check_model_inputs", "class", "config", "create_causal_mask", "decoder_layer", "def", "device", "else", "embed_tokens", "eps", "exactly", "for", "forward", "get_seq_length", "gradient_checkpointing", "hidden_size", "hidden_states", "if", "in", "input_embeds", "input_ids", "inputs_embeds", "is", "kwargs", "last_hidden_state", "layer_idx", "layer_norm_eps", "layers", "must", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "rotary_emb", "self", "shape", "specify", "super", "torch", "unsqueeze", "use_cache", "vocab_size"], "cohere/modeling_cohere.py:CohereForCausalLM": ["BaseModelOutputWithPast", "Cache", "CausalLMOutputWithPast", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Union", "Unpack", "__init__", "_pp_plan", "_tied_weights_keys", "_tp_plan", "attention_mask", "attentions", "auto_docstring", "bias", "bool", "cache_position", "can_return_tuple", "class", "colwise_rep", "config", "def", "else", "forward", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "lm_head", "logit_scale", "logits", "logits_to_keep", "loss", "loss_function", "model", "nn", "not", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "position_ids", "post_init", "r", "return", "self", "slice", "slice_indices", "super", "tie_word_embeddings", "torch", "use_cache", "vocab_size", "weight"], "timm_wrapper/modeling_timm_wrapper.py:TimmWrapperModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "r", "torch", "tuple"], "timm_wrapper/modeling_timm_wrapper.py:_create_timm_model_with_error_handling": ["False", "ImportError", "ModelConfig", "Please", "RuntimeError", "The", "U", "Unknown", "__version__", "_create_timm_model_with_error_handling", "a", "architecture", "as", "config", "create_model", "def", "e", "except", "f", "from", "if", "in", "install", "is", "model", "model_kwargs", "more", "not", "of", "pip", "pretrained", "raise", "recent", "return", "str", "supported", "timm", "to", "try", "upgrade", "version", "with", "your"], "timm_wrapper/modeling_timm_wrapper.py:TimmWrapperPreTrainedModel": ["Exception", "False", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "Model_model", "None", "PreTrainedModel", "True", "_Model_model_supports_gradient_checkpointing", "__init__", "_fix_state_dict_key_on_load", "_fix_state_dict_key_on_save", "_init_weights", "_no_split_modules", "_set_gradient_checkpointing", "accepts_loss_kwargs", "args", "bias", "bool", "class", "config", "data", "def", "enable", "except", "f", "for", "hasattr", "if", "in", "initializer_range", "is", "isinstance", "items", "k", "key", "kwargs", "load_state_dict", "main_input_name", "mean", "model_tags", "module", "nn", "normal_", "not", "pixel_values", "post_init", "replace", "requires_backends", "return", "self", "set_grad_checkpointing", "state_dict", "staticmethod", "std", "str", "super", "supports_gradient_checkpointing", "try", "tuple", "v", "vision", "weight", "zero_"], "timm_wrapper/modeling_timm_wrapper.py:TimmWrapperModel": ["Cannot", "FloatTensor", "Model", "ModelConfig", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "Model_model", "None", "Optional", "Please", "Tensor", "The", "To", "Union", "ValueError", "__init__", "_create_Model_model_with_error_handling", "a", "and", "architecture", "auto_docstring", "available", "be", "bool", "cannot", "class", "compatible", "config", "consider", "def", "device", "different", "do_pooling", "dtype", "else", "enable", "extra_init_kwargs", "feature", "for", "forward", "forward_features", "forward_head", "forward_intermediates", "hasattr", "hidden_states", "if", "implemented", "in", "indices", "int", "is", "isinstance", "kwargs", "last_hidden_state", "list", "method", "model", "model_args", "models", "must", "not", "num_classes", "option", "or", "output", "output_attentions", "output_hidden_states", "outputs", "package", "pixel_values", "pooler_output", "post_init", "r", "raise", "return", "return_dict", "self", "set", "super", "the", "this", "to", "torch", "tuple", "updating", "use_return_dict", "using", "version", "versions"], "timm_wrapper/modeling_timm_wrapper.py:TimmWrapperForImageClassification": ["Cannot", "FloatTensor", "ImageClassifierOutput", "LongTensor", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "Model_model", "None", "Optional", "Please", "Tensor", "The", "To", "Union", "ValueError", "You", "__init__", "_create_Model_model_with_error_handling", "a", "and", "architecture", "are", "auto_docstring", "available", "be", "bool", "cannot", "checkpoint", "class", "classes", "classifier", "compatible", "config", "consider", "def", "device", "different", "dtype", "e", "else", "enable", "extra_init_kwargs", "extraction", "feature", "for", "forward", "forward_head", "forward_intermediates", "from", "from_pretrained", "g", "hasattr", "head", "hidden_states", "if", "implemented", "in", "indices", "int", "into", "is", "isinstance", "kwargs", "labels", "last_hidden_state", "list", "load", "logits", "loss", "loss_function", "method", "model", "model_args", "models", "must", "no", "not", "num_classes", "num_labels", "number", "of", "option", "or", "output", "output_attentions", "output_hidden_states", "outputs", "package", "pixel_values", "post_init", "r", "raise", "return", "return_dict", "self", "set", "specify", "super", "the", "this", "to", "torch", "trying", "tuple", "updating", "use", "use_return_dict", "using", "version", "versions", "weights", "with"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniPreTrainedModel": ["False", "Model_5OmniConfig", "Model_5OmniDecoderLayer", "Model_5OmniPreTrainedModel", "Model_5OmniVisionBlock", "PreTrainedModel", "True", "_can_compile_fullgraph", "_no_split_modules", "_skip_keys_device_placement", "_supports_attention_backend", "_supports_flash_attn", "_supports_sdpa", "base_model_prefix", "class", "config", "model", "past_key_values", "supports_gradient_checkpointing"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniPreTrainedModelForConditionalGeneration": ["False", "LongTensor", "Model_5OmniPreTrainedModel", "Model_5OmniPreTrainedModelForConditionalGeneration", "None", "Optional", "Tensor", "True", "_", "_iter", "_llm_pos_ids", "_prepare_4d_causal_attention_mask_with_cache_position", "and", "append", "arange", "argwhere", "attention_mask", "audio_chunk_index", "audio_chunk_indexes", "audio_idx", "audio_len", "audio_llm_pos_ids", "audio_nums", "audio_seqlens", "audio_start_token_id", "audio_token_id", "batch_size", "bool", "bos_len", "cache_position", "cat", "causal_mask", "class", "clone", "config", "cpu", "cumsum", "current_chunk", "def", "device", "diagonal", "dim", "dtype", "ed_audio", "ed_image", "ed_video", "elif", "else", "enumerate", "eos_len", "expand", "fill_value", "flatten", "float", "for", "full", "get_chunked_index", "get_llm_pos_ids_for_vision", "get_rope_index", "grid_hs", "grid_t", "grid_ws", "h_index", "i", "if", "image_grid_thw", "image_idx", "image_len", "image_nums", "image_token_id", "in", "index", "input_ids", "input_tokens", "int", "is", "j", "keepdim", "len", "list", "llm_grid_h", "llm_grid_w", "llm_pos_ids", "llm_pos_ids_list", "llm_positions", "long", "mask_length", "masked_fill", "masked_fill_", "max", "max_position_ids", "min", "min_dtype", "min_ed", "mrope_position_deltas", "multimodal_nums", "not", "ones", "or", "padding_mask", "position_id_per_seconds", "position_ids", "prod", "range", "remain_audios", "remain_images", "remain_videos", "remove_index", "reshape", "return", "second_per_grids", "seconds_per_chunk", "self", "sequence_length", "shape", "spatial_merge_size", "squeeze", "st", "st_idx", "stack", "start_idx", "sub_len", "sum", "t_index", "t_ntoken_per_chunk", "target_length", "tensor", "text_len", "to", "token_indices", "tokens_per_chunk", "tolist", "torch", "total_input_ids", "triu", "tuple", "unsqueeze", "use_audio_in_video", "video_chunk_index", "video_chunk_indexes", "video_grid_thw", "video_idx", "video_len", "video_llm_pos_ids", "video_nums", "video_token_id", "view", "vision_idx", "vision_start_indices", "vision_start_token_id", "vision_tokens", "w_index", "while", "yield"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniThinkerCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelOutput", "Model_5OmniThinkerCausalLMOutputWithPast", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "torch", "tuple"], "qwen2_5_omni/modeling_qwen2_5_omni.py:repeat_kv": ["Model_kv", "None", "Tensor", "batch", "def", "expand", "head_dim", "hidden_states", "if", "int", "n_rep", "num_key_value_heads", "reshape", "return", "shape", "slen", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "causal_mask", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "key_states", "kwargs", "matmul", "module", "nn", "not", "num_key_value_groups", "p", "query", "repeat_kv", "return", "scaling", "shape", "softmax", "to", "torch", "training", "transpose", "value", "value_states"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniAudioAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model_5OmniAudioAttention", "Model_5OmniAudioEncoderConfig", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_attn_implementation", "and", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "be", "bias", "by", "class", "config", "contiguous", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "d_model", "def", "divisible", "dropout", "eager", "eager_attention_forward", "else", "embed_dim", "encoder_attention_heads", "f", "forward", "got", "head_dim", "hidden_states", "if", "is_causal", "is_decoder", "k_proj", "key_states", "kwargs", "max", "max_length_k", "max_length_q", "max_seqlen", "must", "nn", "not", "num_heads", "num_key_value_groups", "out_proj", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_length", "size", "super", "torch", "training", "transpose", "tuple", "unsqueeze", "v_proj", "value_states"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniAudioEncoderLayer": ["ACT2FN", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model_5OmniAudioAttention", "Model_5OmniAudioEncoderConfig", "Model_5OmniAudioEncoderLayer", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_mask", "clamp", "clamp_value", "class", "config", "cu_seqlens", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_ffn_dim", "fc1", "fc2", "final_layer_norm", "finfo", "float16", "forward", "hidden_states", "if", "kwargs", "max", "min", "nn", "outputs", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:SinusoidsPositionEmbedding": ["False", "ModelPositionEmbedding", "Module", "ValueError", "__init__", "arange", "cat", "channels", "class", "cos", "def", "dim", "even", "exp", "float", "forward", "if", "input", "int", "inv_timescales", "length", "log", "log_timescale_increment", "max_timescale", "needs", "newaxis", "nn", "np", "persistent", "positional_embedding", "raise", "register_buffer", "return", "scaled_time", "self", "seqlen", "sin", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniAudioEncoder": ["AvgPool1d", "BaseModelOutput", "Conv1d", "Embedding", "F", "False", "LayerNorm", "Linear", "LongTensor", "Model_5OmniAudioEncoder", "Model_5OmniAudioEncoderConfig", "Model_5OmniAudioEncoderLayer", "Model_5OmniPreTrainedModel", "Module", "ModuleList", "None", "SinusoidsPositionEmbedding", "Tensor", "True", "_", "__init__", "_attn_implementation", "_freeze_parameters", "_get_feat_extract_output_lengths", "_no_split_modules", "_prepare_attention_mask", "_requires_grad", "_supports_sdpa", "aftercnn_lens", "append", "attention_mask", "audio_bos_eos_token", "auto_docstring", "avg_pooler", "batch_mask", "batch_mask_after_cnn", "bool", "cat", "ceil", "chunk_lengths", "chunk_list", "chunk_num", "class", "config", "conv1", "conv2", "cu_seqlens", "cumsum", "d_model", "def", "device", "dim", "dropout", "dtype", "each_audio_states", "else", "embed_dim", "embed_scale", "encoder_layer", "encoder_layers", "enumerate", "feature_lens", "feature_lens_after_cnn", "fill_value", "finfo", "flash_attention_2", "for", "forward", "full", "functional", "gelu", "get_input_embeddings", "gradient_checkpointing", "hidden_states", "hidden_states_list", "i", "if", "in", "input_features", "input_lengths", "inputs_tensor", "int32", "kernel_size", "kwargs", "last_hidden_state", "layer_outputs", "layers", "len", "length", "ln_post", "long", "main_input_name", "math", "max", "max_len", "max_len_after_cnn", "max_source_positions", "min", "n_window", "nn", "num_mel_bins", "output_dim", "output_lengths", "pad", "padded_and_mask_function", "padded_embed", "padded_feature", "padded_mask", "padded_mask_after_cnn", "padded_tensor", "padding", "padding_side", "padding_value", "param", "parameters", "positional_embedding", "post_init", "proj", "r", "range", "requires_grad", "return", "right", "scale_embedding", "self", "seq_length", "set_input_embeddings", "shape", "size", "split", "sqrt", "stride", "sum", "super", "tail_chunk_index", "tensor", "tensor_len", "tensor_list", "to", "token_audio", "token_audio_list", "tolist", "torch", "transpose", "transpose_", "unsqueeze", "value", "where", "zeros"], "qwen2_5_omni/modeling_qwen2_5_omni.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "qwen2_5_omni/modeling_qwen2_5_omni.py:apply_rotary_pos_emb_vision": ["Model_rotary_pos_emb_vision", "Tensor", "cos", "def", "dtype", "float", "freqs", "orig_dtype", "output", "repeat", "return", "rotate_half", "sin", "tensor", "to", "torch", "unsqueeze"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniVisionAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "Model_5OmniVisionAttention", "Model_5OmniVisionEncoderConfig", "Module", "None", "Optional", "Tensor", "True", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb_vision", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_outputs", "bias", "cat", "class", "config", "contiguous", "cu_seq_lens_k", "cu_seq_lens_q", "cu_seqlens", "def", "dim", "dropout", "eager", "eager_attention_forward", "else", "flash_attention_2", "for", "forward", "head_dim", "hidden_size", "hidden_states", "if", "in", "is_causal", "k", "key_states", "kwargs", "lengths", "max", "max_length_k", "max_length_q", "max_seqlen", "nn", "not", "num_heads", "num_key_value_groups", "proj", "q", "query_states", "reshape", "return", "rotary_pos_emb", "scaling", "self", "seq_length", "shape", "split", "splits", "squeeze", "super", "tensor", "tolist", "torch", "training", "transpose", "unsqueeze", "v", "value_states", "zip"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniMLP": ["ACT2FN", "False", "Linear", "Model_5OmniMLP", "Module", "__init__", "act_fn", "bias", "bool", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "nn", "return", "self", "super", "up_proj"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniVisionBlock": ["GradientCheckpointingLayer", "ModelRMSNorm", "Model_5OmniMLP", "Model_5OmniVisionAttention", "Model_5OmniVisionBlock", "Model_5OmniVisionEncoderConfig", "None", "Optional", "Tensor", "True", "__init__", "attn", "bias", "class", "config", "cu_seqlens", "def", "eps", "forward", "hidden_size", "hidden_states", "kwargs", "mlp", "norm1", "norm2", "return", "rotary_pos_emb", "self", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5_VisionPatchEmbed": ["Conv3d", "False", "Model_5_VisionPatchEmbed", "Module", "None", "Tensor", "__init__", "bias", "class", "def", "dtype", "embed_dim", "forward", "hidden_states", "in_channels", "int", "kernel_size", "nn", "patch_size", "proj", "return", "self", "stride", "super", "target_dtype", "temporal_patch_size", "to", "torch", "view", "weight"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5_VisionRotaryEmbedding": ["False", "Model_5_VisionRotaryEmbedding", "Module", "None", "Tensor", "__init__", "arange", "class", "def", "device", "dim", "dtype", "float", "forward", "freqs", "int", "inv_freq", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq", "seqlen", "super", "theta", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniPatchMerger": ["GELU", "Linear", "ModelRMSNorm", "Model_5OmniPatchMerger", "Module", "None", "Sequential", "Tensor", "__init__", "class", "context_dim", "def", "dim", "eps", "forward", "hidden_size", "int", "ln_q", "mlp", "nn", "return", "self", "spatial_merge_size", "super", "torch", "view", "x"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniVisionEncoder": ["F", "False", "Model_5OmniPatchMerger", "Model_5OmniPreTrainedModel", "Model_5OmniVisionBlock", "Model_5OmniVisionEncoder", "Model_5OmniVisionEncoderConfig", "Model_5_VisionPatchEmbed", "Model_5_VisionRotaryEmbedding", "ModuleList", "None", "Tensor", "_", "__init__", "_no_split_modules", "append", "arange", "argsort", "blk", "blocks", "cat", "class", "config", "constant", "context_dim", "cu_seqlens", "cu_seqlens_now", "cu_seqlens_tmp", "cu_window_seqlens", "cumsum", "def", "depth", "device", "dim", "dtype", "else", "embed_dim", "enumerate", "expand", "extend", "flatten", "for", "forward", "fullatt_block_indexes", "get_window_index", "gradient_checkpointing", "grid_h", "grid_t", "grid_thw", "grid_w", "h", "head_dim", "hidden_size", "hidden_states", "hpos_ids", "if", "in", "in_channels", "index", "index_new", "index_padded", "inputs", "int32", "is_tracing", "item", "jit", "kwargs", "layer_num", "list", "llm_grid_h", "llm_grid_w", "max", "max_grid_size", "merger", "nn", "num_heads", "num_windows_h", "num_windows_w", "out_hidden_size", "pad", "pad_h", "pad_w", "patch_embed", "patch_size", "permute", "pos_ids", "range", "repeat", "repeat_interleave", "reshape", "return", "reverse_indices", "rot_pos_emb", "rotary_pos_emb", "rotary_pos_emb_full", "self", "seq_len", "seqlens", "size", "spatial_merge_size", "spatial_merge_unit", "stack", "sum", "super", "t", "temporal_patch_size", "tensor", "tolist", "torch", "unique_consecutive", "unsqueeze", "value", "vit_merger_window_size", "w", "window_index", "window_index_id", "window_size", "wpos_ids"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniRotaryEmbedding": ["False", "Model_5OmniRotaryEmbedding", "Model_5OmniThinkerConfig", "Module", "None", "ROPE_INIT_FUNCTIONS", "Tensor", "__init__", "and", "attention_scaling", "autocast", "cat", "class", "config", "cos", "cpu", "def", "default", "device", "device_type", "dim", "dtype", "dynamic_rope_update", "else", "emb", "enabled", "expand", "float", "forward", "freqs", "get", "hasattr", "if", "inv_freq", "inv_freq_expanded", "is", "isinstance", "max_position_embeddings", "max_seq_len_cached", "mps", "nn", "no_grad", "not", "original_inv_freq", "original_max_seq_len", "persistent", "position_ids", "position_ids_expanded", "register_buffer", "return", "rope_init_fn", "rope_scaling", "rope_type", "self", "shape", "sin", "str", "super", "to", "torch", "transpose", "type", "with", "x"], "qwen2_5_omni/modeling_qwen2_5_omni.py:apply_multimodal_rotary_pos_emb": ["Model_multimodal_rotary_pos_emb", "cat", "cos", "def", "dim", "enumerate", "for", "i", "in", "k", "k_embed", "m", "mrope_section", "q", "q_embed", "return", "rotate_half", "sin", "split", "torch", "unsqueeze", "unsqueeze_dim"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniAttention": ["ALL_ATTENTION_FUNCTIONS", "Cache", "Callable", "False", "FlashAttentionKwargs", "Instantiating", "Linear", "LongTensor", "Model_5OmniAttention", "Model_5OmniConfig", "Model_5OmniRotaryEmbedding", "Module", "None", "Optional", "Please", "Tensor", "True", "Unpack", "_", "__class__", "__init__", "__name__", "_attn_implementation", "a", "and", "apply_multimodal_rotary_pos_emb", "attention_dropout", "attention_interface", "attention_mask", "attn_output", "attn_weights", "bias", "bool", "bsz", "cache_kwargs", "cache_position", "caching", "call", "class", "config", "contiguous", "cos", "creating", "def", "deprecate_kwarg", "dropout", "during", "eager", "eager_attention_forward", "else", "errors", "f", "forward", "getattr", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "k_proj", "key_states", "kwargs", "layer_idx", "layer_types", "logger", "make", "mrope_section", "new_name", "nn", "not", "num_attention_heads", "num_heads", "num_key_value_groups", "num_key_value_heads", "o_proj", "output_attentions", "passing", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "provide", "q_len", "q_proj", "query_states", "recommended", "reshape", "return", "rope_scaling", "rotary_emb", "scaling", "self", "sin", "size", "sliding_attention", "sliding_window", "super", "sure", "the", "this", "to", "torch", "training", "transpose", "tuple", "update", "use_cache", "used", "v_proj", "value_states", "version", "view", "warning_once", "when", "will", "without"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2MLP": ["ACT2FN", "False", "Linear", "ModelMLP", "Module", "__init__", "act_fn", "bias", "bool", "class", "config", "def", "down_proj", "forward", "gate_proj", "hidden_act", "hidden_size", "hidden_state", "intermediate_size", "nn", "return", "self", "super", "up_proj"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniDecoderLayer": ["Attention", "Cache", "False", "FlashAttentionKwargs", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelMLP", "ModelRMSNorm", "Model_5OmniAttention", "Model_5OmniDecoderLayer", "Model_5OmniTextConfig", "None", "Optional", "Sliding", "Tensor", "Unpack", "Window", "__init__", "_attn_implementation", "and", "attention_mask", "attention_type", "be", "bool", "but", "cache_position", "class", "config", "def", "deprecate_kwarg", "enabled", "encountered", "eps", "f", "flash_attention_2", "for", "forward", "hidden_size", "hidden_states", "if", "implemented", "input_layernorm", "int", "is", "kwargs", "layer_idx", "layer_types", "logger", "may", "mlp", "new_name", "not", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embeddings", "position_ids", "post_attention_layernorm", "residual", "results", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "unexpected", "use_cache", "use_sliding_window", "version", "warning_once"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniThinkerTextModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelRMSNorm", "Model_5OmniDecoderLayer", "Model_5OmniPreTrainedModel", "Model_5OmniRotaryEmbedding", "Model_5OmniTextConfig", "Model_5OmniThinkerTextModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_no_split_modules", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "checkpointing", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "elif", "else", "embed_tokens", "eps", "exactly", "expand", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "isinstance", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layer_types", "layers", "logger", "mask_kwargs", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "return_dict", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "text_position_ids", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warning_once", "with"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniThinkerForConditionalGeneration": ["Cache", "False", "FloatTensor", "GenerationMixin", "Image", "Linear", "LongTensor", "Model_5OmniAudioEncoder", "Model_5OmniPreTrainedModelForConditionalGeneration", "Model_5OmniThinkerCausalLMOutputWithPast", "Model_5OmniThinkerConfig", "Model_5OmniThinkerForConditionalGeneration", "Model_5OmniThinkerTextModel", "Model_5OmniVisionEncoder", "None", "Optional", "Tensor", "TransformersKwargs", "True", "Union", "Unpack", "ValueError", "Videos", "_", "__init__", "_from_config", "_get_feat_extract_output_lengths", "_no_split_modules", "_tied_weights_keys", "add", "aftercnn_lens", "all", "and", "arange", "attention_mask", "attentions", "audio_config", "audio_feat_lengths", "audio_feature_lengths", "audio_features", "audio_mask", "audio_output_lengths", "audio_outputs", "audio_token_id", "audio_tower", "auto_docstring", "base_model_prefix", "batch_size", "bias", "bool", "cache_position", "class", "config", "def", "delta", "delta0", "device", "dim", "do", "dtype", "else", "embed_tokens", "expand", "expand_as", "f", "feature_attention_mask", "feature_lens", "features", "forward", "get_audio_features", "get_image_features", "get_input_embeddings", "get_placeholder_mask", "get_rope_index", "get_text_config", "get_video_features", "grid_thw", "hidden_size", "hidden_states", "if", "image", "image_embeds", "image_features", "image_grid_thw", "image_mask", "image_token_id", "input_features", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "last_hidden_state", "length", "lm_head", "logits", "long", "loss", "loss_function", "masked_scatter", "match", "model", "model_inputs", "n_image_tokens", "n_video_tokens", "nn", "not", "numel", "of", "or", "output", "output_attentions", "output_hidden_states", "outputs", "pad_token_id", "past_key_values", "permute", "pixel_values", "pixel_values_videos", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "rope_deltas", "self", "seq_length", "set_input_embeddings", "shape", "should", "spatial_merge_size", "special_audio_mask", "special_image_mask", "special_video_mask", "sum", "super", "tensor", "text_config", "thinker", "to", "tokens", "tolist", "torch", "tuple", "type", "unsqueeze", "use_audio_in_video", "use_cache", "use_return_dict", "value", "video_embeds", "video_features", "video_grid_thw", "video_mask", "video_second_per_grid", "video_token_id", "view", "vision_config", "visual", "vocab_size", "weight"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniTalkerCausalLMOutputWithPast": ["Cache", "FloatTensor", "LongTensor", "ModelOutput", "Model_5OmniTalkerCausalLMOutputWithPast", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "r", "rope_deltas", "thinker_reply_part", "torch", "tuple"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniTalkerModel": ["BaseModelOutputWithPast", "Cache", "DynamicCache", "Embedding", "False", "FlashAttentionKwargs", "FloatTensor", "LongTensor", "ModelRMSNorm", "Model_5OmniDecoderLayer", "Model_5OmniPreTrainedModel", "Model_5OmniRotaryEmbedding", "Model_5OmniTalkerConfig", "Model_5OmniTalkerDecoderLayer", "Model_5OmniTalkerModel", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "Unpack", "ValueError", "You", "__init__", "_attn_implementation", "_no_split_modules", "all_hidden_states", "all_self_attns", "and", "arange", "attention_mask", "attention_type", "attentions", "auto_docstring", "bool", "cache_position", "causal_mask_mapping", "checkpointing", "class", "config", "create_causal_mask", "create_sliding_window_causal_mask", "decoder_layer", "def", "device", "dict", "elif", "else", "embed_tokens", "embedding_size", "eps", "exactly", "expand", "for", "forward", "full_attention", "get_seq_length", "gradient", "gradient_checkpointing", "has_sliding_layers", "hidden_size", "hidden_states", "if", "in", "incompatible", "input_embeds", "input_ids", "inputs_embeds", "is", "is_tracing", "isinstance", "jit", "kwargs", "last_hidden_state", "layer_idx", "layer_outputs", "layer_types", "layers", "logger", "mask_kwargs", "must", "ndim", "nn", "norm", "not", "num_hidden_layers", "of", "one", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_seen_tokens", "position_embeddings", "position_ids", "post_init", "raise", "range", "return", "return_dict", "rms_norm_eps", "rotary_emb", "self", "shape", "sliding_attention", "specify", "super", "text_position_ids", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warning_once", "with"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniTalkerForConditionalGeneration": ["Any", "Cache", "False", "FloatTensor", "GenerationMixin", "Linear", "LongTensor", "ModelOutput", "Model_5OmniPreTrainedModelForConditionalGeneration", "Model_5OmniTalkerCausalLMOutputWithPast", "Model_5OmniTalkerConfig", "Model_5OmniTalkerForConditionalGeneration", "Model_5OmniTalkerModel", "None", "Optional", "Tensor", "True", "Union", "__init__", "_get_initial_cache_position", "_update_model_kwargs_for_generation", "add", "and", "arange", "attention_mask", "attentions", "audio_feature_attention_mask", "audio_feature_lengths", "auto_docstring", "base_model_prefix", "batch_size", "bias", "bool", "cache_position", "class", "codebook_size", "codec_bos_token", "codec_embeds", "codec_eos_token", "codec_head", "codec_mask_token", "codec_pad_token", "config", "def", "delta", "device", "dict", "dtype", "else", "embedding_size", "expand", "float", "forward", "get_input_embeddings", "get_rope_index", "getattr", "hidden_size", "hidden_states", "if", "image_grid_thw", "input_audio_features", "input_ids", "input_text_ids", "inputs_embeds", "int", "is", "is_encoder_decoder", "kwargs", "logits", "long", "loss", "model", "model_inputs", "model_kwargs", "nn", "not", "num_new_tokens", "or", "output", "output_attentions", "output_hidden_states", "outputs", "past_key_values", "pixel_values", "pixel_values_videos", "pop", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "return", "return_dict", "rope_deltas", "self", "seq_length", "set_input_embeddings", "shape", "spatial_merge_size", "str", "super", "talker", "talker_lm_input", "tensor", "text_bos_token", "text_eos_token", "text_pad_token", "thinker_reply_part", "thinker_to_talker_proj", "to", "torch", "tts_codec_end_token_id", "tts_codec_mask_token_id", "tts_codec_pad_token_id", "tts_codec_start_token_id", "tts_text_end_token_id", "tts_text_pad_token_id", "tts_text_start_token_id", "tuple", "unsqueeze", "use_audio_in_video", "use_cache", "use_return_dict", "value", "video_grid_thw", "video_second_per_grid", "view", "vocab_size"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniDiTRotaryEmbedding": ["False", "Model_5OmniDiTRotaryEmbedding", "Module", "Tensor", "__init__", "arange", "autocast", "base", "batch_size", "class", "cos", "cpu", "def", "device", "device_type", "dim", "dtype", "else", "enabled", "float", "forward", "freqs", "if", "inv_freq", "mps", "nn", "register_buffer", "repeat", "reshape", "return", "self", "seq_len", "shape", "sin", "stack", "super", "t", "to", "torch", "type", "unsqueeze", "with", "x"], "qwen2_5_omni/modeling_qwen2_5_omni.py:TimeDelayNetBlock": ["Conv1d", "ModelDelayNetBlock", "Module", "ReLU", "Tensor", "__init__", "activation", "class", "conv", "def", "dilation", "forward", "hidden_states", "in_channels", "kernel_size", "nn", "out_channels", "padding", "padding_mode", "reflect", "return", "same", "self", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Res2NetBlock": ["ModelNetBlock", "Module", "ModuleList", "TimeDelayNetBlock", "__init__", "append", "blocks", "cat", "chunk", "class", "def", "dilation", "dim", "elif", "else", "enumerate", "for", "forward", "hidden_channel", "hidden_part", "hidden_states", "i", "if", "in", "in_channel", "in_channels", "kernel_size", "nn", "out_channels", "output", "output_part", "outputs", "range", "return", "scale", "self", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:SqueezeExcitationBlock": ["Conv1d", "ModelExcitationBlock", "Module", "ReLU", "Sigmoid", "True", "__init__", "class", "conv1", "conv2", "def", "dim", "forward", "hidden_states", "hidden_states_mean", "in_channels", "inplace", "keepdim", "kernel_size", "mean", "nn", "out_channels", "padding", "padding_mode", "reflect", "relu", "return", "same", "se_channels", "self", "sigmoid", "super"], "qwen2_5_omni/modeling_qwen2_5_omni.py:AttentiveStatisticsPooling": ["Conv1d", "F", "ModelStatisticsPooling", "Module", "None", "Tanh", "TimeDelayNetBlock", "True", "__init__", "_compute_statistics", "_length_to_mask", "arange", "as_tensor", "attention", "attention_channels", "cat", "channels", "clamp", "class", "conv", "def", "device", "dim", "dtype", "eps", "expand", "float", "forward", "hidden_states", "if", "in_channels", "inf", "is", "item", "keepdim", "kernel_size", "len", "length", "lengths", "long", "m", "mask", "masked_fill", "max", "max_len", "mean", "nn", "ones", "out_channels", "padding", "padding_mode", "pooled_stats", "pow", "reflect", "repeat", "return", "same", "self", "seq_length", "shape", "softmax", "sqrt", "std", "sum", "super", "tanh", "tdnn", "torch", "total", "unsqueeze", "x"], "qwen2_5_omni/modeling_qwen2_5_omni.py:SqueezeExcitationRes2NetBlock": ["ModelExcitationBlock", "ModelExcitationRes2NetBlock", "Module", "Res2NetBlock", "TimeDelayNetBlock", "__init__", "class", "def", "dilation", "forward", "hidden_state", "in_channels", "kernel_size", "nn", "out_channels", "res2net_block", "res2net_scale", "residual", "return", "se_block", "se_channels", "self", "super", "tdnn1", "tdnn2"], "qwen2_5_omni/modeling_qwen2_5_omni.py:ECAPA_TimeDelayNet": ["AttentiveStatisticsPooling", "Conv1d", "Model_TimeDelayNet", "Module", "ModuleList", "Qwen2_5OmniDiTConfig", "SqueezeExcitationRes2NetBlock", "TimeDelayNetBlock", "ValueError", "__init__", "and", "append", "asp", "attention_channels", "blocks", "cat", "channels", "class", "config", "def", "dilation", "dim", "enc_attention_channels", "enc_channels", "enc_dilations", "enc_dim", "enc_kernel_sizes", "enc_res2net_scale", "enc_se_channels", "fc", "for", "forward", "have", "hidden_states", "hidden_states_list", "i", "if", "in", "in_channels", "kernel_size", "layer", "len", "length", "mel_dim", "mfa", "nn", "or", "out_channels", "padding", "padding_mode", "raise", "range", "reflect", "res2net_scale", "return", "same", "se_channels", "self", "should", "squeeze", "super", "torch", "transpose"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DiTInputEmbedding": ["DiTInputEmbedding", "ECAPA_TimeDelayNet", "False", "Linear", "Module", "None", "Optional", "Qwen2_5OmniDiTConfig", "Tensor", "True", "__init__", "apply_cfg", "bool", "cat", "class", "code_embed", "code_embed_uncond", "condition_vector", "config", "def", "dim", "drop_audio_cond", "elif", "emb_dim", "enc_dim", "enc_emb_dim", "forward", "hidden_size", "hidden_states", "if", "mel_dim", "nn", "proj", "repeat", "return", "self", "size", "speaker_embedding", "spk_encoder", "super", "torch", "unsqueeze", "zeros_like"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DiTCodecEmbedding": ["DiTCodecEmbedding", "Embedding", "False", "Module", "__init__", "class", "code", "code_embed", "codec_dim", "codec_embed", "codec_num_embeds", "def", "dim", "drop_code", "forward", "if", "nn", "repeat_interleave", "repeats", "return", "self", "super", "torch", "zeros_like"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5_OmniAdaLayerNormZero": ["False", "LayerNorm", "Linear", "ModelAdaLayerNormZero", "Module", "None", "SiLU", "__init__", "chunk", "class", "def", "dim", "elementwise_affine", "emb", "eps", "forward", "gate_mlp", "gate_msa", "hidden_states", "linear", "nn", "norm", "return", "scale_mlp", "scale_msa", "self", "shift_mlp", "shift_msa", "silu", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5_OmniAdaLayerNormZero_Final": ["False", "LayerNorm", "Linear", "ModelAdaLayerNormZero_Final", "Module", "None", "SiLU", "__init__", "chunk", "class", "def", "dim", "elementwise_affine", "emb", "eps", "forward", "hidden_states", "linear", "nn", "norm", "return", "scale", "self", "shift", "silu", "super", "torch"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DiTMLP": ["DiTMLP", "Dropout", "GELU", "Linear", "Module", "ModuleList", "__init__", "approximate", "class", "def", "dim", "dropout", "ff", "for", "forward", "hidden_states", "in", "inner_dim", "int", "layer", "mult", "nn", "return", "self", "super", "tanh"], "qwen2_5_omni/modeling_qwen2_5_omni.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "None", "cos", "def", "dim", "k", "k_embed", "position_ids", "q", "q_embed", "reshape", "return", "rotate_half_codec", "shape", "sin", "stack", "torch", "unbind", "unsqueeze", "unsqueeze_dim", "x", "x1", "x2"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DiTAttention": ["ALL_ATTENTION_FUNCTIONS", "DiTAttention", "Dropout", "False", "Linear", "Module", "ModuleList", "None", "Qwen2_5OmniDiTConfig", "Tensor", "_", "__init__", "_attn_implementation", "apply_rotary_pos_emb", "attention_interface", "attention_mask", "attention_output", "attention_weights", "batch_size", "class", "config", "cos", "def", "dim", "dropout", "dtype", "forward", "head_dim", "heads", "hidden_size", "hidden_states", "inner_dim", "is_causal", "key", "nn", "num_attention_heads", "position_embeddings", "query", "reshape", "return", "self", "shape", "sin", "super", "to", "to_k", "to_out", "to_q", "to_v", "torch", "transpose", "value", "view"], "qwen2_5_omni/modeling_qwen2_5_omni.py:SinusPositionEmbedding": ["ModelPositionEmbedding", "Module", "__init__", "arange", "cat", "class", "cos", "def", "device", "dim", "emb", "exp", "float", "forward", "half_dim", "hidden_states", "log", "math", "nn", "return", "scale", "self", "sin", "super", "torch", "type_as", "unsqueeze"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DiTTimestepEmbedding": ["DiTTimestepEmbedding", "Linear", "Module", "ModuleList", "SiLU", "SinusPositionEmbedding", "__init__", "class", "def", "dim", "dtype", "for", "forward", "freq_embed_dim", "in", "layer", "nn", "return", "self", "super", "time_embed", "time_hidden", "time_mlp", "timestep", "to"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DiTDecoderLayer": ["DiTAttention", "DiTDecoderLayer", "DiTMLP", "False", "LayerNorm", "ModelAdaLayerNormZero", "Module", "None", "Qwen2_5OmniDiTConfig", "__init__", "attention_mask", "attn", "attn_norm", "attn_output", "block_diff", "class", "config", "def", "dim", "dropout", "elementwise_affine", "emb", "eps", "ff", "ff_mult", "ff_norm", "ff_output", "float", "forward", "gate_mlp", "gate_msa", "hidden_size", "hidden_states", "look_ahead_block", "look_backward_block", "mult", "nn", "norm", "position_embeddings", "return", "scale_mlp", "self", "shift_mlp", "super", "timestep", "unsqueeze"], "qwen2_5_omni/modeling_qwen2_5_omni.py:SnakeBeta": ["ModelBeta", "Module", "Parameter", "__init__", "alpha", "beta", "class", "def", "exp", "forward", "hidden_states", "in_features", "nn", "no_div_by_zero", "pow", "return", "self", "sin", "super", "torch", "unsqueeze", "zeros"], "qwen2_5_omni/modeling_qwen2_5_omni.py:kaiser_sinc_filter1d": ["False", "Model_sinc_filter1d", "Model_window", "arange", "attenuation", "beta", "cutoff", "def", "delta_f", "dtype", "elif", "else", "float32", "half_size", "half_width", "if", "is_even", "kernel_size", "math", "normalized_filter", "periodic", "pi", "return", "sinc", "sinc_filter", "sum", "time_indices", "torch", "view", "zeros"], "qwen2_5_omni/modeling_qwen2_5_omni.py:UpSample1d": ["F", "False", "Module", "None", "UpSample1d", "__init__", "channels", "class", "conv_transpose1d", "cutoff", "def", "else", "expand", "filter", "forward", "groups", "half_width", "hidden_states", "if", "int", "is", "kaiser_sinc_filter1d", "kernel_size", "mode", "nn", "pad", "pad_left", "pad_right", "persistent", "ratio", "register_buffer", "replicate", "return", "self", "shape", "stride", "super"], "qwen2_5_omni/modeling_qwen2_5_omni.py:DownSample1d": ["A", "F", "False", "Minimum", "ModelSample1d", "Module", "None", "ValueError", "__init__", "above", "be", "channels", "class", "conv1d", "cutoff", "def", "does", "even", "expand", "filter", "forward", "groups", "half_width", "hidden_states", "if", "int", "kaiser_sinc_filter1d", "kernel_size", "larger", "make", "mode", "must", "nn", "not", "out", "pad", "pad_left", "pad_right", "persistent", "raise", "ratio", "register_buffer", "replicate", "return", "self", "sense", "shape", "stride", "super", "than", "zero"], "qwen2_5_omni/modeling_qwen2_5_omni.py:TorchActivation1d": ["Activation", "DownSample1d", "ModelActivation1d", "Module", "TypeError", "UpSample1d", "__init__", "act", "activation", "be", "callable", "class", "def", "down_kernel_size", "down_ratio", "downsample", "forward", "function", "hidden_states", "if", "int", "must", "nn", "not", "raise", "return", "self", "super", "up_kernel_size", "up_ratio", "upsample"], "qwen2_5_omni/modeling_qwen2_5_omni.py:AMPBlock": ["Conv1d", "Model", "Module", "ModuleList", "SnakeBeta", "TorchActivation1d", "_", "__init__", "_get_padding", "act1", "act2", "activation", "activations", "acts1", "acts2", "channels", "class", "conv1", "conv2", "convs1", "convs2", "def", "dilation", "for", "forward", "hidden_states", "in", "int", "kernel_size", "len", "nn", "num_layers", "padding", "range", "residual", "return", "self", "super", "torch", "zip"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniToken2WavBigVGANModel": ["AMPBlock", "Conv1d", "ConvTranspose1d", "False", "Model_5OmniBigVGANConfig", "Model_5OmniPreTrainedModel", "Model_5OmniToken2WavBigVGANModel", "ModuleList", "SnakeBeta", "TorchActivation1d", "__init__", "activation", "activation_post", "amplitude", "amplitude_spectrum", "amplitude_to_db", "bias", "block_index", "clamp", "class", "config", "conv_post", "conv_pre", "cpu", "decibel_spectrum", "def", "device", "dilation", "dtype", "enumerate", "exp", "for", "forward", "hidden_representation", "in", "kernel_size", "layer_idx", "layer_index", "len", "log", "log10", "max", "max_value", "mel_dim", "mel_spectrogram", "min", "min_db", "min_db_level", "min_level", "nn", "normalize_spectrogram", "np", "num_residual_blocks", "num_upsample_layers", "output_waveform", "padding", "process_mel_spectrogram", "processed_spectrogram", "range", "resblock_dilation_sizes", "resblock_kernel_sizes", "resblocks", "residual_output", "return", "self", "spectrogram", "squeeze", "stride", "sum", "super", "tensor", "torch", "ups", "upsample_initial_channel", "upsample_kernel_sizes", "upsample_rates", "zip"], "qwen2_5_omni/modeling_qwen2_5_omni.py:RungeKutta4ODESolver": ["ModelKutta4ODESolver", "None", "_", "__init__", "_compute_step", "_linear_interpolation", "_one_third", "_rk4_step", "_two_thirds", "and", "class", "current_index", "current_value", "def", "delta_value", "device", "dtype", "else", "empty", "for", "function", "function_value_start", "if", "in", "initial_value", "integrate", "is", "k1", "k2", "k3", "k4", "len", "next_value", "not", "return", "self", "shape", "solution", "time_end", "time_point", "time_points", "time_start", "time_step", "torch", "value_end", "value_start", "weight", "while", "zip"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniToken2WavDiTModel": ["DiTCodecEmbedding", "DiTDecoderLayer", "DiTInputEmbedding", "DiTTimestepEmbedding", "False", "Linear", "ModelAdaLayerNormZero_Final", "Model_5OmniDiTConfig", "Model_5OmniDiTRotaryEmbedding", "Model_5OmniPreTrainedModel", "Model_5OmniToken2WavDiTModel", "ModuleList", "None", "Only", "RungeKutta4ODESolver", "True", "ValueError", "__init__", "_create_block_diff", "_no_split_modules", "append", "apply_cfg", "arange", "batch", "batch_size", "block_diff", "block_i", "block_indices", "block_j", "block_size", "blockwise_difference", "chunk", "class", "code_embed_uncond", "condition_vector", "conditioning_vector", "config", "cos", "currently", "def", "device", "dim", "drop_audio_cond", "drop_audio_conditioning", "drop_code", "dtype", "else", "emb_dim", "expand", "for", "forward", "function", "generated_mel_spectrogram", "generated_waveform", "guidance_scale", "guided_prediction", "head_dim", "hidden_size", "hidden_states", "i", "if", "in", "initial_state", "initial_time", "initial_value", "input_embed", "integrate", "is", "layers", "linspace", "look_ahead_block", "look_ahead_layers", "look_backward_block", "look_backward_layers", "maximum_duration", "mel_dim", "model_output", "ndim", "nn", "no_grad", "noise_initialization", "norm_out", "not", "null_prediction", "num_attention_heads", "num_embeds", "num_hidden_layers", "num_steps", "ode_function", "ode_solver", "output", "permute", "pi", "position_embeddings", "prediction", "proj_out", "quantized_code", "raise", "randn", "range", "reference_mel_spectrogram", "repeat", "repeats", "return", "rotary_embed", "sample", "self", "seq_len", "shape", "size", "solution_trajectory", "speaker_embedding", "super", "supported", "sway_coefficient", "text_embed", "text_embedding", "text_embedding_unconditioned", "time_embed", "time_embedding", "time_step", "to", "torch", "transformer_block", "transformer_blocks", "unsqueeze"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniToken2WavModel": ["Model_5OmniPreTrainedModel", "Model_5OmniToken2WavBigVGANModel", "Model_5OmniToken2WavConfig", "Model_5OmniToken2WavDiTModel", "Model_5OmniToken2WavModel", "__init__", "_attn_implementation", "_from_config", "_no_split_modules", "and", "attention", "attn_impl", "attn_implementation", "back", "base_model_prefix", "bf16", "bigvgan_config", "but", "class", "code", "code2wav_bigvgan_model", "code2wav_dit_model", "conditioning", "config", "def", "dit_config", "does", "eager", "elif", "fall", "fallback", "flash_attention_2", "forward", "fp16", "fp32", "guidance_scale", "if", "implementation", "inference", "kwargs", "logger", "mel_spectrogram", "model", "must", "not", "num_steps", "of", "only", "reference_mel", "return", "sample", "sdpa", "self", "super", "support", "supports", "sway_coefficient", "to", "warning_once", "waveform", "will", "with"], "qwen2_5_omni/modeling_qwen2_5_omni.py:Qwen2_5OmniForConditionalGeneration": ["Cannot", "Chelsie", "False", "GenerationMixin", "Model", "Model_5OmniConfig", "Model_5OmniForConditionalGeneration", "Model_5OmniPreTrainedModel", "Model_5OmniTalkerForConditionalGeneration", "Model_5OmniThinkerForConditionalGeneration", "Model_5OmniToken2WavModel", "None", "NotImplementedError", "Omni", "Optional", "Speaker", "Tensor", "True", "Use", "ValueError", "__init__", "_no_split_modules", "and", "attention_mask", "audio", "audio_feature_lengths", "audio_ids_mask", "audio_mask", "audio_mask_tensor", "audio_token_index", "available", "batched", "bool", "bos_token", "cache_dir", "cached_file", "cat", "check_torch_load_is_safe", "class", "classmethod", "clone", "cls", "codec_bos_token", "codec_mask_token", "codec_pad_token", "cond", "conditioning", "config", "currently", "def", "del", "device", "dim", "disable_talker", "do_sample", "does", "dtype", "elif", "else", "embeds_to_talker", "enable", "enable_audio_output", "enable_talker", "eos_embedding", "eos_token", "eos_token_id", "expand_as", "f", "feature_attention_mask", "fill_value", "float", "for", "force_download", "from_pretrained", "full_like", "generate", "generate_audio", "get", "get_input_embeddings", "has_talker", "hasattr", "hidden_states", "if", "ignore_mismatched_sizes", "image_ids_mask", "image_mask", "image_mask_tensor", "image_token_index", "in", "inference", "info", "initialized", "input_features", "input_ids", "input_text_ids", "inputs_embeds", "int", "is", "is_tensor", "items", "k", "key", "keys", "kwargs", "len", "list", "load", "load_speakers", "loaded", "local_files_only", "logger", "long", "main", "masked_scatter_", "max_new_tokens", "method", "model", "model_args", "module", "new_ones", "no_grad", "not", "or", "output", "output_hidden_states", "pad_embedding", "pad_token", "path", "pixel_values", "pixel_values_videos", "pop", "post_init", "pretrained_model_name_or_path", "processed_thinker_hidden", "proxies", "pt", "r", "raise", "ref_mel", "reference_mel", "repetition_penalty", "resume_download", "return", "return_audio", "return_dict_in_generate", "revision", "self", "sequences", "set", "shape", "shared_kwargs", "size", "speaker", "speaker_map", "speaker_params", "speakers", "spk_dict", "spk_path", "startswith", "str", "subfolder", "sum", "super", "support", "suppress_tokens", "talker", "talker_", "talker_attention_mask", "talker_config", "talker_do_sample", "talker_eos_token_id", "talker_generate_codes", "talker_input_ids", "talker_input_text_ids", "talker_inputs_embeds", "talker_kwargs", "talker_max_new_tokens", "talker_repetition_penalty", "talker_result", "talker_temperature", "talker_text_bos_embed", "talker_text_bos_token", "talker_top_k", "talker_top_p", "temperature", "tensor", "text_eos_token", "text_pad_token", "thinker", "thinker_", "thinker_config", "thinker_embed_tokens", "thinker_generate_ids", "thinker_hidden_states", "thinker_kwargs", "thinker_max_new_tokens", "thinker_reply_part", "thinker_result", "thinker_token_embeds", "to", "token", "token2wav", "token2wav_", "token2wav_config", "token2wav_kwargs", "token_hidden_states", "top_k", "top_p", "torch", "unsqueeze", "use", "use_audio_in_video", "use_auth_token", "use_safetensors", "v", "value", "video_ids_mask", "video_mask", "video_mask_tensor", "video_token_index", "wav", "weights_only", "when", "with", "zeros"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersPatchEmbeddings": ["Conv2d", "Expected", "Iterable", "Make", "ModelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "but", "channel", "class", "collections", "config", "configuration", "def", "dimension", "else", "embeddings", "f", "flatten", "forward", "got", "hidden_size", "if", "image_size", "in", "isinstance", "kernel_size", "match", "nn", "num_channels", "num_patches", "of", "one", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "transpose", "values", "with"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersEmbeddings": ["Dropout", "False", "ModelConfig", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "None", "Optional", "Parameter", "Tensor", "True", "ValueError", "Width", "_", "__init__", "align_corners", "and", "antialias", "batch_size", "bicubic", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "does", "dropout", "dtype", "embeddings", "expand", "float32", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "interpolated", "is", "is_tracing", "jit", "mask_token", "match", "mode", "nn", "not", "num_patches", "num_positions", "num_register_tokens", "or", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position", "position_embeddings", "projection", "raise", "randn", "register_tokens", "reshape", "return", "self", "shape", "size", "sqrt_num_positions", "super", "target_dtype", "the", "to", "torch", "torch_int", "unsqueeze", "view", "weight", "where", "width", "with", "zeros"], "dinov2_with_registers/modeling_dinov2_with_registers.py:eager_attention_forward": ["Model_attention_forward", "Module", "None", "Optional", "Tensor", "attention_mask", "attn_output", "attn_weights", "contiguous", "def", "dim", "dropout", "dtype", "float", "float32", "functional", "if", "is", "key", "kwargs", "matmul", "module", "nn", "not", "p", "query", "return", "scaling", "softmax", "to", "torch", "training", "transpose", "value"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersSelfAttention": ["ALL_ATTENTION_FUNCTIONS", "Callable", "False", "Linear", "ModelConfig", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "_attn_implementation", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_interface", "attention_probs", "attention_probs_dropout_prob", "batch_size", "bias", "class", "config", "context_layer", "def", "dropout", "dropout_prob", "eager", "eager_attention_forward", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_causal", "key", "key_layer", "multiple", "new_context_layer_shape", "new_shape", "nn", "not", "num_attention_heads", "number", "of", "qkv_bias", "query", "query_layer", "raise", "reshape", "return", "scaling", "self", "shape", "size", "super", "the", "torch", "training", "transpose", "tuple", "value", "value_layer", "view"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersAttention": ["ModelAttention", "ModelConfig", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "_", "__init__", "all_head_size", "attention", "attention_head_size", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_attn_output", "set", "super", "torch", "union", "value"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersLayerScale": ["ModelLayerScale", "Module", "None", "Parameter", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_size", "hidden_state", "lambda1", "layerscale_value", "nn", "ones", "return", "self", "super", "torch"], "dinov2_with_registers/modeling_dinov2_with_registers.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersMLP": ["ACT2FN", "Linear", "ModelMLP", "Module", "None", "Tensor", "True", "__init__", "activation", "bias", "class", "config", "def", "else", "fc1", "fc2", "forward", "hidden_act", "hidden_features", "hidden_size", "hidden_state", "if", "in_features", "int", "isinstance", "mlp_ratio", "nn", "out_features", "return", "self", "str", "super", "torch"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersSwiGLUFFN": ["Linear", "ModelSwiGLUFFN", "Module", "None", "Tensor", "True", "__init__", "bias", "chunk", "class", "config", "def", "dim", "forward", "functional", "hidden", "hidden_features", "hidden_size", "hidden_state", "in_features", "int", "mlp_ratio", "nn", "out_features", "return", "self", "silu", "super", "torch", "weights_in", "weights_out", "x1", "x2"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersLayer": ["GradientCheckpointingLayer", "Identity", "LayerNorm", "ModelAttention", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelLayerScale", "ModelMLP", "ModelSwiGLUFFN", "None", "Optional", "Tensor", "__init__", "attention", "class", "config", "def", "drop_path", "drop_path_rate", "else", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "hidden_states_norm", "if", "layer_norm_eps", "layer_output", "layer_scale1", "layer_scale2", "mlp", "nn", "norm1", "norm2", "return", "self", "self_attention_output", "super", "torch", "use_swiglu_ffn"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "all_hidden_states", "append", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "nn", "not", "num_hidden_layers", "output_hidden_states", "range", "return", "self", "super", "torch", "tuple"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelEmbeddings", "ModelLayer", "ModelLayerScale", "ModelPreTrainedModel", "ModelSelfAttention", "None", "PreTrainedModel", "True", "Union", "_can_record_outputs", "_init_weights", "_no_split_modules", "_supports_attention_backend", "_supports_flash_attn", "_supports_flex_attn", "_supports_sdpa", "attentions", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "dtype", "elif", "fill_", "float32", "if", "init", "initializer_range", "is", "isinstance", "lambda1", "layerscale_value", "main_input_name", "mask_token", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "register_tokens", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "weight", "zero_"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "ModelConfig", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Tensor", "ValueError", "You", "__init__", "_prune_heads", "attention", "auto_docstring", "bool", "bool_masked_pos", "check_model_inputs", "class", "config", "def", "dict", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "int", "is", "items", "kwargs", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "nn", "num_hidden_layers", "output_hidden_states", "patch_embeddings", "pixel_values", "pooled_output", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "self", "sequence_output", "specify", "super", "to", "torch"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersForImageClassification": ["BaseModelOutputWithPooling", "Identity", "ImageClassifierOutput", "Linear", "Model", "ModelConfig", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "Tensor", "TransformersKwargs", "Unpack", "__init__", "attentions", "auto_docstring", "can_return_tuple", "cat", "class", "classifier", "cls_token", "config", "def", "dim", "else", "forward", "head_mask", "hidden_size", "hidden_states", "if", "is", "kwargs", "labels", "last_hidden_state", "linear_input", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "num_register_tokens", "outputs", "patch_tokens", "pixel_values", "post_init", "r", "return", "self", "sequence_output", "super", "torch"], "dinov2_with_registers/modeling_dinov2_with_registers.py:Dinov2WithRegistersBackbone": ["BackboneMixin", "BackboneOutput", "BaseModelOutput", "LayerNorm", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPatchEmbeddings", "ModelPreTrainedModel", "None", "Optional", "Tensor", "True", "_", "__init__", "_init_backbone", "append", "apply_layernorm", "auto_docstring", "batch_size", "bool", "check_model_inputs", "class", "config", "contiguous", "def", "else", "embedding_output", "embeddings", "encoder", "eps", "feature_maps", "for", "forward", "get_input_embeddings", "height", "hidden_size", "hidden_state", "hidden_states", "if", "in", "is", "kwargs", "layer_norm_eps", "layernorm", "nn", "num_features", "num_hidden_layers", "num_register_tokens", "out_features", "output", "output_hidden_states", "patch_embeddings", "patch_size", "permute", "pixel_values", "post_init", "r", "range", "reshape", "reshape_hidden_states", "return", "self", "shape", "stage", "stage_names", "super", "torch", "tuple", "width", "zip"], "deprecated/realm/modeling_realm.py:RealmEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "deprecated/realm/modeling_realm.py:RealmSelfAttention": ["Cache", "Dropout", "Embedding", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bhld", "bhlr", "bhrd", "bool", "cat", "class", "config", "context_layer", "contiguous", "def", "deprecate_kwarg", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "functional", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "key_length", "long", "lrd", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_name", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "past_key_value", "past_key_values", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "query", "query_layer", "query_length", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "shape", "size", "softmax", "sqrt", "super", "tensor", "the", "to", "torch", "transpose", "transpose_for_scores", "tuple", "use_cache", "value", "value_layer", "version", "view", "x"], "deprecated/realm/modeling_realm.py:RealmSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "deprecated/realm/modeling_realm.py:RealmAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelSelfOutput", "Model_SELF_ATTENTION_CLASSES", "Module", "None", "Optional", "Tensor", "__init__", "_attn_implementation", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "version"], "deprecated/realm/modeling_realm.py:RealmIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deprecated/realm/modeling_realm.py:RealmOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "deprecated/realm/modeling_realm.py:RealmLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "a", "absolute", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "bool", "by", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "cross_attn_past_key_value", "cross_attn_present_key_value", "crossattention", "decoder", "def", "deprecate_kwarg", "else", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_output", "layers", "model", "new_name", "not", "output", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "position_embedding_type", "present_key_value", "raise", "return", "self", "self_attention_outputs", "self_attn_past_key_value", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "version", "with"], "deprecated/realm/modeling_realm.py:RealmEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "_", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "checkpointing", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "logger", "next_decoder_cache", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "past_key_values", "range", "return", "return_dict", "self", "super", "torch", "training", "tuple", "use_cache", "v", "warning_once", "with"], "deprecated/realm/modeling_realm.py:RealmPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "deprecated/realm/modeling_realm.py:RealmEmbedderOutput": ["FloatTensor", "ModelEmbedderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "projected_score", "torch", "tuple"], "deprecated/realm/modeling_realm.py:RealmScorerOutput": ["FloatTensor", "ModelOutput", "ModelScorerOutput", "None", "Optional", "candidate_score", "class", "query_score", "relevance_score", "torch"], "deprecated/realm/modeling_realm.py:RealmReaderOutput": ["BoolTensor", "FloatTensor", "IntTensor", "LongTensor", "ModelOutput", "ModelReaderOutput", "None", "Optional", "attentions", "block_idx", "candidate", "class", "end_pos", "hidden_states", "loss", "reader_correct", "reader_loss", "retriever_correct", "retriever_loss", "start_pos", "torch", "tuple"], "deprecated/realm/modeling_realm.py:RealmForOpenQAOutput": ["LongTensor", "ModelForOpenQAOutput", "ModelOutput", "None", "Optional", "class", "dict", "predicted_answer_ids", "reader_output", "torch"], "deprecated/realm/modeling_realm.py:RealmPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "transform_act_fn"], "deprecated/realm/modeling_realm.py:RealmLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "deprecated/realm/modeling_realm.py:RealmOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super"], "deprecated/realm/modeling_realm.py:RealmScorerProjection": ["LayerNorm", "Linear", "ModelLMPredictionHead", "ModelScorerProjection", "Module", "__init__", "class", "config", "def", "dense", "eps", "forward", "hidden_size", "hidden_states", "layer_norm_eps", "nn", "predictions", "retriever_proj_size", "return", "self", "super"], "deprecated/realm/modeling_realm.py:RealmReaderProjection": ["LayerNorm", "Linear", "ModelReaderProjection", "Module", "ReLU", "_", "__init__", "_spans_given_width", "arange", "block_mask", "candidate_end_projections", "candidate_ends", "candidate_hidden", "candidate_mask", "candidate_start_projections", "candidate_starts", "cat", "chunk", "class", "config", "current_ends", "current_starts", "def", "dense_intermediate", "dense_output", "device", "dim", "dtype", "end_masks", "end_projection", "ends", "eps", "finfo", "float32", "for", "forward", "hidden_size", "hidden_states", "in", "index", "index_select", "layer_normalization", "mask", "mask_to_score", "masks", "max_sequence_len", "max_span_width", "min", "nn", "range", "reader_layer_norm_eps", "reader_logits", "relu", "return", "self", "shape", "span_candidates", "span_hidden_size", "span_masks", "squeeze", "start_masks", "start_projection", "starts", "super", "torch", "type", "w", "width", "zip"], "deprecated/realm/modeling_realm.py:RealmPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_flatten_inputs", "_init_weights", "append", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "fill_", "flattened_inputs", "for", "if", "in", "initializer_range", "input_shape", "inputs", "is", "isinstance", "len", "mean", "module", "nn", "normal_", "not", "padding_idx", "return", "self", "shape", "std", "tensor", "view", "weight", "zero_"], "deprecated/realm/modeling_realm.py:RealmBertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "False", "ModelBertModel", "ModelEmbeddings", "ModelEncoder", "ModelPooler", "ModelPreTrainedModel", "None", "Tensor", "True", "ValueError", "You", "_", "__init__", "_prune_heads", "add_pooling_layer", "and", "at", "attention", "attention_mask", "attentions", "batch_size", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "class", "config", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "expand", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "past_key_values", "past_key_values_length", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "deprecated/realm/modeling_realm.py:RealmEmbedder": ["FloatTensor", "LongTensor", "Model", "ModelBertModel", "ModelEmbedder", "ModelEmbedderOutput", "ModelPreTrainedModel", "ModelScorerProjection", "Model_INPUTS_DOCSTRING", "Model_outputs", "None", "Optional", "Union", "_CONFIG_FOR_DOC", "__init__", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bias", "bool", "class", "cls", "config", "config_class", "decoder", "def", "else", "embeddings", "format", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "not", "output_attentions", "output_hidden_states", "output_type", "pooler_output", "position_ids", "post_init", "predictions", "projected_score", "r", "replace_return_docstrings", "return", "return_dict", "self", "sequence_length", "set_input_embeddings", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "value", "word_embeddings"], "deprecated/realm/modeling_realm.py:RealmScorer": ["FloatTensor", "LongTensor", "ModelEmbedder", "ModelPreTrainedModel", "ModelScorer", "ModelScorerOutput", "Model_INPUTS_DOCSTRING", "None", "Optional", "Union", "ValueError", "You", "_CONFIG_FOR_DOC", "__init__", "_flatten_inputs", "add_start_docstrings_to_model_forward", "and", "attention_mask", "batch_size", "bd", "bn", "bnd", "bool", "candidate_attention_mask", "candidate_input_ids", "candidate_inputs_embeds", "candidate_outputs", "candidate_score", "candidate_token_type_ids", "class", "config", "config_class", "def", "einsum", "either", "else", "embedder", "flattened_attention_mask", "flattened_input_ids", "flattened_token_type_ids", "format", "forward", "have", "head_mask", "if", "input_embeds", "input_ids", "inputs_embeds", "is", "not", "num_candidates", "or", "output_attentions", "output_hidden_states", "output_type", "position_ids", "post_init", "query_embedder", "query_outputs", "query_score", "r", "raise", "relevance_score", "replace_return_docstrings", "retriever_proj_size", "return", "return_dict", "self", "sequence_length", "specify", "super", "to", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/realm/modeling_realm.py:RealmKnowledgeAugEncoder": ["CrossEntropyLoss", "FloatTensor", "LongTensor", "MaskedLMOutput", "Model", "ModelBertModel", "ModelKnowledgeAugEncoder", "ModelOnlyMLMHead", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Union", "ValueError", "You", "_CONFIG_FOR_DOC", "__init__", "_flatten_inputs", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bias", "bool", "candidate_log_prob", "candidate_score", "class", "cls", "compute", "config", "config_class", "decoder", "def", "dtype", "else", "embeddings", "flattened_attention_mask", "flattened_input_ids", "flattened_token_type_ids", "float32", "format", "forward", "get_input_embeddings", "get_output_embeddings", "have", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "is", "joint_gold_log_prob", "joint_output", "joint_outputs", "labels", "log_softmax", "logits", "logsumexp", "loss", "loss_fct", "marginal_gold_log_probs", "masked_lm_log_prob", "masked_lm_loss", "mlm_logits", "mlm_mask", "mlm_targets", "nansum", "new_embeddings", "none", "not", "num_candidates", "ones_like", "order", "output", "output_attentions", "output_hidden_states", "output_type", "position_ids", "post_init", "prediction_scores", "predictions", "r", "raise", "reduction", "relevance_score", "replace_return_docstrings", "return", "return_dict", "self", "seq_length", "sequence_length", "set_input_embeddings", "set_output_embeddings", "size", "specified", "specify", "sum", "super", "tile", "to", "token_type_ids", "torch", "tuple", "type", "unsqueeze", "use_return_dict", "value", "view", "vocab_size", "when", "word_embeddings"], "deprecated/realm/modeling_realm.py:RealmReader": ["BoolTensor", "FloatTensor", "LongTensor", "Model", "ModelBertModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "ModelReader", "ModelReaderOutput", "ModelReaderProjection", "Model_INPUTS_DOCSTRING", "None", "Optional", "The", "Union", "ValueError", "You", "_CONFIG_FOR_DOC", "__init__", "add_start_docstrings_to_model_forward", "and", "any", "any_reader_correct", "any_retriever_correct", "argmax", "attention_mask", "attentions", "be", "block", "block_idx", "block_mask", "bool", "calculate", "candidate", "candidate_ends", "candidate_starts", "clamp", "class", "cls", "compute_correct_candidates", "config", "config_class", "def", "dim", "dtype", "else", "end_pos", "end_positions", "eq", "equal", "evidence", "finfo", "float32", "format", "forward", "gold_ends", "gold_starts", "greater", "has_answers", "have", "head_mask", "hidden_states", "if", "ignored_index", "index", "index_select", "input", "input_ids", "inputs_embeds", "is", "is_correct", "is_gold_end", "is_gold_start", "length", "log_denominator", "log_numerator", "logical_and", "logits", "logsumexp", "loss", "marginal_log_loss", "mask", "mask_to_score", "max", "max_span_width", "mean", "min", "must", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "position_ids", "post_init", "predicted_block_index", "predicted_candidate", "predicted_end", "predicted_start", "qa_outputs", "question", "r", "raise", "reader_beam_size", "reader_correct", "reader_logits", "reader_loss", "relevance_score", "replace_return_docstrings", "retriever_correct", "retriever_logits", "retriever_loss", "return", "return_dict", "self", "separate", "sequence", "sequence_length", "sequence_output", "size", "specify", "start_pos", "start_positions", "super", "than", "to", "token_type_ids", "torch", "total_loss", "tuple", "type", "unsqueeze", "use_return_dict", "values", "view"], "deprecated/realm/modeling_realm.py:RealmForOpenQA": ["B", "BD", "D", "FloatTensor", "LongTensor", "ModelEmbedder", "ModelForOpenQA", "ModelForOpenQAOutput", "ModelPreTrainedModel", "ModelReader", "Model_FOR_OPEN_QA_DOCSTRING", "None", "Optional", "QB", "QD", "The", "True", "Union", "ValueError", "_", "_CONFIG_FOR_DOC", "__init__", "add_start_docstrings_to_model_forward", "and", "answer_ids", "attention_mask", "batch_scores", "batch_size", "be", "block_emb", "block_embedding_to", "block_idx", "block_mask", "bool", "class", "concat_inputs", "config", "config_class", "cpu", "def", "device", "dim", "dtype", "einsum", "else", "embedder", "end_pos", "end_positions", "float32", "format", "forward", "has_answers", "if", "index", "index_select", "input_ids", "inputs", "is", "k", "logical_and_", "logical_not_", "long", "max_length", "must", "new_empty", "not", "num_block_records", "of", "output_type", "post_init", "predicted_answer_ids", "predicted_block", "property", "question_outputs", "question_projection", "r", "raise", "reader", "reader_beam_size", "reader_output", "reader_seq_len", "register_buffer", "relevance_score", "replace_return_docstrings", "retrieved_block_emb", "retrieved_block_ids", "retrieved_logits", "retriever", "retriever_proj_size", "return", "return_dict", "searcher_beam_size", "self", "sequence_length", "shape", "size", "special_tokens_mask", "squeeze", "start_pos", "start_positions", "super", "tensor", "the", "to", "token_type_ids", "topk", "torch", "training", "tuple", "type", "use_return_dict", "zeros"], "deprecated/transfo_xl/modeling_transfo_xl_utilities.py:ProjectedAdaptiveLogSoftmax": ["False", "FloatTensor", "Input", "Linear", "ModelAdaptiveLogSoftmax", "Module", "ModuleList", "None", "Parameter", "ParameterList", "RuntimeError", "__init__", "_compute_logit", "and", "append", "batch", "bias", "bias_i", "biases", "cat", "class", "cluster_bias", "cluster_prob_idx", "cluster_weight", "contiguous", "continue", "copy_", "cutoff_ends", "cutoff_values", "cutoffs", "d_emb_i", "d_embed", "d_proj", "def", "device", "dim", "dimension", "div_val", "dtype", "else", "for", "forward", "functional", "gather", "hasattr", "have", "head_bias", "head_logit", "head_logprob", "head_logprob_i", "head_proj", "head_size", "head_weight", "hidden", "hidden_i", "i", "if", "in", "index_copy_", "index_select", "indices_i", "is", "keep_order", "l_idx", "labels", "len", "linear", "log_prob", "log_softmax", "logit", "logprob_i", "mask", "mask_i", "n_clusters", "n_token", "new_empty", "nn", "nonzero", "not", "numel", "offset", "or", "out", "out_layers", "out_projs", "proj", "proj_hid", "proj_i", "r", "r_idx", "raise", "range", "return", "same", "self", "shortlist_size", "should", "size", "squeeze", "start_idx", "stop_idx", "super", "t", "tail_logit_i", "tail_logprob_i", "target_i", "the", "torch", "unsqueeze", "view", "weight", "weight_i", "weights", "zeros", "zeros_like"], "deprecated/transfo_xl/modeling_transfo_xl.py:PositionalEmbedding": ["ModelEmbedding", "Module", "None", "__init__", "arange", "bsz", "cat", "class", "cos", "def", "demb", "dim", "else", "expand", "forward", "if", "inv_freq", "is", "nn", "not", "outer", "pos_emb", "pos_seq", "register_buffer", "return", "self", "sin", "sinusoid_inp", "super", "torch"], "deprecated/transfo_xl/modeling_transfo_xl.py:PositionwiseFF": ["CoreNet", "Dropout", "False", "LayerNorm", "Linear", "ModelFF", "Module", "ReLU", "Sequential", "True", "__init__", "class", "core_out", "d_inner", "d_model", "def", "dropout", "else", "eps", "forward", "if", "inp", "inplace", "layer_norm", "layer_norm_epsilon", "nn", "output", "pre_lnorm", "return", "self", "super"], "deprecated/transfo_xl/modeling_transfo_xl.py:RelPartialLearnableMultiHeadAttn": ["AC", "BD", "Dropout", "False", "FloatTensor", "LayerNorm", "Linear", "ModelPartialLearnableMultiHeadAttn", "Module", "None", "Parameter", "_Model_shift", "__init__", "and", "append", "attn_mask", "attn_out", "attn_prob", "attn_score", "attn_vec", "bias", "bsz", "cat", "chunk", "class", "contiguous", "d_head", "d_model", "def", "device", "dim", "drop", "dropatt", "dropout", "dtype", "einsum", "elif", "else", "eps", "finfo", "float", "forward", "functional", "head_mask", "ibnd", "if", "ijbn", "is", "item", "jbnd", "jnd", "klen", "layer_norm", "layer_norm_epsilon", "mask_value", "masked_fill", "mems", "min", "mul_", "n_head", "nn", "not", "o_net", "or", "output_attentions", "outputs", "pre_lnorm", "qkv_net", "qlen", "r", "r_head_k", "r_net", "r_r_bias", "r_w_bias", "return", "rlen", "rr_head_q", "rw_head_q", "scale", "self", "size", "softmax", "sum", "super", "torch", "type_as", "view", "view_as", "w", "w_head_k", "w_head_q", "w_head_v", "w_heads", "x", "x_padded", "x_padded_shape", "zero_pad", "zero_pad_shape", "zeros"], "deprecated/transfo_xl/modeling_transfo_xl.py:RelPartialLearnableDecoderLayer": ["False", "ModelPartialLearnableDecoderLayer", "ModelPartialLearnableMultiHeadAttn", "Module", "None", "PositionwiseFF", "__init__", "attn_mask", "attn_outputs", "class", "d_head", "d_inner", "d_model", "dec_attn", "dec_attn_mask", "dec_inp", "def", "dropout", "ff_output", "forward", "get", "head_mask", "kwargs", "layer_norm_epsilon", "mems", "n_head", "nn", "output_attentions", "outputs", "pos_ff", "pre_lnorm", "r", "return", "self", "super"], "deprecated/transfo_xl/modeling_transfo_xl.py:AdaptiveEmbedding": ["Embedding", "False", "FloatTensor", "ModelEmbedding", "Module", "ModuleList", "Parameter", "ParameterList", "__init__", "append", "class", "continue", "cutoff_ends", "cutoffs", "d_emb_i", "d_embed", "d_proj", "def", "device", "div_val", "dtype", "else", "emb_flat", "emb_i", "emb_layers", "emb_projs", "emb_scale", "embed", "embed_shape", "for", "forward", "functional", "i", "if", "in", "index_copy_", "index_select", "indices_i", "inp", "inp_flat", "inp_i", "l_idx", "len", "linear", "mask_i", "mul_", "n_token", "next", "nn", "nonzero", "numel", "param", "parameters", "r_idx", "range", "return", "sample_softmax", "self", "size", "sparse", "squeeze", "super", "torch", "view", "zeros"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLPreTrainedModel": ["AdaptiveEmbedding", "Embedding", "LayerNorm", "Linear", "ModelXLConfig", "ModelXLPreTrainedModel", "Modelrmer", "None", "Optional", "PreTrainedModel", "ProjectedAdaptiveLogSoftmax", "The", "__class__", "__name__", "_get_embedding_shapes", "_get_new_num_tokens_layer", "_get_resized_embeddings", "_init_bias", "_init_weight", "_init_weights", "_resize_cutoffs", "_resize_token_embeddings", "and", "assert", "base_model", "base_model_prefix", "be", "bias", "cannot", "class", "classname", "cluster_bias", "cluster_weight", "config", "constant_", "cutoff_ends", "cutoffs", "def", "elif", "else", "emb", "emb_layers", "emb_projs", "embedding", "embeddings", "find", "for", "get_input_embeddings", "getattr", "hasattr", "i", "if", "in", "init", "init_range", "init_std", "int", "is", "layer", "len", "less", "m", "model_embeds", "n_token", "new", "new_emb_size", "new_embedding_shapes", "new_embeddings_layer", "new_num_tokens", "new_num_tokens_layer", "nn", "normal", "normal_", "not", "of", "or", "out_projs", "proj_init_std", "r_bias", "r_emb", "r_r_bias", "r_w_bias", "range", "resize_token_embeddings", "return", "self", "set_input_embeddings", "shape", "size", "sum", "the", "tie_weights", "uniform", "uniform_", "vocab_size", "weight"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLModelOutput": ["FloatTensor", "ModelOutput", "ModelXLModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "list", "mems", "torch", "tuple"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLSequenceClassifierOutputWithPast": ["FloatTensor", "ModelOutput", "ModelXLSequenceClassifierOutputWithPast", "None", "Optional", "attentions", "class", "hidden_states", "list", "logits", "loss", "mems", "torch", "tuple"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLLMHeadModelOutput": ["FloatTensor", "ModelOutput", "ModelXLLMHeadModelOutput", "None", "Optional", "attentions", "class", "def", "hidden_states", "list", "logits", "loss", "losses", "mems", "prediction_scores", "property", "return", "self", "torch", "tuple"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLModel": ["AdaptiveEmbedding", "Dropout", "FloatTensor", "Head", "LongTensor", "ModelXLModel", "ModelXLModelOutput", "ModelXLPreTrainedModel", "Model_XL_INPUTS_DOCSTRING", "Modelrmer", "ModuleList", "None", "NotImplementedError", "Optional", "Parameter", "PositionalEmbedding", "RelPartialLearnableDecoderLayer", "Union", "ValueError", "XL", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_prune_heads", "_update_mems", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "all_ones", "and", "append", "arange", "assert", "at", "attentions", "attn_type", "backward_compatible", "beg_idx", "bool", "both", "bsz", "cannot", "cat", "checkpoint", "clamp_", "clamp_len", "class", "config", "config_class", "contiguous", "core_out", "cutoffs", "d_embed", "d_head", "d_inner", "d_model", "dec_attn_mask", "def", "detach", "device", "diagonal", "dim", "div_val", "drop", "dropatt", "dropout", "dtype", "either", "elif", "else", "empty", "end_idx", "enumerate", "expand", "for", "forward", "get_input_embeddings", "have", "head_mask", "heads", "hidden_states", "hids", "i", "if", "implemented", "in", "info", "init_mems", "input_ids", "inputs_embeds", "int64", "is", "klen", "last_hidden_state", "layer", "layer_norm_epsilon", "layer_outputs", "layers", "len", "list", "logger", "mask_len", "mask_shift_len", "max", "mem_len", "mems", "mems_i", "mlen", "model", "n_head", "n_layer", "n_token", "new_embeddings", "new_mems", "new_ones", "next", "nn", "no_grad", "not", "or", "output_attentions", "output_hidden_states", "output_type", "param", "parameters", "pass", "permute", "pos_emb", "pos_seq", "post_init", "pre_lnorm", "pruning", "qlen", "r_r_bias", "r_w_bias", "raise", "range", "reset_memory_length", "return", "return_dict", "same", "same_length", "sample_softmax", "self", "set_input_embeddings", "shape", "size", "specify", "super", "t", "the", "time", "to", "torch", "transpose", "tril", "triu", "tuple", "type_as", "unsqueeze", "untie_r", "use_return_dict", "v", "vocab_size", "with", "word_emb", "zeros"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLLMHeadModel": ["DeprecationWarning", "False", "FloatTensor", "In", "LongTensor", "ModelXL", "ModelXLLMHeadModel", "ModelXLLMHeadModelOutput", "ModelXLModel", "ModelXLPreTrainedModel", "Model_XL_INPUTS_DOCSTRING", "Modelrmer", "Modelrmer_outputs", "Modelrmers", "None", "Optional", "Parameter", "Please", "ProjectedAdaptiveLogSoftmax", "Sampling", "Tensor", "The", "True", "Union", "ValueError", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_reorder_cache", "_resize_cutoffs", "_tie_or_clone_weights", "_tied_weights_keys", "a", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "and", "argument", "as", "assert", "at", "attentions", "attribute", "be", "beam_idx", "bool", "bsz", "checkpoint", "class", "clone", "com", "config", "config_class", "configuration", "crit", "cutoff_ends", "cutoffs", "d", "d_embed", "d_model", "def", "device", "div_val", "either", "elif", "else", "emb_layers", "emb_projs", "enumerate", "eos_token_id", "first", "for", "forward", "from", "get_output_embeddings", "getattr", "github", "have", "head_mask", "hidden_states", "https", "huggingface", "i", "if", "implemented", "in", "index_select", "init_mems", "input_ids", "inputs", "inputs_embeds", "is", "issue", "issues", "labels", "last_hidden", "layer", "layer_past", "len", "list", "look", "loss", "losses", "mean", "mem_len", "mems", "miss_valid_label", "model_kwargs", "n_token", "new_cutoffs", "new_emb_size", "new_embedding_shapes", "new_num_tokens", "nn", "not", "of", "or", "order", "out_layer", "out_layers", "out_projs", "output", "output_attentions", "output_hidden_states", "output_type", "past_key_values", "please", "post_init", "pred_hid", "prediction_scores", "prepare_inputs_for_generation", "r", "raise", "range", "reset_memory_length", "return", "return_dict", "sample_softmax", "self", "single", "size", "softmax", "softmax_output", "specify", "staticmethod", "sum", "super", "support", "tgt_len", "that", "the", "tie_proj", "tie_projs", "tie_weights", "tie_word_embeddings", "to", "torch", "torchscript", "trainer_compatible", "tuple", "unsqueeze", "updated", "use", "use_return_dict", "v5", "view", "vocab_size", "warn", "warnings", "weight", "will", "word_emb", "yet", "your"], "deprecated/transfo_xl/modeling_transfo_xl.py:TransfoXLForSequenceClassification": ["BCEWithLogitsLoss", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "ModelXLForSequenceClassification", "ModelXLModel", "ModelXLPreTrainedModel", "ModelXLSequenceClassifierOutputWithPast", "Model_XL_INPUTS_DOCSTRING", "Modelrmer", "Modelrmer_outputs", "None", "Optional", "Results", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__class__", "__init__", "__name__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "and", "argmax", "assert", "attentions", "batch", "batch_size", "be", "bias", "bool", "checkpoint", "class", "config", "config_class", "conjunction", "d_embed", "def", "defined", "detect", "device", "dtype", "elif", "else", "eq", "f", "forward", "handle", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "int", "is", "labels", "list", "logger", "logits", "long", "loss", "loss_fct", "may", "mems", "multi_label_classification", "nn", "no", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "output_type", "pad_token_id", "padding", "pooled_logits", "post_init", "problem_type", "r", "range", "regression", "return", "return_dict", "score", "self", "sequence_length", "sequence_lengths", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "tokens", "torch", "tuple", "unexpected", "use_return_dict", "using", "view", "warning_once", "will", "with"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "Model", "Module", "None", "Optional", "Tensor", "__init__", "absolute", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "getattr", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "int", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embedding_type", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertSelfAttention": ["Dropout", "Embedding", "False", "Model", "Module", "None", "QuantLinear", "Softmax", "TensorQuantizer", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bhld", "bhlr", "bhrd", "cat", "class", "config", "context_layer", "contiguous", "def", "default_quant_desc_input", "deprecate_kwarg", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "long", "lrd", "math", "matmul", "matmul_a_input_quantizer", "matmul_k_input_quantizer", "matmul_q_input_quantizer", "matmul_v_input_quantizer", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_name", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "past_key_value", "past_key_values", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "quant_nn", "query", "query_layer", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "seq_length", "size", "sqrt", "super", "the", "to", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "version", "view", "x"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertSelfOutput": ["Dropout", "LayerNorm", "Model", "Module", "QuantLinear", "TensorQuantizer", "__init__", "add_local", "add_local_input_quantizer", "add_residual", "add_residual_input_quantizer", "class", "config", "def", "default_quant_desc_input", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "quant_nn", "return", "self", "super"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertAttention": ["False", "Model", "Module", "None", "QDQBertSelfAttention", "QDQBertSelfOutput", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value", "version"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertIntermediate": ["ACT2FN", "Model", "Module", "QuantLinear", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "quant_nn", "return", "self", "str", "super"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertOutput": ["Dropout", "LayerNorm", "Model", "Module", "QuantLinear", "TensorQuantizer", "__init__", "add_local", "add_local_input_quantizer", "add_residual", "add_residual_input_quantizer", "class", "config", "def", "default_quant_desc_input", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "quant_nn", "return", "self", "super"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertLayer": ["False", "GradientCheckpointingLayer", "If", "Model", "None", "QDQBertAttention", "QDQBertIntermediate", "QDQBertOutput", "True", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "are", "as", "attention", "attention_mask", "attention_output", "be", "by", "class", "config", "cross", "cross_attention_outputs", "cross_attn_past_key_value", "cross_attn_present_key_value", "crossattention", "decoder", "def", "deprecate_kwarg", "else", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_output", "layers", "model", "new_name", "not", "output", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "present_key_value", "raise", "return", "self", "self_attention_outputs", "self_attn_past_key_value", "seq_len_dim", "setting", "should", "super", "to", "used", "version", "with"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "False", "Model", "Module", "ModuleList", "None", "QDQBertLayer", "True", "_", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "next_decoder_cache", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "past_key_values", "range", "return", "return_dict", "self", "super", "tuple", "use_cache", "v"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertPooler": ["Linear", "Model", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "Model", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertLMPredictionHead": ["False", "Linear", "Model", "Module", "Parameter", "QDQBertPredictionHeadTransform", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertOnlyMLMHead": ["Model", "Module", "QDQBertLMPredictionHead", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertOnlyNSPHead": ["Linear", "Model", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertPreTrainingHeads": ["Linear", "Model", "Module", "QDQBertLMPredictionHead", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "None", "PreTrainedModel", "QDQBertConfig", "True", "_init_weights", "base_model_prefix", "bert", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "False", "FloatTensor", "LongTensor", "Model", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertEmbeddings", "QDQBertEncoder", "QDQBertPooler", "QDQBertPreTrainedModel", "Tensor", "True", "Union", "ValueError", "You", "_", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_prune_heads", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "at", "attention", "attention_mask", "attentions", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "checkpoint", "class", "config", "config_class", "cross_attentions", "def", "device", "dict", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "expand", "extended_attention_mask", "for", "format", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "int", "invert_attention_mask", "is", "is_decoder", "items", "last_hidden_state", "layer", "list", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "output_type", "past_key_values", "past_key_values_length", "pooled_output", "pooler", "pooler_output", "position_ids", "post_init", "prune_heads", "pytorch_quantization", "r", "raise", "requires_backends", "return", "return_dict", "same", "self", "seq_length", "sequence_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertLMHeadModel": ["CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "Model", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertOnlyMLMHead", "QDQBertPreTrainedModel", "Tensor", "True", "Union", "_CONFIG_FOR_DOC", "__init__", "_reorder_cache", "_tied_weights_keys", "a", "add", "add_pooling_layer", "add_start_docstrings_to_model_forward", "as", "attention_mask", "attentions", "batch_size", "beam_idx", "bert", "bias", "bool", "class", "cls", "config", "config_class", "contiguous", "cross_attentions", "decoder", "def", "device", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "format", "forward", "get_output_embeddings", "get_seq_length", "head_mask", "hidden_states", "if", "in", "index_select", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "layer_past", "lm_loss", "logger", "logits", "loss", "loss_fct", "model_kwargs", "new_embeddings", "new_ones", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "past_length", "past_state", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "remove_prefix_length", "reordered_past", "replace_return_docstrings", "return", "return_dict", "self", "sequence_length", "sequence_output", "set_output_embeddings", "shape", "shifted_prediction_scores", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertForMaskedLM": ["CrossEntropyLoss", "False", "FloatTensor", "If", "LongTensor", "MaskedLMOutput", "Model", "None", "Optional", "PAD", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertOnlyMLMHead", "QDQBertPreTrainedModel", "The", "Union", "ValueError", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_tied_weights_keys", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention", "attention_mask", "attentions", "batch_size", "be", "bert", "bi", "bias", "bool", "cat", "checkpoint", "class", "cls", "config", "config_class", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "format", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pad_token_id", "position_ids", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_length", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertForNextSentencePrediction": ["CrossEntropyLoss", "FloatTensor", "FutureWarning", "LongTensor", "Model", "NextSentencePredictorOutput", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertOnlyNSPHead", "QDQBertPreTrainedModel", "The", "Union", "_CONFIG_FOR_DOC", "__init__", "a", "add_start_docstrings_to_model_forward", "and", "argument", "attention_mask", "attentions", "batch_size", "be", "bert", "bool", "class", "cls", "config", "config_class", "def", "else", "format", "forward", "future", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "pop", "position_ids", "post_init", "r", "removed", "replace_return_docstrings", "return", "return_dict", "self", "seq_relationship_scores", "sequence_length", "super", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "version", "view", "warn", "warnings", "will"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "MSELoss", "Model", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertPreTrainedModel", "SequenceClassifierOutput", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bert", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "dropout", "dtype", "elif", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "position_ids", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_length", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "MultipleChoiceModelOutput", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertPreTrainedModel", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bert", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "dropout", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertPreTrainedModel", "TokenClassifierOutput", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bert", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "dropout", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "position_ids", "post_init", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/qdqbert/modeling_qdqbert.py:QDQBertForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "None", "Optional", "QDQBERT_INPUTS_DOCSTRING", "QDQBertModel", "QDQBertPreTrainedModel", "QuestionAnsweringModelOutput", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bert", "bool", "checkpoint", "clamp", "class", "config", "config_class", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "format", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "position_ids", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "deprecated/tvlt/modeling_tvlt.py:TvltModelOutput": ["FloatTensor", "LongTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "audio_ids_restore", "audio_label_masks", "class", "hidden_states", "last_audio_hidden_state", "last_hidden_state", "last_pixel_hidden_state", "pixel_ids_restore", "pixel_label_masks", "torch", "tuple"], "deprecated/tvlt/modeling_tvlt.py:TvltDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "torch", "tuple"], "deprecated/tvlt/modeling_tvlt.py:TvltForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "audio_logits", "class", "hidden_states", "loss", "matching_logits", "pixel_logits", "torch", "tuple"], "deprecated/tvlt/modeling_tvlt.py:generate_pixel_mask_noise": ["Model_pixel_mask_noise", "None", "batch_size", "def", "device", "int", "len_keep", "mask_ratio", "noise", "pixel_mask", "pixel_values", "rand", "return", "seq_len", "shape", "torch"], "deprecated/tvlt/modeling_tvlt.py:generate_audio_mask_noise": ["Model_audio_mask_noise", "None", "audio_mask", "audio_values", "batch_size", "def", "device", "elif", "frame", "freq_len", "if", "int", "len_keep", "level", "mask_ratio", "mask_type", "noise", "num_time_patches", "patch", "rand", "repeat", "return", "seq_len", "shape", "torch", "unsqueeze", "view"], "deprecated/tvlt/modeling_tvlt.py:random_masking": ["Model_masking", "None", "argsort", "attention_masks", "batch_size", "def", "device", "dim", "gather", "hidden_dim", "ids_keep", "ids_restore", "ids_shuffle", "if", "index", "is", "label_masks", "len_keep", "noise", "not", "ones", "repeat", "return", "seq_len", "sequence", "sequence_masked", "shape", "torch", "unsqueeze"], "deprecated/tvlt/modeling_tvlt.py:TvltPixelEmbeddings": ["ModelPixelEmbeddings", "ModelPixelPatchEmbeddings", "Module", "None", "Parameter", "__init__", "attention_masks", "batch_size", "class", "config", "def", "dim", "embeddings", "forward", "height", "hidden_size", "nn", "num_channels", "num_frames", "num_patches_per_image", "patch_embeddings", "pixel_values", "pos_embed_v", "repeat", "repeat_interleave", "return", "self", "shape", "super", "temporal_embed", "torch", "type_embed_v", "width", "zeros"], "deprecated/tvlt/modeling_tvlt.py:TvltAudioEmbeddings": ["ModelAudioEmbeddings", "ModelAudioPatchEmbeddings", "Module", "None", "Parameter", "__init__", "attention_masks", "audio_patch_size", "audio_values", "class", "config", "def", "dim", "embeddings", "forward", "freq_embed", "frequency_length", "hidden_size", "nn", "num_freq_patches", "num_patches", "num_time_patches", "patch_embeddings", "pos_embed_a", "repeat", "repeat_interleave", "return", "self", "size", "super", "torch", "type_embed_a", "zeros"], "deprecated/tvlt/modeling_tvlt.py:TvltPixelPatchEmbeddings": ["Conv2d", "Input", "Iterable", "Make", "ModelPixelPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "flatten", "forward", "height", "hidden_size", "if", "image", "image_patch_size", "image_size", "in", "isinstance", "kernel_size", "match", "model", "nn", "num_channels", "num_frames", "num_image_channels", "num_patches_per_image", "of", "one", "or", "patch_size", "pixel", "pixel_values", "projection", "raise", "reshape", "return", "self", "set", "shape", "size", "stride", "super", "sure", "t", "that", "the", "torch", "transpose", "values", "width", "with"], "deprecated/tvlt/modeling_tvlt.py:TvltAudioPatchEmbeddings": ["Conv2d", "Input", "Iterable", "Make", "ModelAudioPatchEmbeddings", "Module", "Tensor", "ValueError", "__init__", "abc", "audio", "audio_patch_size", "audio_values", "batch_size", "channel", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "flatten", "forward", "frequency_length", "height", "hidden_size", "if", "in", "isinstance", "kernel_size", "match", "model", "nn", "num_audio_channels", "num_channels", "num_patches", "of", "one", "or", "patch_shape", "patch_size", "pixel", "projection", "raise", "return", "self", "set", "shape", "size", "spectrogram_length", "spectrogram_size", "stride", "super", "sure", "t", "that", "the", "torch", "transpose", "values", "width", "with"], "deprecated/tvlt/modeling_tvlt.py:TvltSelfAttention": ["Dropout", "False", "Linear", "ModelSelfAttention", "Module", "None", "Softmax", "The", "ValueError", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bias", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "self", "size", "sqrt", "super", "the", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "view", "x"], "deprecated/tvlt/modeling_tvlt.py:TvltSelfOutput": ["Dropout", "Linear", "ModelConfig", "ModelSelfOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "deprecated/tvlt/modeling_tvlt.py:TvltAttention": ["False", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "__init__", "all_head_size", "attention", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "deprecated/tvlt/modeling_tvlt.py:TvltIntermediate": ["ACT2FN", "Linear", "ModelConfig", "ModelIntermediate", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deprecated/tvlt/modeling_tvlt.py:TvltOutput": ["Dropout", "Linear", "ModelConfig", "ModelOutput", "Module", "None", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "deprecated/tvlt/modeling_tvlt.py:TvltLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "__init__", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "device", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "to"], "deprecated/tvlt/modeling_tvlt.py:TvltEncoder": ["BaseModelOutput", "False", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attention_mask", "attentions", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "deprecated/tvlt/modeling_tvlt.py:TvltPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "deprecated/tvlt/modeling_tvlt.py:TvltModel": ["False", "FloatTensor", "LayerNorm", "ModelAudioEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPixelEmbeddings", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Parameter", "Union", "_CONFIG_FOR_DOC", "__init__", "_prune_heads", "add_start_docstrings_to_model_forward", "and", "attention", "attention_mask", "attention_masks", "attentions", "audio_embedding_output", "audio_embeddings", "audio_ids_restore", "audio_label_masks", "audio_len_keep", "audio_mask", "audio_mask_noise", "audio_mask_ratio", "audio_mask_type", "audio_patch_size", "audio_sequence_output", "audio_values", "batch_size", "bool", "cat", "class", "cls_embedding", "config", "config_class", "def", "else", "embedding_output", "encoder", "encoder_outputs", "eps", "extended_attention_mask", "for", "forward", "freq_len", "frequency_length", "generate_audio_mask_noise", "generate_pixel_mask_noise", "get_extended_attention_mask", "get_input_embeddings", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "input_shape", "is", "items", "last_audio_hidden_state", "last_hidden_state", "last_pixel_hidden_state", "layer", "layer_norm_eps", "layernorm", "mask_audio", "mask_pixel", "mask_ratio", "mask_type", "masked_pixel_len", "nn", "not", "num_freq_patches", "output_attentions", "output_hidden_states", "output_type", "patch_embeddings", "pixel_embedding_output", "pixel_embeddings", "pixel_ids_restore", "pixel_label_masks", "pixel_len_keep", "pixel_mask", "pixel_mask_noise", "pixel_mask_ratio", "pixel_sequence_output", "pixel_values", "post_init", "prune_heads", "r", "random_masking", "repeat", "replace_return_docstrings", "return", "return_dict", "self", "sequence_output", "size", "super", "torch", "tuple", "use_mean_pooling", "use_return_dict", "zeros"], "deprecated/tvlt/modeling_tvlt.py:TvltDecoder": ["False", "LayerNorm", "ModelDecoder", "ModelDecoderOutput", "ModelLayer", "Module", "ModuleList", "None", "True", "_", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "class", "config", "decoder_config", "decoder_hidden_size", "decoder_intermediate_size", "decoder_layers", "decoder_num_attention_heads", "decoder_num_hidden_layers", "deepcopy", "def", "else", "enumerate", "eps", "for", "forward", "gradient_checkpointing", "hidden_size", "hidden_states", "i", "if", "in", "intermediate_size", "is", "layer_module", "layer_norm_eps", "layer_outputs", "layernorm", "logits", "nn", "not", "num_attention_heads", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "tuple", "v"], "deprecated/tvlt/modeling_tvlt.py:TvltForPreTraining": ["BCEWithLogitsLoss", "FloatTensor", "Linear", "LongTensor", "MAE", "Matching", "Model", "ModelDecoder", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelMAEHead", "ModelMatchingHead", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "Must", "None", "Optional", "Parameter", "True", "Union", "ValueError", "_CONFIG_FOR_DOC", "__init__", "add_start_docstrings_to_model_forward", "and", "at", "attentions", "audio_decoder_input", "audio_decoder_outputs", "audio_embeddings", "audio_ids_restore", "audio_label_masks", "audio_logits", "audio_mae_head", "audio_mae_loss", "audio_mae_output_dim", "audio_mask", "audio_mask_token", "audio_patch_size", "audio_predictions", "audio_sequence_output", "audio_values", "batch_size", "bias", "bool", "cat", "class", "concatenate_mask", "config", "config_class", "decoder", "decoder_audio_pos_embed", "decoder_audio_type_embed", "decoder_freq_embed", "decoder_hidden_size", "decoder_pixel_pos_embed", "decoder_pixel_type_embed", "decoder_temporal_embed", "def", "dim", "einsum", "else", "encoder_to_decoder", "forward", "frequency_length", "gather", "height", "hidden_size", "hidden_states", "ids_restore", "if", "image_patch_size", "index", "is", "labels", "last_audio_hidden_state", "last_pixel_hidden_state", "least", "logits", "loss", "loss_fct", "mask", "mask_audio", "mask_pixel", "mask_token", "mask_tokens", "matching", "matching_head", "matching_logits", "mean", "nchpwq", "nhwpqc", "nn", "not", "ntchpwq", "nthwpqc", "num_audio_channels", "num_audio_patches", "num_channels", "num_frames", "num_freq_patches", "num_image_channels", "num_patches", "num_patches_height", "num_patches_per_image", "num_patches_width", "num_time_patches", "of", "one", "or", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "padded_sequence", "patchified_audio_values", "patchified_pixel_values", "patchify_audio", "patchify_pixel", "pixel_decoder_input", "pixel_decoder_outputs", "pixel_embeddings", "pixel_ids_restore", "pixel_label_masks", "pixel_logits", "pixel_mae_head", "pixel_mae_loss", "pixel_mae_output_dim", "pixel_mask", "pixel_mask_mixed", "pixel_mask_token", "pixel_predictions", "pixel_sequence_output", "pixel_values", "pixel_values_mixed", "post_init", "r", "raise", "repeat", "repeat_interleave", "replace_return_docstrings", "requires", "reshape", "return", "return_dict", "self", "seq_length", "sequence", "sequence_output", "set", "shape", "size", "sum", "super", "task", "task_mae", "task_matching", "to", "torch", "total_loss", "training", "true", "tuple", "unsqueeze", "use_return_dict", "view", "width", "zeros"], "deprecated/tvlt/modeling_tvlt.py:TvltPooler": ["Linear", "ModelPooler", "Module", "Tanh", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super"], "deprecated/tvlt/modeling_tvlt.py:TvltMatchingHead": ["Linear", "ModelMatchingHead", "ModelPooler", "Module", "__init__", "class", "config", "def", "fc", "forward", "hidden_size", "hidden_states", "nn", "pooler", "return", "self", "super"], "deprecated/tvlt/modeling_tvlt.py:TvltMAEHead": ["Linear", "ModelMAEHead", "Module", "None", "__init__", "class", "config", "decoder", "decoder_hidden_size", "def", "forward", "hidden_states", "nn", "output_dim", "return", "self", "super"], "deprecated/tvlt/modeling_tvlt.py:TvltForAudioVisualClassification": ["CrossEntropyLoss", "FloatTensor", "GELU", "LayerNorm", "Linear", "LongTensor", "MSELoss", "Model", "ModelForAudioVisualClassification", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "SequenceClassifierOutput", "Sequential", "Union", "_CONFIG_FOR_DOC", "__init__", "add_start_docstrings_to_model_forward", "attentions", "audio_mask", "audio_values", "bool", "class", "classification", "classifier", "config", "config_class", "def", "elif", "else", "eps", "forward", "hidden_size", "hidden_states", "if", "is", "labels", "layer_norm_eps", "logits", "loss", "loss_fct", "loss_type", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pixel_mask", "pixel_values", "post_init", "r", "regression", "replace_return_docstrings", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "deprecated/deta/modeling_deta.py:load_cuda_kernels": ["DCUDA_HAS_FP16", "DWITH_CUDA", "D__CUDA_NO_HALF2_OPERATORS__", "D__CUDA_NO_HALF_CONVERSIONS__", "D__CUDA_NO_HALF_OPERATORS__", "Model", "Model_cuda_kernels", "MultiScaleDeformableAttention", "Path", "True", "__file__", "cpp", "cpu", "cu", "cuda", "def", "deta", "extra_cflags", "extra_cuda_cflags", "extra_include_paths", "filename", "for", "global", "in", "join", "kernels", "ms_deform_attn_cpu", "ms_deform_attn_cuda", "os", "parent", "path", "resolve", "root", "src_files", "str", "vision", "with_cuda"], "deprecated/deta/modeling_deta.py:MultiScaleDeformableAttentionFunction": ["Function", "ModelScaleDeformableAttention", "ModelScaleDeformableAttentionFunction", "None", "attention_weights", "backward", "class", "context", "def", "forward", "grad_attn_weight", "grad_output", "grad_sampling_loc", "grad_value", "im2col_step", "ms_deform_attn_backward", "ms_deform_attn_forward", "once_differentiable", "output", "return", "sampling_locations", "save_for_backward", "saved_tensors", "staticmethod", "value", "value_level_start_index", "value_spatial_shapes"], "deprecated/deta/modeling_deta.py:DetaDecoderOutput": ["FloatTensor", "ModelDecoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "torch", "tuple"], "deprecated/deta/modeling_deta.py:DetaModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "output_proposals", "torch", "tuple"], "deprecated/deta/modeling_deta.py:DetaObjectDetectionOutput": ["FloatTensor", "ModelObjectDetectionOutput", "ModelOutput", "None", "Optional", "auxiliary_outputs", "class", "cross_attentions", "decoder_attentions", "decoder_hidden_states", "dict", "enc_outputs_class", "enc_outputs_coord_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "init_reference_points", "intermediate_hidden_states", "intermediate_reference_points", "last_hidden_state", "list", "logits", "loss", "loss_dict", "output_proposals", "pred_boxes", "torch", "tuple"], "deprecated/deta/modeling_deta.py:_get_clones": ["ModuleList", "N", "_get_clones", "copy", "deepcopy", "def", "for", "i", "in", "module", "nn", "range", "return"], "deprecated/deta/modeling_deta.py:inverse_sigmoid": ["Model_sigmoid", "clamp", "def", "eps", "log", "max", "min", "return", "torch", "x", "x1", "x2"], "deprecated/deta/modeling_deta.py:DetaFrozenBatchNorm2d": ["ModelFrozenBatchNorm2d", "Module", "__init__", "_load_from_state_dict", "bias", "class", "def", "del", "epsilon", "error_msgs", "forward", "if", "in", "local_metadata", "missing_keys", "n", "nn", "num_batches_tracked", "num_batches_tracked_key", "ones", "prefix", "register_buffer", "reshape", "return", "rsqrt", "running_mean", "running_var", "scale", "self", "state_dict", "strict", "super", "torch", "unexpected_keys", "weight", "x", "zeros"], "deprecated/deta/modeling_deta.py:replace_batch_norm": ["BatchNorm2d", "DetaFrozenBatchNorm2d", "Model_batch_norm", "_modules", "bias", "children", "copy_", "data", "def", "device", "for", "if", "in", "isinstance", "len", "list", "meta", "model", "module", "name", "named_children", "new_module", "nn", "num_features", "r", "running_mean", "running_var", "torch", "weight"], "deprecated/deta/modeling_deta.py:DetaBackboneWithPositionalEncodings": ["False", "ModelBackboneWithPositionalEncodings", "Module", "None", "Tensor", "__init__", "and", "append", "backbone", "backbone_config", "bool", "build_position_encoding", "channels", "class", "config", "def", "dtype", "feature_map", "feature_maps", "features", "float", "for", "forward", "functional", "if", "in", "intermediate_channel_sizes", "interpolate", "load_backbone", "mask", "model", "model_type", "name", "named_parameters", "nn", "no_grad", "not", "out", "parameter", "pixel_mask", "pixel_values", "pos", "position_embedding", "position_embeddings", "replace_batch_norm", "requires_grad_", "resnet", "return", "self", "shape", "size", "stages", "super", "to", "torch", "with"], "deprecated/deta/modeling_deta.py:DetaSinePositionEmbedding": ["False", "ModelSinePositionEmbedding", "Module", "No", "None", "True", "ValueError", "__init__", "and", "arange", "be", "cat", "class", "cos", "cumsum", "def", "device", "dim", "dim_t", "div", "dtype", "embedding_dim", "eps", "flatten", "float", "float32", "floor", "forward", "if", "int64", "is", "mask", "math", "nn", "normalize", "not", "passed", "permute", "pi", "pixel", "pixel_mask", "pixel_values", "pos", "pos_x", "pos_y", "provided", "raise", "return", "rounding_mode", "scale", "self", "should", "sin", "stack", "super", "temperature", "torch", "x_embed", "y_embed"], "deprecated/deta/modeling_deta.py:DetaLearnedPositionEmbedding": ["Embedding", "ModelLearnedPositionEmbedding", "Module", "None", "__init__", "arange", "cat", "class", "column_embeddings", "def", "device", "dim", "embedding_dim", "forward", "height", "height_values", "nn", "permute", "pixel_mask", "pixel_values", "pos", "repeat", "return", "row_embeddings", "self", "shape", "super", "torch", "unsqueeze", "width", "width_values", "x_emb", "y_emb"], "deprecated/deta/modeling_deta.py:build_position_encoding": ["DetaLearnedPositionEmbedding", "DetaSinePositionEmbedding", "Model_position_encoding", "Not", "True", "ValueError", "config", "d_model", "def", "elif", "else", "f", "if", "learned", "n_steps", "normalize", "position_embedding", "position_embedding_type", "raise", "return", "sine", "supported"], "deprecated/deta/modeling_deta.py:multi_scale_deformable_attention": ["False", "Model_scale_deformable_attention", "Tensor", "_", "align_corners", "append", "attention_weights", "batch_size", "bilinear", "contiguous", "def", "dim", "enumerate", "flatten", "for", "functional", "grid_sample", "height", "hidden_dim", "in", "item", "level_id", "mode", "nn", "num_heads", "num_levels", "num_points", "num_queries", "output", "padding_mode", "reshape", "return", "sampling_grid_l_", "sampling_grids", "sampling_locations", "sampling_value_l_", "sampling_value_list", "shape", "split", "stack", "sum", "torch", "transpose", "value", "value_l_", "value_list", "value_spatial_shapes", "view", "width", "zeros"], "deprecated/deta/modeling_deta.py:DetaMultiscaleDeformableAttention": ["CUDA", "Could", "Exception", "F", "False", "Last", "Linear", "Make", "ModelConfig", "ModelMultiscaleDeformableAttention", "Module", "MultiScaleDeformableAttention", "MultiScaleDeformableAttentionFunction", "None", "Optional", "Parameter", "Tensor", "True", "ValueError", "You", "_", "__init__", "_reset_parameters", "a", "abs", "align", "and", "apply", "arange", "as", "attention", "attention_mask", "attention_weights", "authors", "batch_size", "be", "better", "bias", "bool", "but", "by", "class", "config", "constant_", "cos", "custom", "d", "d_model", "data", "def", "default_dtype", "deformable", "dim", "dim_per_head", "dimension", "disable_custom_kernels", "divisible", "dtype", "e", "each", "efficient", "elif", "else", "embed_dim", "encoder", "encoder_attention_mask", "encoder_hidden_states", "except", "f", "float", "for", "forward", "get_default_dtype", "got", "grid_init", "head", "hidden", "hidden_states", "i", "if", "im2col_step", "implementation", "in", "init", "int", "int64", "is", "is_ninja_available", "is_torch_cuda_available", "keepdim", "kernel", "kernel_loaded", "length", "level_start_index", "load", "load_cuda_kernels", "logger", "make", "masked_fill", "math", "max", "more", "multi", "multi_scale_deformable_attention", "must", "n_heads", "n_levels", "n_points", "nn", "no_grad", "not", "num_coordinates", "num_feature_levels", "num_heads", "num_queries", "of", "offset_normalizer", "or", "output", "output_attentions", "output_proj", "pi", "position_embeddings", "power", "raise", "range", "reference_points", "repeat", "return", "sampling_locations", "sampling_offsets", "scale", "self", "sequence", "sequence_length", "set", "shape", "shapes", "sin", "softmax", "spatial", "spatial_shapes", "stack", "states", "sum", "super", "sure", "tensor", "the", "thetas", "to", "torch", "try", "value", "value_proj", "view", "warn", "warning", "warnings", "weight", "which", "with", "with_pos_embed", "xavier_uniform_"], "deprecated/deta/modeling_deta.py:DetaMultiheadAttention": ["Attention", "False", "Linear", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_prepare_4d_attention_mask", "_shape", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bias", "bmm", "bool", "but", "by", "class", "contiguous", "def", "dim", "divisible", "dropout", "dtype", "else", "embed_dim", "f", "float", "forward", "functional", "got", "head_dim", "hidden_states", "hidden_states_original", "if", "inf", "int", "is", "k_proj", "key_states", "mask", "masked_fill_", "must", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "position_embeddings", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "should", "size", "softmax", "source_len", "super", "target_len", "tensor", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "view", "weights", "with_pos_embed", "zeros_like"], "deprecated/deta/modeling_deta.py:DetaEncoderLayer": ["ACT2FN", "False", "LayerNorm", "Linear", "ModelConfig", "ModelEncoderLayer", "ModelMultiscaleDeformableAttention", "Module", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "any", "attention_mask", "attn_weights", "bool", "clamp", "clamp_value", "class", "config", "d_model", "def", "dropout", "dtype", "embed_dim", "encoder_attention_heads", "encoder_attention_mask", "encoder_ffn_dim", "encoder_hidden_states", "encoder_n_points", "fc1", "fc2", "final_layer_norm", "finfo", "forward", "functional", "hidden_states", "if", "isinf", "isnan", "level_start_index", "max", "min", "n_points", "nn", "num_heads", "or", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "spatial_shapes", "super", "torch", "training"], "deprecated/deta/modeling_deta.py:DetaDecoderLayer": ["ACT2FN", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelConfig", "ModelDecoderLayer", "ModelMultiheadAttention", "ModelMultiscaleDeformableAttention", "None", "Optional", "Tensor", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "decoder_n_points", "def", "dropout", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "level_start_index", "n_points", "nn", "num_heads", "output_attentions", "outputs", "p", "position_embeddings", "reference_points", "residual", "return", "second_residual", "self", "self_attn", "self_attn_layer_norm", "self_attn_weights", "spatial_shapes", "super", "torch", "training"], "deprecated/deta/modeling_deta.py:DetaPreTrainedModel": ["BatchNorm2d", "Conv2d", "Embedding", "Linear", "ModelBackboneWithPositionalEncodings", "ModelConfig", "ModelDecoderLayer", "ModelEncoderLayer", "ModelLearnedPositionEmbedding", "ModelMultiscaleDeformableAttention", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "_reset_parameters", "and", "base_model_prefix", "bias", "class", "column_embeddings", "config", "constant_", "data", "def", "elif", "gain", "hasattr", "if", "init", "init_std", "is", "isinstance", "level_embed", "main_input_name", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "pixel_values", "r", "reference_points", "row_embeddings", "self", "std", "supports_gradient_checkpointing", "two_stage", "uniform_", "weight", "xavier_uniform_", "zero_"], "deprecated/deta/modeling_deta.py:DetaEncoder": ["BaseModelOutput", "False", "ModelConfig", "ModelEncoder", "ModelEncoderLayer", "ModelPreTrainedModel", "ModuleList", "None", "_", "__init__", "all_attentions", "append", "attention_mask", "attentions", "cat", "class", "config", "def", "device", "dropout", "dtype", "else", "encoder_layer", "encoder_layers", "encoder_states", "enumerate", "float32", "for", "forward", "functional", "get_reference_points", "gradient_checkpointing", "height", "hidden_states", "i", "if", "ij", "in", "indexing", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "level", "level_start_index", "linspace", "meshgrid", "nn", "not", "output_attentions", "output_hidden_states", "p", "position_embeddings", "post_init", "r", "range", "ref", "ref_x", "ref_y", "reference_points", "reference_points_list", "reshape", "return", "return_dict", "self", "spatial_shapes", "stack", "staticmethod", "super", "torch", "training", "tuple", "use_return_dict", "v", "valid_ratios", "width"], "deprecated/deta/modeling_deta.py:DetaDecoder": ["False", "ModelConfig", "ModelDecoder", "ModelDecoderLayer", "ModelDecoderOutput", "ModelPreTrainedModel", "Modelch", "ModuleList", "None", "Reference", "ValueError", "_", "__init__", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "attentions", "bbox_embed", "be", "but", "cat", "class", "class_embed", "config", "cross_attentions", "decoder_layer", "decoder_layers", "def", "dim", "dimension", "dropout", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "gradient_checkpointing", "hidden_states", "idx", "if", "in", "inputs_embeds", "intermediate", "intermediate_hidden_states", "intermediate_reference_points", "inverse_sigmoid", "is", "last", "last_hidden_state", "layer_outputs", "layers", "level_start_index", "must", "new_reference_points", "nn", "not", "of", "output_attentions", "output_hidden_states", "points", "position_embeddings", "post_init", "r", "raise", "range", "reference_points", "reference_points_input", "return", "return_dict", "self", "shape", "sigmoid", "size", "spatial_shapes", "stack", "super", "tmp", "torch", "tuple", "use_return_dict", "v", "valid_ratios"], "deprecated/deta/modeling_deta.py:DetaModel": ["BaseModelOutput", "Conv2d", "Embedding", "False", "FloatTensor", "GroupNorm", "LayerNorm", "Linear", "LongTensor", "ModelBackboneWithPositionalEncodings", "ModelConfig", "ModelDecoder", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "Modelch", "ModuleList", "No", "None", "Optional", "Parameter", "Sequential", "Tensor", "True", "Union", "ValueError", "WARNING", "_", "_CONFIG_FOR_DOC", "__init__", "_cur", "_len_sources", "add_start_docstrings_to_model_forward", "all", "and", "any", "append", "arange", "as_tensor", "assign_first_stage", "attention", "attention_mask", "attentions", "b", "backbone", "batch_size", "batched_nms", "bbox_embed", "bool", "cat", "center_to_corners_format", "clamp", "class", "class_embed", "config", "config_class", "cos", "cross_attentions", "cumsum", "d_model", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_outputs", "def", "delta_bbox", "device", "dim", "dim_t", "div", "dtype", "elif", "else", "enc_output", "enc_output_norm", "enc_outputs", "enc_outputs_class", "enc_outputs_coord_logits", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "enumerate", "expand", "f", "features", "flatten", "float", "float32", "floor", "for", "forward", "freeze_backbone", "functional", "gather", "gen_encoder_output_proposals", "get_encoder", "get_proposal_pos_embed", "get_valid_ratio", "grid", "grid_x", "grid_y", "height", "hidden_states", "if", "ij", "in", "in_channels", "indexing", "inf", "init_reference_points", "input_proj", "input_proj_list", "inputs_embeds", "int64", "intermediate_channel_sizes", "intermediate_hidden_states", "intermediate_reference_points", "interpolate", "is", "is_level_ordered", "isinstance", "keep_inds", "keep_inds_mask", "keep_inds_topk", "keepdim", "kernel_size", "last_hidden_state", "len", "level", "level_embed", "level_ids", "level_start_index", "linspace", "log", "long", "lvl", "lvl_mask", "lvl_pos_embed", "lvl_pos_embed_flatten", "m", "mask", "mask_flatten", "mask_flatten_", "masked_fill", "masks", "math", "meshgrid", "model", "naive", "name", "named_parameters", "new_ones", "new_zeros", "nms", "nn", "nonzero", "not", "num_backbone_outs", "num_channels", "num_feature_levels", "num_pos_feats", "num_queries", "num_to_add", "object_query", "object_query_embedding", "ones", "ones_like", "output_attentions", "output_hidden_states", "output_proposals", "output_proposals_valid", "output_type", "pad_inds", "padding", "padding_mask", "param", "pi", "pix_trans", "pix_trans_norm", "pixel_mask", "pixel_values", "pos", "pos_embed", "pos_l", "pos_trans", "pos_trans_norm", "pos_trans_out", "position_embedding", "position_embeddings", "position_embeddings_list", "post_init", "post_nms_inds", "pre_nms_inds", "pre_nms_topk", "print", "prod", "prop_boxes_b", "prop_logits_b", "proposal", "proposal_boxes", "proposal_logit", "proposals", "provided", "q_per_l", "query_embed", "query_embeds", "query_position_embeddings", "r", "raise", "range", "reference_points", "repeat", "replace_return_docstrings", "requires_backends", "requires_grad_", "return", "return_dict", "rounding_mode", "running", "scale", "self", "shape", "sigmoid", "sin", "size", "source", "source_flatten", "sources", "spatial_shapes", "split", "stack", "stride", "sum", "super", "target", "temperature", "to", "topk", "topk_coords_logits", "topk_feats", "topk_proposals", "torch", "torchvision", "transpose", "tuple", "tuple_outputs", "two_stage", "two_stage_num_proposals", "unfreeze_backbone", "unsqueeze", "use_return_dict", "valid_height", "valid_ratio", "valid_ratio_height", "valid_ratio_width", "valid_ratios", "valid_width", "value", "view", "was", "weight", "width", "width_height"], "deprecated/deta/modeling_deta.py:DetaForObjectDetection": ["FloatTensor", "Linear", "LongTensor", "ModelConfig", "ModelForObjectDetection", "ModelHungarianMatcher", "ModelLoss", "ModelMLPPredictionHead", "ModelModel", "ModelObjectDetectionOutput", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "ModuleList", "None", "Optional", "Union", "ValueError", "_", "_CONFIG_FOR_DOC", "__init__", "_enc", "_get_clones", "_no_split_modules", "_set_aux_loss", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "anchors", "append", "assign_first_stage", "assign_second_stage", "aux_loss", "aux_weight_dict", "auxiliary_loss", "auxiliary_outputs", "bbox_cost", "bbox_embed", "bbox_loss_coefficient", "be", "bias", "bias_value", "bool", "box_embed", "boxes", "but", "cardinality", "class", "class_cost", "class_embed", "config", "config_class", "constant_", "criterion", "cross_attentions", "d", "d_model", "data", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_hidden_states", "decoder_inputs_embeds", "decoder_layers", "def", "delta_bbox", "device", "dict", "dict_outputs", "dim", "elif", "else", "enc_outputs", "enc_outputs_class", "enc_outputs_coord", "enc_outputs_coord_logits", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "f", "fill_", "focal_alpha", "for", "forward", "giou_cost", "giou_loss_coefficient", "got", "hidden_dim", "hidden_states", "i", "if", "in", "init", "init_reference", "init_reference_points", "input_dim", "inputs_embeds", "inter_references", "intermediate_hidden_states", "intermediate_reference_points", "inverse_sigmoid", "is", "items", "jit", "k", "labels", "last_hidden_state", "layers", "level", "list", "log", "logits", "loss", "loss_bbox", "loss_ce", "loss_dict", "loss_giou", "losses", "matcher", "math", "model", "nn", "not", "num_classes", "num_labels", "num_layers", "num_pred", "num_queries", "or", "output", "output_attentions", "output_dim", "output_hidden_states", "output_proposals", "output_type", "outputs", "outputs_class", "outputs_classes", "outputs_coord", "outputs_coord_logits", "outputs_coords", "outputs_loss", "pixel_mask", "pixel_values", "post_init", "pred_boxes", "prior_prob", "r", "raise", "range", "reference", "replace_return_docstrings", "return", "return_dict", "self", "shape", "should", "sigmoid", "stack", "sum", "super", "to", "torch", "transpose", "tuple", "tuple_outputs", "two_stage", "unused", "update", "use_return_dict", "v", "weight", "weight_dict", "with_box_refine", "zip"], "deprecated/deta/modeling_deta.py:dice_loss": ["Model_loss", "def", "denominator", "flatten", "inputs", "loss", "num_boxes", "numerator", "return", "sigmoid", "sum", "targets"], "deprecated/deta/modeling_deta.py:sigmoid_focal_loss": ["Model", "Model_focal_loss", "alpha", "alpha_t", "binary_cross_entropy_with_logits", "ce_loss", "def", "float", "functional", "gamma", "if", "inputs", "loss", "mean", "nn", "none", "num_boxes", "p_t", "prob", "reduction", "return", "sum", "targets"], "deprecated/deta/modeling_deta.py:DetaLoss": ["False", "J", "KeyError", "Loss", "ModelLoss", "ModelStage1Assigner", "ModelStage2Assigner", "Module", "No", "PartialState", "ValueError", "_", "__init__", "_enc", "_get_source_permutation_idx", "_get_target_permutation_idx", "_shared_state", "alpha", "argmax", "as_tensor", "assign_first_stage", "assign_second_stage", "auxiliary_outputs", "batch_idx", "bin_targets", "boxes", "bt", "card_err", "card_pred", "cardinality", "cardinality_error", "cat", "center_to_corners_format", "clamp", "class", "class_labels", "copy", "deepcopy", "def", "device", "diag", "dim", "dtype", "else", "enc_outputs", "enumerate", "f", "float", "focal_alpha", "for", "forward", "found", "full", "full_like", "functional", "gamma", "generalized_box_iou", "get_loss", "i", "idx", "if", "in", "indices", "int64", "is_accelerate_available", "item", "items", "iter", "k", "l1_loss", "l_dict", "labels", "layout", "len", "logits", "loss", "loss_bbox", "loss_boxes", "loss_cardinality", "loss_ce", "loss_giou", "loss_labels", "loss_map", "losses", "matcher", "min", "next", "nn", "no_grad", "none", "not", "num_boxes", "num_classes", "num_processes", "num_queries", "outputs", "outputs_without_aux", "pred_boxes", "predicted", "raise", "reduce", "reduction", "return", "scatter_", "self", "shape", "sigmoid_focal_loss", "source", "source_boxes", "source_idx", "source_logits", "stg1_assigner", "stg2_assigner", "sum", "super", "supported", "t", "target", "target_boxes", "target_classes", "target_classes_o", "target_classes_onehot", "target_idx", "target_lengths", "targets", "the", "torch", "unsqueeze", "update", "v", "values", "were", "world_size", "zeros", "zeros_like", "zip"], "deprecated/deta/modeling_deta.py:DetaMLPPredictionHead": ["Linear", "ModelMLPPredictionHead", "Module", "ModuleList", "__init__", "class", "def", "else", "enumerate", "for", "forward", "functional", "h", "hidden_dim", "i", "if", "in", "input_dim", "k", "layer", "layers", "n", "nn", "num_layers", "output_dim", "relu", "return", "self", "super", "x", "zip"], "deprecated/deta/modeling_deta.py:DetaHungarianMatcher": ["All", "Matcher", "ModelHungarianMatcher", "Module", "ValueError", "__init__", "alpha", "and", "as_tensor", "batch_size", "bbox_cost", "be", "boxes", "c", "can", "cat", "cdist", "center_to_corners_format", "class", "class_cost", "class_labels", "cost_matrix", "costs", "cpu", "def", "dtype", "enumerate", "flatten", "float", "for", "forward", "gamma", "generalized_box_iou", "giou_cost", "i", "if", "in", "indices", "int64", "j", "len", "linear_sum_assignment", "log", "logits", "neg_cost_class", "nn", "no_grad", "num_queries", "of", "out_bbox", "out_prob", "outputs", "p", "pos_cost_class", "pred_boxes", "raise", "requires_backends", "return", "scipy", "self", "shape", "sigmoid", "sizes", "split", "super", "t", "target_bbox", "target_ids", "targets", "the", "torch", "v", "view"], "deprecated/deta/modeling_deta.py:_upcast": ["Tensor", "_upcast", "def", "dtype", "else", "float", "float32", "float64", "if", "in", "int", "int32", "int64", "is_floating_point", "return", "t", "torch"], "deprecated/deta/modeling_deta.py:box_area": ["Model_area", "Modeles", "Tensor", "_upcast", "def", "return"], "deprecated/deta/modeling_deta.py:box_iou": ["Model_area", "Model_iou", "Modeles1", "Modeles2", "None", "area1", "area2", "clamp", "def", "inter", "iou", "left_top", "max", "min", "return", "right_bottom", "torch", "union", "width_height"], "deprecated/deta/modeling_deta.py:generalized_box_iou": ["Model_box_iou", "None", "ValueError", "all", "area", "be", "bottom_right", "box_iou", "boxes1", "boxes2", "but", "clamp", "corner", "def", "f", "format", "got", "if", "in", "iou", "max", "min", "must", "not", "raise", "return", "top_left", "torch", "union", "width_height", "x0", "x1", "y0", "y1"], "deprecated/deta/modeling_deta.py:nonzero_tuple": ["Model", "Model_tuple", "True", "as_tuple", "def", "dim", "else", "if", "is_scripting", "jit", "return", "torch", "unbind", "unsqueeze", "x"], "deprecated/deta/modeling_deta.py:DetaMatcher": ["All", "False", "ModelMatcher", "None", "Number", "Thresholds", "ValueError", "_", "__call__", "__init__", "all", "allow_low_quality_matches", "append", "assert", "be", "bool", "class", "def", "default_match_labels", "default_matches", "dim", "dtype", "either", "equal", "float", "for", "high", "highest_quality_foreach_gt", "if", "in", "inf", "insert", "int", "int64", "int8", "l", "labels", "len", "list", "low", "low_high", "match_labels", "match_quality_matrix", "matched_vals", "matches", "max", "new_full", "nonzero_tuple", "not", "number", "numel", "of", "or", "positive", "pred_inds_with_highest_quality", "raise", "return", "self", "set_low_quality_matches_", "should", "size", "sorted", "thresholds", "to", "torch", "zip"], "deprecated/deta/modeling_deta.py:subsample_labels": ["Model_labels", "Tensor", "bg_label", "def", "device", "float", "int", "labels", "min", "neg_idx", "negative", "nonzero_tuple", "num_neg", "num_pos", "num_samples", "numel", "perm1", "perm2", "pos_idx", "positive", "positive_fraction", "randperm", "return", "torch"], "deprecated/deta/modeling_deta.py:sample_topk_per_gt": ["Model_topk_per_gt", "None", "True", "c", "cat", "counts", "def", "dim", "for", "gt", "gt_inds", "gt_inds2", "gt_inds3", "if", "in", "iou", "k", "len", "pr", "pr_inds", "pr_inds2", "pr_inds3", "repeat", "return", "return_counts", "scores", "topk", "torch", "unique", "zip"], "deprecated/deta/modeling_deta.py:DetaStage2Assigner": ["False", "ModelMatcher", "ModelStage2Assigner", "Modelch", "Module", "Tensor", "True", "_", "__init__", "_sample_proposals", "allow_low_quality_matches", "append", "b", "batch_size_per_image", "bg_label", "box_iou", "boxes", "bs", "cat", "center_to_corners_format", "class", "class_labels", "def", "dim", "else", "for", "forward", "gt_classes", "gt_inds", "has_gt", "if", "in", "indices", "init_reference", "iou", "ious", "k", "labels", "len", "matched_idxs", "matched_labels", "max_k", "nn", "num_queries", "numel", "outputs", "pos_gt_inds", "pos_pr_inds", "positive_fraction", "postprocess_indices", "pr_inds", "proposal_matcher", "range", "return", "return_cost_matrix", "sample_topk_per_gt", "sampled_bg_idxs", "sampled_fg_idxs", "sampled_gt_classes", "sampled_idxs", "self", "subsample_labels", "super", "targets", "thresholds", "torch", "zeros_like"], "deprecated/deta/modeling_deta.py:DetaStage1Assigner": ["ModelMatcher", "ModelStage1Assigner", "Module", "True", "_", "__init__", "_subsample_labels", "all_pr_inds", "allow_low_quality_matches", "anchor_matcher", "anchors", "append", "arange", "b", "batch_size_per_image", "box_iou", "boxes", "bs", "center_to_corners_format", "class", "continue", "def", "device", "dtype", "fill_", "for", "forward", "gt_inds", "if", "in", "indices", "iou", "k", "label", "labels", "len", "long", "matched_idxs", "matched_labels", "max_k", "neg_idx", "nn", "outputs", "pos_gt_inds", "pos_idx", "pos_pr_inds", "positive_fraction", "postprocess_indices", "pr_inds", "range", "return", "sample_topk_per_gt", "scatter_", "self", "subsample_labels", "super", "t_high", "t_low", "targets", "tensor", "thresholds", "to", "torch"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:softmax": ["False", "Model", "def", "dim", "dtype", "else", "float", "float32", "functional", "hidden_state", "if", "nn", "onnx_trace", "return", "torch"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:ngram_attention_bias": ["False", "Model", "Model_attention_bias", "cat", "clone", "def", "detach", "device", "dim", "dtype", "fill_diagonal_", "finfo", "for", "in", "left_block", "min", "ones", "range", "return", "right_block", "sequence_length", "stream_idx", "torch", "triu_", "wrap"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:compute_relative_buckets": ["False", "Model_relative_buckets", "abs", "def", "else", "float", "if", "int", "inv_relative_positions", "is_bidirectional", "is_small", "log", "lt", "math", "max", "max_distance", "max_exact", "min", "num_buckets", "ones_like", "rel_positions_bucket", "relative_positions", "return", "torch", "val_if_large", "where", "zeros_like"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:compute_all_stream_relative_buckets": ["False", "Model_all_stream_relative_buckets", "Model_relative_buckets", "cat", "def", "dim", "is_bidirectional", "main_relative_position_buckets", "main_stream_relative_positions", "max_distance", "num_buckets", "position_ids", "predict_relative_position_buckets", "predicting_stream_relative_positions", "repeat", "return", "size", "torch", "unsqueeze"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetSeq2SeqLMOutput": ["Cache", "FloatTensor", "FutureWarning", "Model", "ModelOutput", "None", "Optional", "Please", "and", "be", "class", "cross_attentions", "decoder_attentions", "decoder_cross_attentions", "decoder_hidden_states", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "def", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "instead", "is", "logits", "logits_ngram", "loss", "past_key_values", "property", "removed", "return", "self", "soon", "torch", "tuple", "use", "warn", "warnings", "will"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetSeq2SeqModelOutput": ["Cache", "FloatTensor", "FutureWarning", "Model", "ModelOutput", "None", "Optional", "Please", "and", "be", "class", "cross_attentions", "decoder_attentions", "decoder_cross_attentions", "decoder_hidden_states", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "def", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "instead", "is", "last_hidden_state", "last_hidden_state_ngram", "past_key_values", "property", "removed", "return", "self", "soon", "torch", "tuple", "use", "warn", "warnings", "will"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetDecoderModelOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "hidden_states_ngram", "last_hidden_state", "last_hidden_state_ngram", "ngram_attentions", "past_key_values", "torch", "tuple"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetDecoderLMOutput": ["Cache", "FloatTensor", "Model", "ModelOutput", "None", "Optional", "attentions", "class", "cross_attentions", "hidden_states", "hidden_states_ngram", "logits", "logits_ngram", "loss", "ngram_attentions", "past_key_values", "torch", "tuple"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetPreTrainedModel": ["Embedding", "In", "Linear", "Model", "None", "PreTrainedModel", "See", "True", "Verify", "XLMProphetNet", "XLMProphetNetConfig", "_init_weights", "_shift_right", "all", "assert", "base_model_prefix", "be", "bias", "class", "clone", "config", "data", "decoder_start_token_id", "def", "defined", "docs", "elif", "for", "has", "if", "information", "init_std", "input_ids", "is", "isinstance", "it", "item", "masked_fill_", "mean", "model", "module", "more", "new_zeros", "nn", "normal_", "not", "only", "pad_token_id", "padding_idx", "positive", "prophetnet", "return", "self", "set", "shape", "shifted_input_ids", "std", "supports_gradient_checkpointing", "that", "the", "to", "torch", "usually", "values", "weight", "zero_"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetPositionalEmbeddings": ["Embedding", "If", "Model", "None", "XLMProphetNetConfig", "__init__", "_forward", "assert", "attention_mask", "be", "clamp", "class", "computed", "config", "cumsum", "def", "device", "dim", "dtype", "else", "forward", "get_seq_length", "hidden_size", "if", "inputs_shape", "int", "is", "long", "max_length", "max_position_embeddings", "nn", "not", "num_input_ids", "ones", "or", "pad_token_id", "padding_idx", "past_key_values", "position_ids", "pre", "prev_num_input_ids", "return", "self", "set", "should", "super", "then", "torch", "type_as"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetAttention": ["Attention", "Cache", "False", "Head", "Linear", "Model", "Module", "None", "Optional", "Size", "Tensor", "ValueError", "XLMProphetNetConfig", "__init__", "_shape", "a", "and", "assert", "attention_dropout", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "batch_size", "be", "bool", "bsij", "bsik", "bsjk", "bsz", "but", "by", "class", "config", "contiguous", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "einsum", "elif", "else", "expected_shape", "f", "for", "forward", "functional", "have", "head_dim", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "key_proj", "key_states", "key_value_states", "layer", "layer_head_mask", "list", "mask", "must", "new_name", "nn", "not", "num_attn_heads", "num_decoder_attention_heads", "num_encoder_attention_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "query_proj", "query_states", "raise", "reshape", "return", "self", "seq_len", "shape", "should", "single", "size", "softmax", "src_len", "states", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "value_proj", "value_states", "version", "view", "weights"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetFeedForward": ["ACT2FN", "Linear", "Model", "Module", "XLMProphetNetConfig", "__init__", "activation_dropout", "activation_fn", "activation_function", "class", "config", "def", "dropout", "ffn_dim", "forward", "functional", "hidden_size", "hidden_states", "int", "intermediate", "nn", "output", "p", "return", "self", "super", "training"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetNgramSelfAttention": ["Cache", "False", "Head", "Linear", "Model", "Module", "None", "Optional", "They", "True", "XLMProphetNetConfig", "__init__", "_shape", "a", "arange", "are", "assert", "attention_dropout", "attention_mask", "attn_output", "attn_weights", "batch_size", "be", "bncs", "bnhsc", "bnhtc", "bnhts", "bntc", "bnts", "but", "by", "cat", "chunk", "class", "compute_relative_buckets", "config", "contiguous", "def", "deprecate_kwarg", "device", "dim", "divisible", "dropout", "dtype", "einsum", "extended_predict_attention_mask", "f", "for", "format", "forward", "functional", "gather", "get_main_relative_pos_embeddings", "get_predict_relative_pos_embeddings", "head_dim", "hidden_size", "hidden_states", "hidden_states_list", "hidden_states_predict_list", "if", "in", "incorrect", "index", "is", "key", "key_proj", "key_sequence_length", "key_states", "key_states_list", "layer", "layer_head_mask", "list", "long", "main_attn_output", "main_attn_probs", "main_attn_weights", "main_hidden_states", "main_key_states", "main_query_states", "main_relative_pos_embeddings", "main_relative_position_buckets", "main_value_states", "mask", "must", "new_name", "ngram", "ngram_sequence_length", "nn", "not", "num_attn_heads", "num_buckets", "num_decoder_attention_heads", "of", "onnx_trace", "out_proj", "p", "past_key_value", "past_key_values", "permute", "position_ids", "predict_attn_output", "predict_attn_probs", "predict_attn_weights", "predict_hidden_states", "predict_key_states", "predict_key_states_list", "predict_query_states", "predict_query_states_list", "predict_relative_pos_embeddings", "predict_relative_position_buckets", "predict_value_states", "predict_value_states_list", "prepare_for_onnx_export_", "prev_main_key_states", "prev_main_value_states", "proj_shape", "query_proj", "query_states", "query_states_list", "rel_pos_embeddings", "relative_max_distance", "relative_pos_embeddings", "relative_positions", "repeat", "reshape", "return", "self", "seq_len", "sequence_length", "shape", "should", "single", "size", "softmax", "src_len", "stack", "super", "tensor", "tgt_len", "the", "to", "torch", "training", "transpose", "type_as", "unsqueeze", "v_p", "value_proj", "value_states", "value_states_list", "version", "view"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetEncoderLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "Model", "XLMProphetNetAttention", "XLMProphetNetConfig", "XLMProphetNetFeedForward", "_", "__init__", "attention_mask", "attention_output", "attn_weights", "bool", "class", "config", "def", "encoder_ffn_dim", "feed_forward", "feed_forward_layer_norm", "feed_forward_output", "forward", "hidden_size", "hidden_states", "if", "layer_head_mask", "num_encoder_attention_heads", "output_attentions", "outputs", "return", "self", "self_attn", "self_attn_layer_norm", "super"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetDecoderLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "Model", "None", "True", "XLMProphetNetAttention", "XLMProphetNetConfig", "XLMProphetNetFeedForward", "XLMProphetNetNgramSelfAttention", "__init__", "add_cross_attention", "attention_mask", "attention_output", "bool", "class", "config", "cross_attn", "cross_attn_layer_head_mask", "cross_attn_layer_norm", "cross_attn_past_key_value", "cross_attn_present_key_value", "cross_attn_weights", "decoder_ffn_dim", "def", "deprecate_kwarg", "else", "encoder_attn_mask", "encoder_hidden_states", "extended_predict_attention_mask", "feed_forward", "feed_forward_layer_norm", "feed_forward_output", "forward", "hidden_size", "hidden_states", "if", "is", "key_value_states", "layer_head_mask", "main_relative_position_buckets", "new_name", "ngram_attention_output", "not", "num_decoder_attention_heads", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "predict_relative_position_buckets", "present_key_value", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_past_key_value", "self_attn_weights", "self_attn_weights_ngram", "super", "use_cache", "version"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetEncoder": ["BaseModelOutput", "Either", "Embedding", "False", "LayerNorm", "Make", "Model", "ModelLayer", "ModuleList", "None", "Optional", "Tensor", "The", "Union", "ValueError", "XLMProphetNetConfig", "XLMProphetNetPositionalEmbeddings", "XLMProphetNetPreTrainedModel", "XLM_PROPHETNET_STANDALONE_INPUTS_DOCSTRING", "_", "_CONFIG_FOR_DOC", "__init__", "add_start_docstrings_to_model_forward", "all_attentions", "and", "assert", "attention_mask", "attentions", "be", "bool", "but", "class", "config", "config_class", "def", "device", "dropout", "dtype", "elif", "else", "embeddings_layer_norm", "encoder_hidden_states", "encoder_layer", "enumerate", "extended_attention_mask", "f", "finfo", "for", "forward", "functional", "get_input_embeddings", "gradient_checkpointing", "has", "head_mask", "hidden_size", "hidden_states", "idx", "if", "in", "input_ids", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layers", "len", "min", "nn", "not", "num_encoder_attention_heads", "num_encoder_layers", "only", "or", "output_attentions", "output_hidden_states", "output_type", "p", "pad_token_id", "padding_idx", "pass", "passed", "position_embeddings", "position_ids", "post_init", "r", "raise", "range", "repeat", "replace_return_docstrings", "return", "return_dict", "self", "set_input_embeddings", "shape", "should", "size", "specified", "super", "sure", "to", "torch", "training", "tuple", "use_return_dict", "v", "value", "vocab_size", "word_embeddings"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetDecoder": ["At", "Cache", "Either", "Embedding", "False", "LayerNorm", "Make", "Model", "ModelLayer", "ModelModelOutput", "ModuleList", "None", "Optional", "Setting", "Tensor", "The", "True", "Union", "ValueError", "XLMProphetNetConfig", "XLMProphetNetPositionalEmbeddings", "XLMProphetNetPreTrainedModel", "XLM_PROPHETNET_STANDALONE_INPUTS_DOCSTRING", "_", "_CONFIG_FOR_DOC", "__init__", "_forward", "add_cross_attention", "add_start_docstrings_to_model_forward", "all_cross_attns", "all_main_stream_attns", "all_main_stream_hidden_states", "all_ngram_stream_attns", "all_ngram_stream_hidden_states", "and", "arange", "assert", "attention_mask", "attentions", "attn_mask", "batch_size", "be", "bool", "but", "cat", "causal_mask", "checkpointing", "class", "compute_all_stream_relative_buckets", "compute_buffered_relative_buckets", "config", "config_class", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "def", "device", "dim", "dropout", "dtype", "elif", "else", "embeddings_layer_norm", "encoder_attention_mask", "encoder_attn_mask", "encoder_hidden_states", "enumerate", "expand", "extended_attention_mask", "extended_causal_mask", "extended_encoder_attention_mask", "extended_predict_attention_mask", "extended_predict_causal_mask", "f", "finfo", "for", "forward", "full", "functional", "get_input_embeddings", "gradient", "gradient_checkpointing", "has", "head_mask", "hidden_size", "hidden_states", "hidden_states_ngram", "idx", "if", "in", "incompatible", "input_ids", "inputs_embeds", "is", "it", "last_hidden_state", "last_hidden_state_ngram", "layer_head_mask", "layer_outputs", "layers", "len", "length", "logger", "main_relative_buckets", "main_relative_position_buckets", "main_stream_pos_embed", "mask_name", "max_position_embeddings", "max_target_positions", "min", "moment", "ngram", "ngram_attention_bias", "ngram_attentions", "ngram_embeddings", "ngram_hidden_states", "nn", "not", "num_buckets", "num_decoder_attention_heads", "num_decoder_layers", "of", "only", "or", "output_attentions", "output_hidden_states", "output_type", "p", "pad_token_id", "padding_idx", "pass", "passed", "past_key_values", "position_embeddings", "position_ids", "post_init", "predict_causal_mask", "predict_relative_buckets", "predict_relative_position_buckets", "predicting_stream_pos_embed", "prepare_attention_mask", "prepare_predict_attention_mask", "present_key_values", "r", "raise", "range", "relative_max_distance", "repeat", "replace_return_docstrings", "return", "return_dict", "self", "seq_length", "sequence_length", "set_input_embeddings", "shape", "should", "size", "specified", "super", "supported", "sure", "the", "to", "torch", "training", "triu", "tuple", "use_cache", "use_return_dict", "v", "value", "vocab_size", "warning_once", "weight", "with", "word_embeddings", "zeros_like", "zip"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetModel": ["BoolTensor", "Cache", "Embedding", "False", "Model", "None", "Optional", "Tensor", "True", "Union", "XLMProphetNetConfig", "XLMProphetNetDecoder", "XLMProphetNetEncoder", "XLMProphetNetPreTrainedModel", "XLMProphetNetSeq2SeqModelOutput", "XLM_PROPHETNET_INPUTS_DOCSTRING", "_CONFIG_FOR_DOC", "__init__", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "bool", "class", "config", "config_class", "copy", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_config", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "decoder_outputs", "deepcopy", "def", "else", "encoder", "encoder_attention_mask", "encoder_attentions", "encoder_config", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "forward", "get_encoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "hidden_states_ngram", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "last_hidden_state", "last_hidden_state_ngram", "ngram_attentions", "nn", "not", "output_attentions", "output_hidden_states", "output_type", "pad_token_id", "padding_idx", "past_key_values", "post_init", "r", "replace_return_docstrings", "return", "return_dict", "self", "set_input_embeddings", "super", "tie_word_embeddings", "torch", "tuple", "use_cache", "use_return_dict", "value", "vocab_size", "weight", "word_embeddings"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetForConditionalGeneration": ["BoolTensor", "Cache", "False", "Linear", "Model", "None", "Optional", "Tensor", "True", "Union", "XLMProphetNetConfig", "XLMProphetNetModel", "XLMProphetNetPreTrainedModel", "XLMProphetNetSeq2SeqLMOutput", "XLM_PROPHETNET_INPUTS_DOCSTRING", "_CONFIG_FOR_DOC", "__init__", "_compute_loss", "_reorder_cache", "_shift_right", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "all_logits", "and", "assert", "attention_mask", "batch_size", "be", "beam_idx", "bias", "bool", "break", "class", "config", "config_class", "contiguous", "cross_attentions", "cross_attn_head_mask", "decoder", "decoder_attention_mask", "decoder_attentions", "decoder_head_mask", "decoder_hidden_states", "decoder_input_ids", "decoder_inputs_embeds", "decoder_ngram_attentions", "decoder_ngram_hidden_states", "def", "device", "dim", "disable_ngram_loss", "dtype", "else", "encoder", "encoder_attentions", "encoder_hidden_states", "encoder_last_hidden_state", "encoder_outputs", "eps", "eps_i", "expend_targets", "fill_", "float32", "for", "forward", "functional", "generation", "get_decoder", "get_encoder", "get_input_embeddings", "have", "head_mask", "hidden_size", "i", "if", "ignore_index", "in", "index_select", "input_ids", "inputs_embeds", "is", "is_contiguous", "keepdim", "kwargs", "labels", "layer_past", "lm_head", "log_softmax", "logits", "logits_ngram", "loss", "lprobs", "mean", "ne", "new_zeros", "ngram", "nll_loss", "nn", "non_masked_tokens", "not", "output_attentions", "output_hidden_states", "output_type", "outputs", "pad_token_id", "padding_idx", "passed", "past_key_values", "past_state", "post_init", "predict_logits", "predicting_streams", "prepare_decoder_input_ids_from_labels", "prepare_inputs_for_generation", "prophetnet", "r", "range", "reduction", "reordered_past", "replace_return_docstrings", "return", "return_dict", "self", "sequence_length", "shape", "size", "smooth_loss", "staticmethod", "sum", "super", "tie_word_embeddings", "to", "torch", "transpose", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "weight", "word_embeddings"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetForCausalLM": ["Cache", "False", "Linear", "Model", "None", "Optional", "Tensor", "True", "Union", "XLMProphetNetConfig", "XLMProphetNetDecoderLMOutput", "XLMProphetNetDecoderWrapper", "XLMProphetNetPreTrainedModel", "XLM_PROPHETNET_STANDALONE_INPUTS_DOCSTRING", "_CONFIG_FOR_DOC", "__init__", "_compute_loss", "_reorder_cache", "_tie_or_clone_weights", "_tie_weights", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "all_logits", "and", "attention_mask", "attentions", "batch_size", "beam_idx", "bias", "bool", "break", "class", "config", "config_class", "contiguous", "copy", "cross_attentions", "cross_attn_head_mask", "decoder", "deepcopy", "def", "device", "dim", "disable_ngram_loss", "dtype", "else", "encoder_attention_mask", "encoder_hidden_states", "eps", "eps_i", "expend_targets", "fill_", "float32", "for", "forward", "functional", "get_decoder", "get_input_embeddings", "head_mask", "hidden_size", "hidden_states", "hidden_states_ngram", "i", "if", "ignore_index", "in", "index_select", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "keepdim", "kwargs", "labels", "layer_past", "lm_head", "log_softmax", "logits", "logits_ngram", "loss", "lprobs", "mean", "ne", "new_ones", "new_zeros", "ngram", "ngram_attentions", "nll_loss", "nn", "non_masked_tokens", "not", "output_attentions", "output_hidden_states", "output_type", "outputs", "pad_token_id", "padding_idx", "past_key_values", "past_state", "post_init", "predict_logits", "predicting_streams", "prepare_inputs_for_generation", "prophetnet", "r", "range", "reduction", "reordered_past", "replace_return_docstrings", "return", "return_dict", "self", "sequence_length", "set_decoder", "set_input_embeddings", "shape", "size", "smooth_loss", "staticmethod", "sum", "super", "tie_word_embeddings", "to", "torch", "transpose", "tuple", "use_cache", "use_return_dict", "v", "value", "view", "vocab_size", "weight", "word_embeddings"], "deprecated/xlm_prophetnet/modeling_xlm_prophetnet.py:XLMProphetNetDecoderWrapper": ["Embedding", "Model", "XLMProphetNetConfig", "XLMProphetNetDecoder", "XLMProphetNetPreTrainedModel", "__init__", "_tie_or_clone_weights", "_tie_weights", "args", "class", "config", "decoder", "def", "forward", "get_input_embeddings", "hidden_size", "kwargs", "nn", "pad_token_id", "padding_idx", "post_init", "return", "self", "super", "vocab_size", "word_embeddings"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridEmbeddings": ["BoolTensor", "Dropout", "False", "Module", "None", "Optional", "Parameter", "Tensor", "ViTHybridConfig", "ViTHybridEmbeddings", "ViTHybridPatchEmbeddings", "__init__", "align_corners", "and", "batch_size", "bicubic", "bool", "bool_masked_pos", "cat", "class", "class_pos_embed", "cls_token", "cls_tokens", "config", "def", "dim", "dropout", "else", "embeddings", "expand", "forward", "functional", "height", "hidden_dropout_prob", "hidden_size", "if", "int", "interpolate", "interpolate_pos_encoding", "is", "is_tracing", "jit", "mask", "mask_token", "mask_tokens", "mode", "new_height", "new_width", "nn", "not", "num_channels", "num_patches", "num_positions", "patch_embeddings", "patch_pos_embed", "patch_size", "permute", "pixel_values", "position_embeddings", "randn", "reshape", "return", "self", "seq_length", "shape", "size", "sqrt_num_positions", "super", "torch", "torch_int", "type_as", "unsqueeze", "use_mask_token", "view", "width", "zeros"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridPatchEmbeddings": ["Backbone", "Conv2d", "False", "Input", "Iterable", "Make", "Module", "None", "Tensor", "ValueError", "ViTHybridPatchEmbeddings", "_", "__init__", "abc", "backbone", "backbone_featmap_shape", "bit", "bool", "channel", "channels", "class", "collections", "config", "configuration", "def", "dimension", "doesn", "else", "embeddings", "f", "feature_dim", "feature_map", "feature_maps", "feature_size", "features", "flatten", "forward", "grid_size", "height", "hidden_size", "if", "image", "image_size", "in", "interpolate_pos_encoding", "is", "isinstance", "kernel_size", "load_backbone", "match", "model", "model_type", "nn", "not", "num_channels", "num_patches", "of", "one", "or", "patch_size", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "supported", "sure", "t", "that", "the", "torch", "transpose", "type", "values", "width", "with"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridSelfAttention": ["Dropout", "False", "Linear", "Module", "None", "Optional", "Tensor", "The", "Union", "ValueError", "ViTHybridConfig", "ViTHybridSelfAttention", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "embedding_size", "f", "forward", "functional", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "key", "key_layer", "math", "matmul", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "self", "size", "softmax", "sqrt", "super", "the", "torch", "transpose", "transpose_for_scores", "tuple", "value", "value_layer", "view", "x"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridSdpaSelfAttention": ["False", "None", "Optional", "Tensor", "Union", "ViTHybridConfig", "ViTHybridSdpaSelfAttention", "ViTHybridSelfAttention", "__init__", "all_head_size", "attention_probs_dropout_prob", "bool", "class", "config", "context_layer", "contiguous", "def", "else", "forward", "functional", "head_mask", "hidden_states", "if", "is_causal", "key", "key_layer", "mixed_query_layer", "new_context_layer_shape", "nn", "output_attentions", "permute", "query", "query_layer", "return", "scale", "scaled_dot_product_attention", "self", "size", "super", "torch", "training", "transpose_for_scores", "tuple", "value", "value_layer", "view"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridSelfOutput": ["Dropout", "Linear", "Module", "None", "Tensor", "ViTHybridConfig", "ViTHybridSelfOutput", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridAttention": ["False", "Module", "None", "Optional", "Tensor", "Union", "ViTHybridAttention", "ViTHybridConfig", "ViTHybridSelfAttention", "ViTHybridSelfOutput", "__init__", "all_head_size", "attention", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "int", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridSdpaAttention": ["None", "ViTHybridAttention", "ViTHybridConfig", "ViTHybridSdpaAttention", "ViTHybridSdpaSelfAttention", "__init__", "attention", "class", "config", "def", "self", "super"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridIntermediate": ["ACT2FN", "Linear", "Module", "None", "Tensor", "ViTHybridConfig", "ViTHybridIntermediate", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridOutput": ["Dropout", "Linear", "Module", "None", "Tensor", "ViTHybridConfig", "ViTHybridOutput", "__init__", "class", "config", "def", "dense", "dropout", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "nn", "return", "self", "super", "torch"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridLayer": ["False", "GradientCheckpointingLayer", "LayerNorm", "None", "Optional", "Tensor", "Union", "VIT_HYBRID_ATTENTION_CLASSES", "ViTHybridConfig", "ViTHybridIntermediate", "ViTHybridLayer", "ViTHybridOutput", "__init__", "_attn_implementation", "attention", "attention_output", "bool", "chunk_size_feed_forward", "class", "config", "def", "device", "eps", "forward", "head_mask", "hidden_size", "hidden_states", "intermediate", "layer_norm_eps", "layer_output", "layernorm_after", "layernorm_before", "nn", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super", "to", "torch", "tuple"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridEncoder": ["BaseModelOutput", "False", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "ViTHybridConfig", "ViTHybridEncoder", "ViTHybridLayer", "_", "__init__", "all_hidden_states", "all_self_attentions", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "None", "PreTrainedModel", "True", "Union", "ViTHybridConfig", "ViTHybridEmbeddings", "ViTHybridLayer", "ViTHybridPreTrainedModel", "_init_weights", "_no_split_modules", "_supports_sdpa", "base_model_prefix", "bias", "class", "cls_token", "config", "data", "def", "dtype", "elif", "fill_", "float32", "if", "init", "initializer_range", "is", "isinstance", "main_input_name", "mask_token", "mean", "module", "nn", "not", "pixel_values", "position_embeddings", "self", "std", "supports_gradient_checkpointing", "to", "torch", "trunc_normal_", "vit", "weight", "zero_"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridModel": ["BaseModelOutputWithPooling", "BoolTensor", "False", "LayerNorm", "None", "Optional", "Tensor", "True", "Union", "VIT_INPUTS_DOCSTRING", "ValueError", "ViTHybridConfig", "ViTHybridEmbeddings", "ViTHybridEncoder", "ViTHybridModel", "ViTHybridPatchEmbeddings", "ViTHybridPooler", "ViTHybridPreTrainedModel", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_EXPECTED_OUTPUT_SHAPE", "__init__", "_prune_heads", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention", "attentions", "bool", "bool_masked_pos", "checkpoint", "class", "config", "config_class", "def", "dict", "dtype", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "expected_dtype", "expected_output", "for", "forward", "get_head_mask", "get_input_embeddings", "have", "head_mask", "head_outputs", "heads", "heads_to_prune", "hidden_size", "hidden_states", "if", "in", "int", "interpolate_pos_encoding", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "list", "modality", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "output_type", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "projection", "prune_heads", "r", "raise", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "tuple", "use_mask_token", "use_return_dict", "vision", "weight"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridPooler": ["Linear", "Module", "Tanh", "ViTHybridConfig", "ViTHybridPooler", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super"], "deprecated/vit_hybrid/modeling_vit_hybrid.py:ViTHybridForImageClassification": ["False", "Identity", "ImageClassifierOutput", "Linear", "None", "Optional", "Tensor", "Union", "VIT_INPUTS_DOCSTRING", "ViTHybridConfig", "ViTHybridForImageClassification", "ViTHybridModel", "ViTHybridPreTrainedModel", "_CONFIG_FOR_DOC", "_IMAGE_CLASS_CHECKPOINT", "_IMAGE_CLASS_EXPECTED_OUTPUT", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attentions", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "else", "expected_output", "forward", "head_mask", "hidden_size", "hidden_states", "if", "interpolate_pos_encoding", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict", "vit"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2SinusoidalPositionalEmbedding": ["False", "ModelText2SinusoidalPositionalEmbedding", "Module", "None", "Optional", "Parameter", "Tensor", "__init__", "arange", "bsz", "cat", "class", "cos", "create_position_ids_from_input_ids", "cumsum", "def", "detach", "detach_", "device", "dim", "dtype", "emb", "emb_weights", "embedding_dim", "exp", "float", "forward", "get_default_dtype", "get_embedding", "half_dim", "hasattr", "if", "incremental_indices", "index_select", "input_ids", "int", "int64", "is", "log", "long", "make_weights", "mask", "math", "max_pos", "ne", "nn", "no_grad", "not", "num_embeddings", "num_positions", "offset", "padding_idx", "past_key_values_length", "position_ids", "requires_grad", "return", "self", "seq_len", "sin", "size", "staticmethod", "super", "to", "torch", "type_as", "unsqueeze", "view", "weights", "zeros"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2Attention": ["Attention", "Cache", "False", "Head", "Linear", "ModelText2Attention", "ModelText2Config", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_shape", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cat", "class", "config", "contiguous", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "elif", "else", "embed_dim", "f", "float", "for", "forward", "functional", "got", "head_dim", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "layer", "layer_head_mask", "mask", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "shape", "should", "single", "size", "softmax", "src_len", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "version", "view", "weights"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2DecoderLayer": ["ACT2FN", "Cache", "False", "GradientCheckpointingLayer", "LayerNorm", "Linear", "ModelText2Attention", "ModelText2Config", "ModelText2DecoderLayer", "None", "Optional", "Tensor", "True", "__init__", "activation_dropout", "activation_fn", "activation_function", "attention_dropout", "attention_mask", "bool", "class", "config", "cross_attn_layer_head_mask", "cross_attn_past_key_value", "cross_attn_present_key_value", "cross_attn_weights", "d_model", "decoder_attention_heads", "decoder_ffn_dim", "def", "deprecate_kwarg", "dropout", "else", "embed_dim", "encoder_attention_mask", "encoder_attn", "encoder_attn_layer_norm", "encoder_hidden_states", "fc1", "fc2", "final_layer_norm", "forward", "functional", "hidden_states", "if", "is", "is_decoder", "key_value_states", "layer_head_mask", "new_name", "nn", "not", "num_heads", "output_attentions", "outputs", "p", "past_key_value", "past_key_values", "present_key_value", "residual", "return", "self", "self_attn", "self_attn_layer_norm", "self_attn_past_key_value", "self_attn_weights", "super", "torch", "training", "use_cache", "version"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2PreTrainedModel": ["Conv1d", "Embedding", "False", "Linear", "ModelText2Config", "ModelText2PreTrainedModel", "ModelText2SinusoidalPositionalEmbedding", "None", "Parameter", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "detach_", "elif", "get_embedding", "if", "init_std", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "requires_grad", "self", "shape", "std", "supports_gradient_checkpointing", "weight", "zero_"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2Decoder": ["BaseModelOutputWithPastAndCrossAttentions", "Embedding", "False", "ModelText2Config", "ModelText2Decoder", "ModelText2DecoderLayer", "ModelText2PreTrainedModel", "ModelText2SinusoidalPositionalEmbedding", "ModuleList", "None", "Setting", "The", "True", "ValueError", "You", "_", "__init__", "_prepare_4d_attention_mask", "_prepare_4d_causal_attention_mask", "all_cross_attentions", "all_hidden_states", "all_self_attns", "and", "at", "attention_mask", "attentions", "attn_mask", "be", "both", "but", "cannot", "checkpointing", "class", "config", "continue", "cross_attentions", "cross_attn_head_mask", "cross_attn_layer_head_mask", "d_model", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "decoder_layerdrop", "decoder_layers", "def", "dropout", "dropout_probability", "dtype", "either", "elif", "else", "embed_positions", "embed_scale", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "f", "for", "forward", "functional", "get_seq_length", "gradient", "gradient_checkpointing", "have", "head_mask", "hidden_states", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "is", "it", "last_hidden_state", "layer_head_mask", "layer_outputs", "layerdrop", "layers", "len", "logger", "mask_name", "math", "max_target_positions", "next_cache", "next_decoder_cache", "nn", "not", "or", "output_attentions", "output_hidden_states", "p", "pad_token_id", "padding_idx", "past_key_values", "past_key_values_length", "positions", "post_init", "r", "raise", "rand", "range", "return", "return_dict", "same", "scale_embedding", "self", "should", "size", "specified", "specify", "sqrt", "super", "tgt_len", "the", "time", "to", "torch", "training", "tuple", "use_cache", "use_return_dict", "v", "view", "vocab_size", "warning_once", "with", "zip"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2DecoderWrapper": ["ModelText2Decoder", "ModelText2DecoderWrapper", "ModelText2PreTrainedModel", "__init__", "args", "class", "config", "decoder", "def", "forward", "kwargs", "return", "self", "super"], "deprecated/speech_to_text_2/modeling_speech_to_text_2.py:Speech2Text2ForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelText2DecoderWrapper", "ModelText2ForCausalLM", "ModelText2PreTrainedModel", "None", "Optional", "Tensor", "True", "Union", "_CONFIG_FOR_DOC", "__init__", "_reorder_cache", "_tied_weights_keys", "attention_mask", "attentions", "beam_idx", "bias", "bool", "class", "config", "config_class", "cross_attentions", "cross_attn_head_mask", "decoder", "def", "device", "else", "embed_tokens", "encoder_attention_mask", "encoder_hidden_states", "for", "forward", "get_decoder", "get_input_embeddings", "get_seq_length", "head_mask", "hidden_size", "hidden_states", "if", "in", "index_select", "input_ids", "inputs_embeds", "is", "is_decoder", "is_encoder_decoder", "kwargs", "labels", "layer_past", "lm_head", "logits", "loss", "loss_fct", "model", "new_ones", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "past_length", "past_state", "post_init", "prepare_inputs_for_generation", "r", "remove_prefix_length", "reordered_past", "replace_return_docstrings", "return", "return_dict", "self", "set_decoder", "set_input_embeddings", "shape", "staticmethod", "super", "to", "torch", "tuple", "use_cache", "use_return_dict", "value", "view", "vocab_size", "weight"], "deprecated/jukebox/modeling_jukebox.py:filter_logits": ["F", "Inf", "Model_logits", "Model_value", "True", "bool", "clone", "cumsum", "cumulative_probs", "def", "descending", "dim", "dtype", "float", "if", "index", "indices_to_remove", "logits", "min", "return", "scatter_", "size", "softmax", "sort", "sorted_indices", "sorted_indices_to_remove", "sorted_logits", "src", "top_k", "top_p", "topk", "torch", "zeros_like"], "deprecated/jukebox/modeling_jukebox.py:get_relevant_lyric_tokens": ["Model_relevant_lyric_tokens", "cat", "def", "device", "dim", "dtype", "duration", "else", "full_tokens", "if", "indices", "int", "len", "list", "long", "max", "max_n_lyric_tokens", "midpoint", "min", "offset", "range", "return", "to", "tokens", "torch", "total_length", "unsqueeze", "zeros"], "deprecated/jukebox/modeling_jukebox.py:get_starts": ["Model_starts", "append", "def", "for", "hop_length", "if", "in", "n_ctx", "range", "return", "start", "starts", "total_length"], "deprecated/jukebox/modeling_jukebox.py:get_alignment": ["Computing", "Model_alignment", "Model_attn_weights", "Model_indices", "Model_metadata", "Model_starts", "True", "alignment", "alignment_head", "alignment_hop", "alignment_hops", "alignment_layer", "alignments", "append", "attn_layers", "batch_size", "cat", "chunk", "config", "cpu", "def", "del", "desc", "device", "dim", "dtype", "else", "end", "float", "for", "forward_tokens", "full_tokens", "hop_fraction", "hop_length", "if", "in", "indices", "indices_hop", "indices_hops", "int", "item", "labels", "len", "level", "levels", "lyric", "metadata", "metadata_bs", "metadata_i", "music", "music_tokens", "n_ctx", "np", "numpy", "offset", "padding_length", "prior", "prior_alignment_head", "prior_alignment_layer", "range", "return", "reversed", "sample_length", "shape", "start", "to", "tokens", "tokens_bs", "tokens_i", "torch", "total_length", "tqdm", "w_hop", "w_hops", "weights", "zeros", "zip"], "deprecated/jukebox/modeling_jukebox.py:save_temp_audio": ["Model", "Model_temp_audio", "None", "artists", "aud", "clamp", "cpu", "def", "else", "f", "fname", "for", "genres", "i", "if", "in", "is", "list", "lvl", "lvl_", "lyrics", "metas", "not", "np", "numpy", "path", "range", "sample", "shape", "torch", "values"], "deprecated/jukebox/modeling_jukebox.py:get_mask": ["Model_mask", "None", "autoregressive", "blocks", "contiguous", "def", "device", "elif", "else", "functional", "if", "is", "key_value_length", "mask", "max", "nn", "offset", "ones", "or", "pad", "prime", "query_length", "return", "sample", "sample_t", "spread", "summary", "torch", "tril", "value", "view"], "deprecated/jukebox/modeling_jukebox.py:JukeboxConv1D": ["ModelConv1D", "Module", "Parameter", "__init__", "addmm", "bias", "class", "def", "empty", "forward", "hidden_states", "input_width", "nn", "output_width", "return", "self", "size", "size_out", "super", "torch", "type_as", "view", "weight", "zeros"], "deprecated/jukebox/modeling_jukebox.py:JukeboxResConv1DBlock": ["Conv1d", "ModelResConv1DBlock", "Module", "ReLU", "__init__", "activation", "class", "config", "conv1d_1", "conv1d_2", "conv_width", "def", "depth", "dilation", "forward", "hidden_dim", "hidden_states", "nn", "padding", "res_convolution_multiplier", "res_dilation_growth_rate", "res_scale", "residuals", "return", "self", "super"], "deprecated/jukebox/modeling_jukebox.py:JukeboxResnet1D": ["False", "ModelResConv1DBlock", "ModelResnet1D", "Module", "ModuleList", "None", "__init__", "append", "block", "block_depth", "blocks", "class", "config", "conv_res_scale", "conv_width", "def", "depth", "dilation_cycle", "else", "for", "forward", "hidden_states", "if", "in", "is", "math", "n_depth", "nn", "not", "range", "res_dilation_cycle", "res_scale", "resnet_block", "return", "reverse_dilation", "self", "sqrt", "super"], "deprecated/jukebox/modeling_jukebox.py:JukeboxEncoderConvBlock": ["Conv1d", "ModelEncoderConvBlock", "ModelResnet1D", "Module", "ModuleList", "__init__", "append", "block", "blocks", "class", "config", "def", "depth", "down_t", "downsample_block", "else", "embed_dim", "filter_t", "for", "forward", "hidden_dim", "hidden_states", "i", "if", "in", "nn", "pad_t", "proj_out", "range", "return", "self", "stride_t", "super"], "deprecated/jukebox/modeling_jukebox.py:JukeboxEncoder": ["ModelEncoder", "ModelEncoderConvBlock", "Module", "ModuleList", "__init__", "all_hidden_states", "append", "class", "config", "conv_input_shape", "def", "depth", "down_t", "downs_t", "else", "embed_dim", "for", "forward", "hidden_states", "i", "if", "in", "iterator", "level", "level_block", "level_blocks", "levels", "list", "nn", "range", "return", "self", "stride_t", "strides_t", "super", "width", "zip"], "deprecated/jukebox/modeling_jukebox.py:JukeboxDecoderConvBock": ["Conv1d", "ConvTranspose1d", "ModelDecoderConvBock", "ModelResnet1D", "Module", "ModuleList", "True", "__init__", "append", "block", "blocks", "class", "config", "def", "depth", "down_t", "else", "embed_dim", "filter_t", "for", "forward", "hidden_dim", "hidden_states", "i", "if", "in", "nn", "pad_t", "proj_in", "range", "return", "reverse_dilation", "self", "stride_t", "super", "upsample_block"], "deprecated/jukebox/modeling_jukebox.py:JukeboxDecoder": ["Conv1d", "ModelDecoder", "ModelDecoderConvBock", "Module", "ModuleList", "True", "__init__", "all_levels", "and", "append", "class", "config", "conv_input_shape", "def", "depth", "down_t", "downs_t", "embed_dim", "for", "forward", "hidden_dim", "hidden_state", "hidden_states", "if", "in", "level", "level_block", "level_blocks", "levels", "list", "nn", "out", "range", "return", "reversed", "self", "stride_t", "strides_t", "super", "zip"], "deprecated/jukebox/modeling_jukebox.py:JukeboxBottleneckBlock": ["F", "False", "ModelBottleneckBlock", "ModelVQVAEConfig", "Module", "None", "True", "_", "__init__", "_codebook_elem", "_codebook_prob", "_codebook_sum", "_random_codebook", "_tile", "and", "batch_size", "class", "codebook", "codebook_elem", "codebook_sum", "codebook_weights", "codebook_width", "codes", "commit_loss", "config", "contiguous", "decode", "def", "dequantise", "dequantised_states", "detach", "device", "dict", "dim", "distance", "dk", "elif", "else", "embed_dim", "embed_width", "embedding", "encode", "entropy", "fit", "float", "forward", "hidden_states", "if", "init", "init_codebook", "keepdim", "latent_states", "latent_states_onehot", "linalg", "lmu", "log", "matmul", "mean", "min", "min_distance", "mu", "music_tokens", "n_repeats", "nb_discrete_codes", "nn", "no_grad", "norm", "norm_code", "not", "np", "old_codebook", "ones", "permute", "pn", "postprocess", "prenorm", "preprocess", "prod", "quantise", "randn_like", "randperm", "register_buffer", "repeat", "return", "samples", "scatter_", "self", "seq_len", "shape", "sqrt", "std", "sum", "super", "t", "threshold", "time", "torch", "update_codebook", "update_metrics", "usage", "used_curr", "view", "with", "x1", "x2", "x_shape", "zeros"], "deprecated/jukebox/modeling_jukebox.py:JukeboxBottleneck": ["ModelBottleneck", "ModelBottleneckBlock", "Module", "ModuleList", "None", "__init__", "append", "class", "commit_loss", "commit_losses", "config", "decode", "def", "detach", "encode", "end_level", "for", "forward", "hidden_states", "if", "in", "input_audio", "is", "level", "level_block", "level_blocks", "levels", "metric", "metrics", "music_tokens", "nn", "not", "quantised_audio", "quantised_state", "quantised_states", "range", "raw_audio", "return", "sampled_tokens", "self", "start_level", "super", "training", "update_codebook", "z", "zip"], "deprecated/jukebox/modeling_jukebox.py:JukeboxVQVAE": ["Embedding", "False", "FloatTensor", "LayerNorm", "Linear", "ModelBottleneck", "ModelConv1D", "ModelDecoder", "ModelEncoder", "ModelResConv1DBlock", "ModelVQVAE", "ModelVQVAEConfig", "ModuleList", "None", "PreTrainedModel", "Tensor", "_", "__init__", "_decode", "_encode", "_init_weights", "all_levels", "and", "append", "astype", "audio_chunks", "base_model_prefix", "bias", "bottleneck", "bs_chunks", "cat", "chunk", "chunk_i", "chunks", "class", "commit", "commit_loss", "commit_losses", "config", "conv1d_2", "cpu", "cumprod", "data", "decode", "decoder", "decoders", "def", "depth", "dequantised_state", "dequantised_states", "device", "dim", "down", "downs_t", "downsamples", "elif", "else", "encode", "encoder", "encoders", "end_level", "fill_", "float", "for", "forward", "hop_lengths", "i", "if", "in", "init_scale", "input_audio", "int", "is", "isinstance", "latent_state", "latent_states", "level", "levels", "loss", "mean", "module", "multipliers", "music_tokens", "music_tokens_i", "music_tokens_level", "music_tokens_list", "music_tokens_shape", "music_tokens_shapes", "n_samples", "nb_discrete_codes", "nn", "normal_", "not", "np", "permute", "prod", "randint", "range", "raw_audio", "res_conv_depth", "res_conv_width", "res_downs_t", "res_strides_t", "return", "sample", "sample_length", "sample_length_in_seconds", "sampling_rate", "self", "size", "start_level", "std", "stride", "strides_t", "sum", "super", "token", "token_chunks", "top_raw_to_tokens", "torch", "tuple", "vqvae", "weight", "width", "zero_", "zero_out", "zip"], "deprecated/jukebox/modeling_jukebox.py:JukeboxMLP": ["ACT2FN", "Dropout", "ModelConv1D", "ModelMLP", "Module", "__init__", "act", "act_fn", "c_fc", "c_proj", "class", "config", "def", "dropout", "embed_dim", "forward", "hidden_dim", "hidden_size", "hidden_states", "int", "mlp_multiplier", "nn", "resid_dropout", "return", "self", "super"], "deprecated/jukebox/modeling_jukebox.py:JukeboxLayerNorm": ["F", "FusedLayerNorm", "ModelLayerNorm", "True", "__init__", "bias", "class", "def", "elementwise_affine", "else", "eps", "forward", "if", "input", "layer_norm", "max_numel", "normalized_shape", "np", "numel", "prod", "return", "self", "super", "type_as", "weight", "width"], "deprecated/jukebox/modeling_jukebox.py:JukeboxAttention": ["ATTENTION_MAP", "Dropout", "F", "False", "ModelAttention", "ModelConv1D", "Module", "None", "NotImplementedError", "REQUIRED_CACHE_LEN", "True", "TypeError", "__init__", "_append_cache", "_attn", "_cache_len", "_encoder_len", "_offset", "_pad_to_block_ctx", "_slice_cache", "_suff_cache_len", "and", "attention_multiplier", "attention_prob", "attention_scores", "attention_weight", "attn", "attn_dropout", "attn_func", "attn_mask", "attn_weight_type", "autoregressive", "batch_size", "be", "block", "block_attn", "block_ctx", "block_len", "blocks", "c_attn", "c_enc_kv", "c_proj", "cache", "cat", "chunk", "class", "config", "context_states", "contiguous", "cross_attention", "cross_attn", "curr_ctx", "decode_qkv", "def", "del", "del_cache", "dense_attn", "device", "dim", "dropout", "dtype", "elif", "else", "embed_dim", "encoder_blocks", "encoder_len", "end", "factored_qkv", "float", "forward", "functional", "get_mask", "head_dim", "hidden_dim", "hidden_size", "hidden_states", "if", "in", "int", "is", "is_key", "key", "key_states", "l_cache", "last_encoder_hidden_states", "mask", "matmul", "merge_heads", "min", "mul_", "n_blocks", "n_ctx", "n_heads", "nb_key_blocks", "nb_query_blocks", "nb_relevant_lyric_tokens", "new_hidden_states_shape", "nn", "not", "offset", "old_key", "old_value", "pad", "permute", "prev_block_attn", "prev_l", "previous_block_length", "prime", "prime_attn", "prime_qkv", "property", "qkv", "query", "query_length", "query_states", "raise", "record_attn", "resid_dropout", "return", "sample", "sample_t", "scale", "self", "seq_len", "shape", "should", "size", "softmax", "split_heads", "spread", "start", "summary", "summary_attn", "summary_spread_attn", "super", "torch", "training", "transpose", "transpose_block_attn", "type", "type_as", "value", "value_states", "view", "zeros"], "deprecated/jukebox/modeling_jukebox.py:JukeboxBlock": ["False", "ModelAttention", "ModelBlock", "ModelLayerNorm", "ModelMLP", "Module", "__init__", "attn", "attn_func", "attn_res_scale", "class", "config", "def", "dense_attn", "else", "forward", "hidden_size", "hidden_states", "if", "last_encoder_hidden_states", "layer_norm_0", "layer_norm_1", "mlp", "n_ctx", "nn", "num_layers", "output", "output_states", "res_scale", "residuals", "return", "sample", "self", "super", "width"], "deprecated/jukebox/modeling_jukebox.py:JukeboxLayerStack": ["ATTENTION_PATTERNS", "False", "ModelBlock", "ModelLayerStack", "Module", "ModuleList", "None", "__init__", "_attn_mods", "_should_record_attn", "append", "attention_pattern", "attn", "attn_func", "attn_layer", "block_ctx", "blocks", "bool", "c_attn", "class", "config", "cross_attention", "def", "del_cache", "depth", "else", "encoder_len", "enumerate", "for", "forward", "hidden_size", "hidden_states", "i", "if", "in", "is", "isinstance", "last_encoder_hidden_states", "layer", "layer_idx", "n_ctx", "n_heads", "nb_relevant_lyric_tokens", "nn", "not", "num_layers", "range", "record_attn", "return", "sample", "saved_attn_weights", "self", "set_record_attn", "super", "weight", "width"], "deprecated/jukebox/modeling_jukebox.py:JukeboxPositionalEmbedding": ["ModelPositionalEmbedding", "Module", "Parameter", "__init__", "class", "def", "embed_dim", "empty", "forward", "nn", "pos_emb", "return", "self", "super", "torch", "width"], "deprecated/jukebox/modeling_jukebox.py:JukeboxConditionalAutoregressive": ["Ancestral", "Categorical", "CrossEntropyLoss", "Dropout", "Embedding", "False", "Linear", "ModelConditionalAutoregressive", "ModelLayerStack", "ModelPositionalEmbedding", "Module", "None", "Parameter", "Preparing", "Sampling", "True", "__init__", "_attn_mods", "activations", "add_cond_after_transformer", "append", "audio_conditioning", "batch_size", "bias", "c_fc", "cat", "chunk_size", "chunk_sizes", "class", "clone", "cond", "cond_prime", "conds_prime", "config", "current_chunk_size", "def", "del", "del_cache", "desc", "device", "dim", "distributions", "dtype", "elif", "else", "emb_dropout", "embed_dim", "embed_tokens", "embed_tokens_dropout", "empty", "encoder_len", "f", "fc_proj_out", "filter_logits", "for", "forward", "get_acts", "get_emb", "get_preds", "get_sep_loss", "hidden_size", "hidden_states", "if", "in", "input_tokens", "is", "is_encoder", "iter", "itererator", "key", "last_encoder_hidden_states", "leave", "len", "length", "list", "log", "logits", "long", "loss", "loss_fn", "lyric_and_music_tokens", "lyric_hidden_states", "lyric_loss", "merged_decoder", "metadata_conditioning", "mlp", "music", "music_token_loss", "music_tokens", "music_vocab_size", "n_ctx", "n_passes", "n_samples", "nb_relevant_lyric_tokens", "nn", "no_grad", "not", "np", "num_layers", "past", "pos_emb", "pos_emb_dropout", "preds", "primed_sample", "range", "refresh", "reshape", "return", "sample", "sample_t", "sample_tokens", "sampled_audio", "sampled_audio_prime", "sampled_tokens", "sampling", "self", "set_description", "shape", "share_embed_tokens_fc_proj_out", "split", "split_chunks", "start", "start_token", "super", "target", "temp", "to", "token", "token_hidden_states", "tokens", "top_k", "top_p", "torch", "tqdm", "transformer", "value", "view", "weight", "width", "with", "x_prime", "x_primes", "zeros"], "deprecated/jukebox/modeling_jukebox.py:JukeboxMusicTokenConditioner": ["Embedding", "False", "ModelDecoderConvBock", "ModelLayerNorm", "ModelMusicTokenConditioner", "Module", "None", "__init__", "class", "config", "def", "embed_dim", "embed_tokens", "forward", "hidden_size", "hidden_states", "if", "is", "layer_norm", "level", "long", "music_tokens", "music_vocab_size", "nn", "permute", "raw_audio_conditioning", "res_conv_depth", "res_conv_width", "res_downs_t", "res_strides_t", "return", "reverse_dilation", "self", "super", "upsampler"], "deprecated/jukebox/modeling_jukebox.py:JukeboxRangeEmbedding": ["Embedding", "Expected", "False", "ModelRangeEmbedding", "Module", "None", "Range", "TypeError", "__init__", "all", "and", "arange", "bins_", "clamp", "class", "def", "detach", "device", "dims", "dtype", "else", "emb", "embed_dim", "f", "float", "floor", "forward", "got", "if", "interpolation", "is", "len", "long", "n_time", "nn", "normalised_position", "not", "out_width", "pos_end", "pos_max", "pos_min", "pos_start", "position", "raise", "range", "return", "self", "shape", "super", "torch", "view", "with"], "deprecated/jukebox/modeling_jukebox.py:JukeboxLabelConditioner": ["Embedding", "ModelLabelConditioner", "ModelRangeEmbedding", "Module", "None", "True", "__init__", "absolute_pos_emb", "absolute_pos_range", "artist", "artist_emb", "bow_genre_emb", "clamp", "class", "config", "def", "dim", "else", "embed_dim", "end", "float", "forward", "genre", "genre_emb", "hidden_size", "if", "include_time_signal", "keepdim", "length", "mask", "max_duration", "max_nb_genres", "metadata", "metadata_dims", "min_duration", "music_tokens_shape", "n_ctx", "nb_artists", "nb_genres", "nn", "offset", "pos_emb", "relative_pos_emb", "relative_pos_range", "return", "sampling_rate", "self", "start", "start_emb", "sum", "super", "timing_dims", "total_length", "total_length_emb", "total_length_range", "unsqueeze"], "deprecated/jukebox/modeling_jukebox.py:JukeboxPrior": ["Ancestral", "Cond", "Embedding", "False", "LayerNorm", "Level", "Linear", "LongTensor", "ModelConditionalAutoregressive", "ModelConv1D", "ModelLabelConditioner", "ModelLayerNorm", "ModelMusicTokenConditioner", "ModelPositionalEmbedding", "ModelPrior", "ModelPriorConfig", "ModelRangeEmbedding", "ModelResConv1DBlock", "None", "Optional", "PreTrainedModel", "Primed", "Raw", "Sample", "Tensor", "True", "__init__", "_init_weights", "and", "append", "audio_conditioning", "base_model_prefix", "batch_size", "bias", "bins_shift", "bool", "bpd", "bs_chunks", "cat", "chunk_size", "clamp", "class", "clone", "cond_downsample", "cond_level", "conditioner_block", "conditioner_blocks", "conds", "config", "conv1d_2", "cross_entropy", "data", "decode", "def", "dequantised_states", "detach", "device", "dim", "dims", "down", "downsample", "downsamples", "dtype", "duration", "elif", "else", "emb", "embed_dim", "embed_dim_shift", "embed_tokens", "encode", "encoder", "encoder_config", "encoder_dim", "encoder_loss", "encoder_loss_fraction", "encoder_width", "end", "end_level", "f", "fill_", "final_layer_norm", "for", "forward", "forward_tokens", "full_tokens", "functional", "get_attn_weights", "get_cond", "get_encoder_loss", "get_encoder_states", "get_indices", "get_metadata", "get_music_tokens_conds", "get_preds", "get_relevant_lyric_tokens", "get_sep_loss", "hasattr", "hidden_size", "hidden_states", "i", "idx", "if", "in", "include_time_signal", "indices", "indices_list", "info", "init_cond", "init_scale", "input_shapes", "int", "is", "is_encoder", "is_encoder_decoder", "isinstance", "labels", "last_encoder_hidden_states", "latent_states", "len", "length", "level", "levels", "list", "lm_head", "log", "logger", "long", "loss", "lyric_acts", "lyric_acts_width", "lyric_and_music_tokens", "lyric_conditioning", "lyric_tokens", "lyric_vocab_size", "max_nb_genres", "mean", "metadata", "metadata_conditioning", "metadata_embedding", "metadata_pos", "metrics", "min", "missing_cond_len", "module", "music_tokens", "music_tokens_cond", "music_tokens_conds", "music_vocab_size", "n_ctx", "n_labels", "n_samples", "name", "nb_priors", "nb_relevant_lyric_tokens", "next_token_prediction_loss", "next_token_prediction_loss_dims", "nn", "no_grad", "no_past_context", "normal_", "not", "np", "offset", "or", "output", "pos_emb", "preds", "primed_sample", "prior", "prior_postprocess", "prior_preprocess", "priors", "prod", "proj_in", "range", "raw_to_tokens", "res_downs_t", "res_strides_t", "return", "reversed", "sample", "sample_length", "sample_tokens", "samples", "sampling", "saved_attn_weights", "self", "set_metadata_lyric_tokens", "set_record_attn", "shape", "split", "start", "start_level", "start_token", "std", "stride", "super", "target_lyrics", "temp", "tensor", "to", "tokens", "tokens_list", "top_k", "top_p", "torch", "total_length", "total_loss_dims", "transformer", "view", "vqvae_decoder", "vqvae_encoder", "weight", "width", "with", "zero_", "zero_out", "zeros", "zip"], "deprecated/jukebox/modeling_jukebox.py:JukeboxPreTrainedModel": ["False", "Model", "ModelConfig", "ModelPreTrainedModel", "ModelPrior", "ModelVQVAE", "PreTrainedModel", "__init__", "_init_weights", "apply", "base_model_prefix", "class", "config", "def", "if", "inputs", "isinstance", "kwargs", "module", "self", "super", "supports_gradient_checkpointing"], "deprecated/jukebox/modeling_jukebox.py:JukeboxModel": ["Ancestral", "Conditioning", "False", "LongTensor", "Model", "ModelBlock", "ModelModel", "ModelPreTrainedModel", "ModelPrior", "ModelVQVAE", "Model_SAMPLING_INPUT_DOCSTRING", "ModuleList", "None", "Primed", "Sampling", "Tensor", "True", "TypeError", "Unknown", "_", "__init__", "_no_split_modules", "_sample", "add_start_docstrings", "alignments", "ancestral_sample", "and", "append", "aud", "bs_chunks", "cat", "chunk_size", "class", "compute_alignments", "conditioning_tokens", "config", "continue_sample", "decode", "def", "device", "dim", "dtype", "elif", "else", "encode", "end", "end_level", "exists", "f", "float", "for", "get", "get_alignment", "get_metadata", "get_music_tokens_conds", "get_starts", "hop_fraction", "hop_length", "if", "in", "info", "input", "input_audio", "int", "is", "isinstance", "item", "iterator", "labels", "leave", "len", "level", "level_", "list", "logdir", "logger", "long", "lower_batch_size", "lyric_alignments", "makedirs", "max_batch_size", "max_duration", "max_nb_genres", "metadata", "metadata_conditioning", "metadata_i", "metadata_list", "metas", "min_duration", "model_config", "music_tokens", "music_tokens_conds", "music_tokens_conds_i", "music_tokens_conds_list", "music_tokens_i", "music_tokens_list", "music_tokens_new", "n_ctx", "n_passes", "n_samples", "name", "nb_priors", "nb_relevant_lyric_tokens", "nb_sampled_tokens", "new_tokens", "nn", "no_grad", "not", "obj", "of", "offset", "on", "os", "out", "path", "pop", "previous_sampled_tokens", "primed_sample", "prior", "prior_configs", "priors", "pt", "raise", "range", "raw_audio", "raw_to_tokens", "refresh", "return", "sample", "sample_length", "sample_length_in_seconds", "sample_level", "sample_levels", "sample_partial_window", "sample_single_window", "sample_tokens", "sampled_tokens", "sampling_kwargs", "sampling_rate", "sampling_temperature", "save", "save_results", "save_temp_audio", "self", "set_description", "set_shared_params", "shape", "split", "split_batch", "split_size", "start", "start_level", "super", "temp", "timing_dims", "to", "tokens", "tokens_i", "tokens_to_sample", "top_prior", "torch", "total_length", "total_token_to_sample", "tqdm", "type", "upsample", "vqvae", "vqvae_config", "with", "zeros", "zip"], "deprecated/nat/modeling_nat.py:NatEncoderOutput": ["FloatTensor", "ModelEncoderOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "reshaped_hidden_states", "torch", "tuple"], "deprecated/nat/modeling_nat.py:NatModelOutput": ["FloatTensor", "ModelModelOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "last_hidden_state", "pooler_output", "reshaped_hidden_states", "torch", "tuple"], "deprecated/nat/modeling_nat.py:NatImageClassifierOutput": ["FloatTensor", "ModelImageClassifierOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "reshaped_hidden_states", "torch", "tuple"], "deprecated/nat/modeling_nat.py:NatEmbeddings": ["Dropout", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelPatchEmbeddings", "Module", "Optional", "Tensor", "__init__", "class", "config", "def", "dropout", "embed_dim", "embeddings", "forward", "hidden_dropout_prob", "nn", "norm", "patch_embeddings", "pixel_values", "return", "self", "super", "torch", "tuple"], "deprecated/nat/modeling_nat.py:NatPatchEmbeddings": ["Conv2d", "DiModel", "FloatTensor", "Make", "ModelPatchEmbeddings", "Module", "Optional", "Sequential", "Tensor", "ValueError", "_", "__init__", "at", "channel", "class", "config", "configuration", "def", "dimension", "else", "embed_dim", "embeddings", "forward", "height", "hidden_size", "if", "in", "kernel_size", "match", "moment", "nn", "num_channels", "of", "one", "only", "padding", "pass", "patch", "patch_size", "permute", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "size", "stride", "super", "supports", "sure", "that", "the", "torch", "values", "width", "with"], "deprecated/nat/modeling_nat.py:NatDownsampler": ["Conv2d", "False", "LayerNorm", "ModelDownsampler", "Module", "None", "Tensor", "__init__", "bias", "class", "def", "dim", "forward", "input_feature", "int", "kernel_size", "nn", "norm", "norm_layer", "padding", "permute", "reduction", "return", "self", "stride", "super", "torch"], "deprecated/nat/modeling_nat.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "deprecated/nat/modeling_nat.py:NatDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "deprecated/nat/modeling_nat.py:NeighborhoodAttention": ["Dropout", "False", "Linear", "ModelAttention", "Module", "Optional", "Parameter", "Tensor", "The", "ValueError", "__init__", "a", "all_head_size", "attention", "attention_head_size", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bias", "bool", "class", "config", "context_layer", "contiguous", "def", "dim", "dropout", "else", "f", "forward", "functional", "heads", "hidden", "hidden_states", "if", "int", "is", "kernel_size", "key", "key_layer", "math", "multiple", "natten2dav", "natten2dqkrpb", "new_context_layer_shape", "new_x_shape", "nn", "not", "num_attention_heads", "num_heads", "number", "of", "output_attentions", "outputs", "permute", "qkv_bias", "query", "query_layer", "raise", "return", "rpb", "self", "size", "softmax", "sqrt", "super", "the", "torch", "transpose_for_scores", "tuple", "value", "value_layer", "view", "x", "zeros"], "deprecated/nat/modeling_nat.py:NeighborhoodAttentionOutput": ["Dropout", "Linear", "ModelAttentionOutput", "Module", "Tensor", "__init__", "attention_probs_dropout_prob", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_states", "input_tensor", "nn", "return", "self", "super", "torch"], "deprecated/nat/modeling_nat.py:NeighborhoodAttentionModule": ["False", "ModelAttention", "ModelAttentionModule", "ModelAttentionOutput", "Module", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_output", "bool", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "heads", "hidden_states", "if", "index", "kernel_size", "key", "len", "nn", "num_attention_heads", "num_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value"], "deprecated/nat/modeling_nat.py:NatIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "else", "forward", "hidden_act", "hidden_states", "if", "int", "intermediate_act_fn", "isinstance", "mlp_ratio", "nn", "return", "self", "str", "super", "torch"], "deprecated/nat/modeling_nat.py:NatOutput": ["Dropout", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dim", "dropout", "forward", "hidden_dropout_prob", "hidden_states", "int", "mlp_ratio", "nn", "return", "self", "super", "torch"], "deprecated/nat/modeling_nat.py:NatLayer": ["False", "Identity", "LayerNorm", "ModelDropPath", "ModelIntermediate", "ModelLayer", "ModelOutput", "Module", "NeighborhoodAttentionModule", "None", "Optional", "Parameter", "Tensor", "True", "_", "__init__", "attention", "attention_output", "attention_outputs", "batch_size", "bool", "channels", "chunk_size_feed_forward", "class", "config", "contiguous", "def", "dim", "drop_path", "drop_path_rate", "else", "eps", "forward", "functional", "height", "height_pad", "hidden_states", "if", "intermediate", "is", "kernel_size", "layer_norm_eps", "layer_output", "layer_outputs", "layer_scale_init_value", "layer_scale_parameters", "layernorm_after", "layernorm_before", "max", "maybe_pad", "nn", "not", "num_heads", "ones", "or", "output", "output_attentions", "pad", "pad_b", "pad_l", "pad_r", "pad_t", "pad_values", "requires_grad", "return", "self", "shape", "shortcut", "size", "super", "torch", "tuple", "was_padded", "width", "width_pad", "window_size"], "deprecated/nat/modeling_nat.py:NatStage": ["False", "LayerNorm", "ModelLayer", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "_", "__init__", "bool", "class", "config", "def", "depth", "dim", "downsample", "drop_path_rate", "else", "enumerate", "for", "forward", "height", "hidden_states", "hidden_states_before_downsampling", "i", "if", "in", "is", "layer_module", "layer_outputs", "layers", "nn", "norm_layer", "not", "num_heads", "output_attentions", "pointing", "range", "return", "self", "size", "stage_outputs", "super", "torch", "tuple", "width"], "deprecated/nat/modeling_nat.py:NatEncoder": ["False", "ModelDownsampler", "ModelEncoder", "ModelEncoderOutput", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "__init__", "all_hidden_states", "all_reshaped_hidden_states", "all_self_attentions", "and", "attentions", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "dim", "downsample", "dpr", "drop_path_rate", "elif", "else", "embed_dim", "enumerate", "for", "forward", "hidden_states", "hidden_states_before_downsampling", "i", "i_layer", "if", "in", "int", "is", "item", "last_hidden_state", "layer_module", "layer_outputs", "len", "levels", "linspace", "nn", "not", "num_heads", "num_levels", "output_attentions", "output_hidden_states", "output_hidden_states_before_downsampling", "permute", "range", "reshaped_hidden_state", "reshaped_hidden_states", "return", "return_dict", "self", "sum", "super", "torch", "tuple", "v", "x"], "deprecated/nat/modeling_nat.py:NatPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "weight", "zero_"], "deprecated/nat/modeling_nat.py:NatModel": ["AdaptiveAvgPool1d", "FloatTensor", "LayerNorm", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelModelOutput", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "Modelten", "None", "Optional", "True", "Union", "ValueError", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_EXPECTED_OUTPUT_SHAPE", "__init__", "_prune_heads", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention", "attentions", "bool", "checkpoint", "class", "config", "config_class", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "encoder_outputs", "eps", "expected_output", "flatten", "for", "forward", "get_input_embeddings", "have", "heads", "heads_to_prune", "hidden_states", "if", "in", "int", "is", "items", "last_hidden_state", "layer", "layer_norm_eps", "layernorm", "len", "modality", "nn", "not", "num_features", "num_levels", "output", "output_attentions", "output_hidden_states", "output_type", "patch_embeddings", "pixel_values", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "raise", "requires_backends", "reshaped_hidden_states", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "transpose", "tuple", "use_return_dict", "vision"], "deprecated/nat/modeling_nat.py:NatForImageClassification": ["FloatTensor", "Identity", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelImageClassifierOutput", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "Modelten", "None", "Optional", "Union", "_CONFIG_FOR_DOC", "_IMAGE_CLASS_CHECKPOINT", "_IMAGE_CLASS_EXPECTED_OUTPUT", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attentions", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "else", "expected_output", "forward", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_features", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pixel_values", "pooled_output", "post_init", "r", "requires_backends", "reshaped_hidden_states", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "deprecated/nat/modeling_nat.py:NatBackbone": ["BackboneMixin", "BackboneOutput", "LayerNorm", "ModelBackbone", "ModelEmbeddings", "ModelEncoder", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "Modelten", "ModuleDict", "None", "Optional", "Tensor", "True", "_CONFIG_FOR_DOC", "__init__", "_init_backbone", "add_start_docstrings_to_model_forward", "attentions", "batch_size", "bool", "channels", "class", "config", "config_class", "contiguous", "def", "depths", "else", "embed_dim", "embedding_output", "embeddings", "encoder", "feature_maps", "for", "forward", "get_input_embeddings", "height", "hidden_state", "hidden_states", "hidden_states_norms", "i", "if", "in", "int", "is", "len", "nn", "not", "num_channels", "num_features", "out_features", "output", "output_attentions", "output_hidden_states", "output_hidden_states_before_downsampling", "output_type", "outputs", "patch_embeddings", "permute", "pixel_values", "post_init", "range", "replace_return_docstrings", "requires_backends", "reshaped_hidden_states", "return", "return_dict", "self", "shape", "stage", "stage_names", "super", "torch", "use_return_dict", "view", "width", "zip"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMEmbeddings": ["Dropout", "Embedding", "LayerNorm", "LongTensor", "ModelMEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "class", "config", "cumsum", "def", "device", "dim", "dropout", "dtype", "embeddings", "eps", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "int", "int64", "is", "layer_norm", "layer_norm_eps", "max_position_embeddings", "nn", "normalized_shape", "ones", "p", "pad_token_id", "padding_idx", "past_key_values_length", "position_embeddings", "position_ids", "return", "self", "seq_length", "size", "super", "torch", "vocab_size", "word_embeddings"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMSelfAttention": ["Cache", "Dropout", "Embedding", "False", "FloatTensor", "Linear", "ModelMSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "a", "absolute", "all_head_size", "and", "arange", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "bhld", "bhlr", "bhrd", "bool", "cat", "class", "config", "context_layer", "contiguous", "def", "deprecate_kwarg", "device", "dim", "distance", "distance_embedding", "dropout", "dtype", "einsum", "elif", "else", "embedding_size", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "functional", "getattr", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "k_proj", "key_layer", "key_length", "long", "lrd", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_name", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "or", "output_attentions", "outputs", "past_key_value", "past_key_values", "permute", "position_embedding_type", "position_ids_l", "position_ids_r", "positional_embedding", "q_proj", "query_layer", "query_length", "raise", "relative_key", "relative_key_query", "relative_position_scores", "relative_position_scores_key", "relative_position_scores_query", "return", "self", "shape", "size", "softmax", "sqrt", "super", "tensor", "the", "to", "torch", "transpose", "transpose_for_scores", "tuple", "use_cache", "v_proj", "value_layer", "version", "view", "x"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMAttention": ["Cache", "False", "FloatTensor", "Linear", "ModelMAttention", "ModelMSelfAttention", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "deprecate_kwarg", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_size", "hidden_states", "if", "index", "k_proj", "len", "new_name", "nn", "num_attention_heads", "out_proj", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_embedding_type", "prune_heads", "prune_linear_layer", "pruned_heads", "q_proj", "return", "self", "self_attn", "self_outputs", "set", "super", "torch", "tuple", "union", "v_proj", "version"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMEncoderLayer": ["ACT2FN", "Cache", "Dropout", "FloatTensor", "LayerNorm", "Linear", "ModelMAttention", "ModelMEncoderLayer", "Module", "None", "Optional", "Tensor", "True", "__init__", "act_dropout", "activation", "attention_mask", "attention_opt_weights", "bool", "class", "config", "def", "deprecate_kwarg", "dropout", "dropout1", "dropout2", "else", "eps", "forward", "head_mask", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "intermediate_size", "is", "isinstance", "layer_norm_eps", "linear1", "linear2", "new_name", "nn", "norm1", "norm2", "output_attentions", "past_key_value", "past_key_values", "residual", "return", "self", "self_attn", "str", "super", "torch", "version"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "False", "FloatTensor", "ModelMEncoder", "ModelMEncoderLayer", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "attention_mask", "attentions", "bool", "class", "config", "def", "else", "enumerate", "for", "forward", "head_mask", "hidden_states", "i", "if", "in", "input_embeds", "is", "last_hidden_state", "layer", "layer_head_mask", "layers", "nn", "not", "num_hidden_layers", "opt_attn_weights", "output", "output_attentions", "output_hidden_states", "past_key_values", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMPooler": ["Linear", "ModelMPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelMConfig", "ModelMPreTrainedModel", "Model_m", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMModel": ["BaseModelOutputWithPastAndCrossAttentions", "BaseModelOutputWithPoolingAndCrossAttentions", "FloatTensor", "ModelMEmbeddings", "ModelMEncoder", "ModelMModel", "ModelMPooler", "ModelMPreTrainedModel", "Model_M_INPUTS_DOCSTRING", "None", "Optional", "True", "Union", "ValueError", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_TOKENIZER_FOR_DOC", "__init__", "_prune_heads", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "at", "attention_mask", "attentions", "batch_size", "bool", "both", "cannot", "checkpoint", "class", "concat", "config", "config_class", "def", "dim", "dtype", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "finfo", "float32", "for", "format", "forward", "get_head_mask", "get_input_embeddings", "get_seq_length", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "initializer_range", "input_ids", "inputs_embeds", "is", "items", "last_hidden_state", "layer", "layers", "min", "ndim", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "output_type", "pad_token_id", "past_key_values", "past_key_values_length", "past_mask", "pooler", "pooler_output", "position_ids", "post_init", "processor_class", "prune_heads", "raise", "return", "return_dict", "same", "self", "self_attn", "sequence_length", "sequence_output", "set_input_embeddings", "shape", "specify", "super", "tensor", "the", "time", "to", "torch", "tuple", "unsqueeze", "use_cache", "value", "word_embeddings", "zeros"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "MSELoss", "ModelMForSequenceClassification", "ModelMModel", "ModelMPreTrainedModel", "Model_M_INPUTS_DOCSTRING", "Model_m", "None", "Optional", "SequenceClassifierOutput", "Tensor", "True", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_TOKENIZER_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "dtype", "elif", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "pooled_output", "position_ids", "post_init", "problem_type", "processor_class", "r", "regression", "return", "return_dict", "self", "sequence_length", "single_label_classification", "squeeze", "super", "torch", "tuple", "use_cache", "use_return_dict", "view"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "ModelMForMultipleChoice", "ModelMModel", "ModelMPreTrainedModel", "Model_M_INPUTS_DOCSTRING", "Model_m", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "True", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "position_ids", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "torch", "tuple", "use_return_dict", "view"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMForTokenClassification": ["Cache", "CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "ModelMForTokenClassification", "ModelMModel", "ModelMPreTrainedModel", "Model_M_INPUTS_DOCSTRING", "Model_m", "None", "Optional", "Tensor", "TokenClassifierOutput", "True", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_TOKENIZER_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "position_ids", "post_init", "processor_class", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "super", "torch", "tuple", "use_return_dict", "view"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "ModelMForQuestionAnswering", "ModelMModel", "ModelMPreTrainedModel", "Model_M_INPUTS_DOCSTRING", "Model_m", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "True", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_TOKENIZER_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "clamp", "class", "config", "config_class", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "format", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "position_ids", "post_init", "processor_class", "qa_outputs", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple", "use_return_dict"], "deprecated/ernie_m/modeling_ernie_m.py:ErnieMForInformationExtraction": ["BCEWithLogitsLoss", "FloatTensor", "Linear", "ModelMForInformationExtraction", "ModelMModel", "ModelMPreTrainedModel", "Model_M_INPUTS_DOCSTRING", "Model_m", "None", "Optional", "QuestionAnsweringModelOutput", "Sigmoid", "Tensor", "True", "Union", "__init__", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "clamp", "class", "config", "def", "elif", "end_logits", "end_loss", "end_positions", "for", "format", "forward", "head_mask", "hidden_size", "hidden_states", "i", "if", "ignored_index", "in", "input_ids", "inputs_embeds", "is", "last_hidden_state", "len", "linear_end", "linear_start", "loss", "loss_fct", "nn", "not", "num_choices", "output_attentions", "output_hidden_states", "position_ids", "post_init", "r", "result", "return", "return_dict", "self", "sequence_length", "sequence_output", "sigmoid", "size", "squeeze", "start_logits", "start_loss", "start_positions", "super", "torch", "total_loss", "tuple"], "deprecated/mega/modeling_mega.py:MegaEmbeddings": ["Embedding", "False", "ModelConfig", "ModelEmbeddings", "Module", "Must", "None", "ValueError", "__init__", "add_token_type_embeddings", "and", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dtype", "elif", "else", "embeddings", "expand", "forward", "hasattr", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "long", "max_positions", "nn", "not", "of", "one", "or", "pad_token_id", "padding_idx", "persistent", "provide", "raise", "register_buffer", "return", "self", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "use_token_types", "vocab_size", "word_embeddings", "zeros"], "deprecated/mega/modeling_mega.py:MegaSimpleRelativePositionalBias": ["F", "ModelConfig", "ModelSimpleRelativePositionalBias", "Module", "Parameter", "Sequence", "Tensor", "ValueError", "__init__", "beyond", "bias", "chunk_size", "class", "config", "def", "else", "end", "f", "forward", "going", "if", "length", "max", "max_positions", "nn", "pad", "raise", "rel_pos_bias", "return", "self", "seq_len", "size", "start", "super", "tile", "torch", "view"], "deprecated/mega/modeling_mega.py:MegaRotaryRelativePositionalBias": ["FloatTensor", "ModelConfig", "ModelRotaryRelativePositionalBias", "Module", "None", "Parameter", "Rotary", "RuntimeError", "Tensor", "__init__", "_float_tensor", "a", "alpha", "arange", "b_param", "be", "bias", "cat", "chunk", "chunk_1", "chunk_2", "chunk_size", "class", "config", "cos", "cosine", "def", "dim", "dtype", "einsum", "else", "emb", "embed_dim", "embedding_dim", "exp", "expand", "float", "forward", "get_sinusoid_embeddings", "half_dim", "hidden_size", "if", "input", "int", "int64", "is", "log", "math", "max_positions", "mk", "mn", "multiple", "nk", "nn", "of", "or", "positional", "raise", "register_buffer", "requires", "return", "rotary", "rotary_alpha", "rotary_beta", "self", "seq_len", "shared_representation_size", "sin", "sine", "size", "staticmethod", "super", "to", "torch", "unsqueeze"], "deprecated/mega/modeling_mega.py:MegaDropout": ["F", "False", "Feature", "ModelDropout", "Module", "ValueError", "__init__", "are", "batch", "batch_first", "be", "bool", "class", "def", "dim", "dimension", "dimensional", "dropout", "dropout2d", "dropout_probability", "else", "exactly", "forward", "hidden", "if", "input", "inputs", "is_featurewise", "length", "must", "nn", "ordered", "p", "permute", "raise", "return", "self", "sequence", "size", "super", "training", "transpose"], "deprecated/mega/modeling_mega.py:MegaRMSNorm": ["ModelRMSNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "affine", "class", "def", "dim", "else", "eps", "extra_repr", "f", "forward", "if", "input", "is", "keepdim", "mean", "mean_square", "nn", "not", "num_features", "number_features", "register_parameter", "return", "rsqrt", "self", "square", "super", "torch", "weight"], "deprecated/mega/modeling_mega.py:MegaScaleNorm": ["ModelScaleNorm", "Module", "None", "Parameter", "Tensor", "True", "__init__", "affine", "class", "def", "dim", "else", "eps", "forward", "if", "input", "is", "keepdim", "mean", "mean_square", "nn", "not", "output", "register_parameter", "return", "rsqrt", "scalar", "self", "square", "super", "torch"], "deprecated/mega/modeling_mega.py:MegaSequenceNorm": ["BatchNorm", "BatchNorm1d", "False", "LayerNorm", "ModelRMSNorm", "ModelScaleNorm", "ModelSequenceNorm", "Module", "SyncBatchNorm", "True", "Unknown", "ValueError", "_BatchNorm", "__init__", "affine", "batchnorm", "be", "class", "def", "dim", "dimensional", "elementwise_affine", "elif", "else", "embedding_dim", "eps", "exactly", "export", "f", "forward", "if", "input", "inputs", "isinstance", "layernorm", "modules", "must", "nn", "norm", "norm_type", "permute", "raise", "return", "rmsnorm", "scalenorm", "self", "super", "syncbatchnorm", "type"], "deprecated/mega/modeling_mega.py:MegaMultiDimensionDampedEma": ["Bidirectional", "EMA", "F", "False", "ModelConfig", "ModelMultiDimensionDampedEma", "Module", "None", "Optional", "Parameter", "RuntimeError", "Tensor", "Unexpected", "ValueError", "__init__", "_coeffs", "_compute_efficient_ema_kernel", "_compute_ema_coefficients", "_kernel", "and", "arange", "attention_mask", "bd", "bdl", "bdn", "bidirectional", "bool", "bsz", "class", "config", "convolved_sequence", "damping_factor", "decay_factor", "def", "dim", "dimension", "dims", "dl", "dn", "dnl", "does", "einsum", "else", "ema_expansion_matrix", "ema_output", "ema_projection_size", "ema_step", "embed_dim", "embedding", "exp", "expects", "f", "fft", "fft_convolution", "fft_len", "flip", "float", "forward", "gated_ema_output", "get_ema_coefficients", "get_ema_kernel", "hidden_size", "if", "incremental", "input", "inputs", "inputs_fft", "int", "irfft", "is", "k1", "k2", "kernel", "kernel_dim", "kernel_fft", "kernel_proj", "kernel_projection_matrix", "kernel_size", "length", "log", "math", "min", "model", "n", "ndim", "nn", "not", "one_ema_step", "or", "out", "pad", "past_ema_proj", "past_ema_state", "past_state", "past_vandermonde", "permute", "prev_state", "previous_timestep_weight", "raise", "received", "residual", "residual_weight", "return", "rfft", "s_index", "scale", "self", "seq_len", "sigmoid", "silu", "size", "split", "sqrt", "squeeze", "state", "super", "support", "to", "torch", "training", "truncation", "type_as", "unsqueeze", "updated_hidden_state", "updated_state", "use_cache", "vander", "view"], "deprecated/mega/modeling_mega.py:MegaGatedCrossAttention": ["ACT2FN", "Cache", "F", "False", "Incremental", "Key", "Linear", "ModelConfig", "ModelDropout", "ModelGatedCrossAttention", "ModelRotaryRelativePositionalBias", "ModelSequenceNorm", "ModelSimpleRelativePositionalBias", "Module", "None", "Optional", "Position", "Softmax", "Tensor", "True", "Unexpected", "ValueError", "_", "__init__", "activation", "addcmul", "affine", "align", "and", "attention_activation", "attention_dropout", "attention_probs_dropout_prob", "attention_query", "attn_weights", "batch", "batch_first", "be", "bias", "bmm", "bool", "bsz", "but", "class", "config", "ctx_len", "decoding", "def", "dim", "dimension", "does", "dropout", "dropout_prob", "element_attention", "elif", "else", "embed_dim", "embedding", "expected", "f", "float", "forward", "full_query", "h_proj", "hidden_dropout", "hidden_dropout_prob", "hidden_size", "if", "inf", "input", "is", "is_featurewise", "k_proj", "kernel", "key", "key_padding_mask", "length", "lengths", "longer", "mask", "masked_fill", "max", "must", "nn", "norm", "norm_affine", "normalization_type", "normalize_before_Model", "not", "num_incremental_steps", "offset", "on", "out", "output_attentions", "outputs", "padding", "past_key_values", "pidx", "position", "prenorm", "prev_cross_key", "prev_cross_value", "prev_self_key", "projected_key", "projected_value", "provided", "q_proj", "qk", "queries", "query", "query_projected", "raise", "received", "rel_pos_bias", "relative", "relative_positional_bias", "requested", "residual_weight", "return", "rotary", "scaling", "self", "seq_len", "sequence", "shared_representation_size", "sigmoid", "silu", "simple", "simultaneously", "size", "softmax", "softmax_attention", "split", "src_len", "sum", "super", "target_gate", "tgt_len", "than", "the", "to", "token", "torch", "transpose", "tuple", "type_as", "unknown", "unsqueeze", "updated_cross_key", "updated_cross_value", "use_cache", "use_feature_dropout", "v_proj", "value", "view", "weighted_targets", "with"], "deprecated/mega/modeling_mega.py:MegaMovingAverageGatedAttention": ["ACT2FN", "Cache", "F", "False", "Incremental", "Input", "K", "Linear", "ModelConfig", "ModelDropout", "ModelMovingAverageGatedAttention", "ModelMultiDimensionDampedEma", "ModelRotaryRelativePositionalBias", "ModelSequenceNorm", "ModelSimpleRelativePositionalBias", "Module", "None", "Optional", "Parameter", "Q", "Size", "Softmax", "Standard", "Tensor", "Transformer", "True", "Unknown", "ValueError", "__init__", "activation", "addcmul", "additive_causal_mask", "affine", "all", "and", "as", "attention", "attention_activation", "attention_dropout", "attention_function", "attention_gate", "attention_mask", "attention_probs_dropout_prob", "attn_weights", "base", "batch_first", "be", "between", "bias", "bool", "bsz", "cat", "causal_mask", "chunk_size", "clamp", "class", "config", "ctx_len", "curr_len", "decoding", "def", "dim", "dimension", "dropout", "dropout_prob", "dtype", "element", "element_attention", "elif", "else", "ema_gate", "ema_out", "embed_dim", "embedding", "f", "float", "forward", "h_proj", "hidden_dropout", "hidden_dropout_prob", "hidden_size", "if", "in", "inf", "input", "intermediate_size", "intermediate_state", "is", "is_decoder", "is_featurewise", "keepdim", "kernel", "key", "length", "lengths", "logical_and", "masked_fill", "matmul", "min", "mismatch", "mx_proj", "n_chunks", "nn", "norm", "norm_affine", "normalization_type", "normalize_before_Model", "not", "of", "only", "original", "out", "output_attentions", "padding_mask", "padding_mask_all", "paper", "past_key_values", "positional", "prev_ema_state", "prev_self_key", "prev_self_value", "prev_state", "qk", "qk_bias", "qk_weight", "query", "query_key", "query_key_gates", "raise", "received", "rel_pos_bias", "relative", "relative_positional_bias", "reshape", "residual", "residual_weight", "return", "return_values", "rotary", "scaling", "self", "seq_len", "sequence", "shared_representation_size", "should", "sigmoid", "silu", "simple", "size", "softmax", "softmax_attention", "split", "sum", "super", "supports", "the", "to", "torch", "transpose", "type_as", "unbind", "unsqueeze", "updated_ema_state", "updated_self_key", "updated_self_value", "use_cache", "use_chunking", "use_feature_dropout", "v_proj", "value", "view", "weighted_self_output", "zeros_like"], "deprecated/mega/modeling_mega.py:MegaNormalizedFeedForwardNetwork": ["ACT2FN", "Linear", "ModelConfig", "ModelDropout", "ModelNormalizedFeedForwardNetwork", "ModelSequenceNorm", "Module", "__init__", "act_fn", "activation", "affine", "class", "config", "def", "dropout", "dropout_prob", "fc1", "fc2", "forward", "hidden", "hidden_dim", "hidden_dropout", "hidden_size", "if", "inputs", "is_featurewise", "nffn_activation_dropout_prob", "nffn_hidden_size", "nn", "norm", "norm_affine", "normalization_type", "normalize_before_ffn", "not", "output", "prenorm", "residual", "return", "self", "super", "use_feature_dropout"], "deprecated/mega/modeling_mega.py:MegaBlock": ["Cache", "False", "FloatTensor", "LongTensor", "ModelBlock", "ModelConfig", "ModelGatedCrossAttention", "ModelMovingAverageGatedAttention", "ModelNormalizedFeedForwardNetwork", "Model_layer", "Model_outputs", "Model_padding_mask", "Module", "None", "Optional", "Requested", "Tensor", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "as", "attention", "attention_mask", "be", "bool", "causal_mask", "class", "config", "cross", "cross_attention_weights", "cross_attn", "cross_attn_outputs", "cross_key", "cross_value", "decoder", "def", "deprecate_kwarg", "else", "encoder", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "hidden", "hidden_states", "if", "input", "is", "is_decoder", "key", "key_padding_mask", "model", "new_hidden_states", "new_key_values", "new_name", "nffn", "nn", "not", "output_attentions", "outs", "padding_mask", "past_key_value", "past_key_values", "providing", "query", "raise", "return", "self", "self_attention_weights", "self_ema_state", "self_key", "self_value", "seq_len_dim", "should", "states", "super", "torch", "tuple", "unsqueeze", "use_cache", "use_normalized_ffn", "used", "value", "version", "without"], "deprecated/mega/modeling_mega.py:MegaPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "deprecated/mega/modeling_mega.py:MegaPreTrainedModel": ["Embedding", "False", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelMovingAverageGatedAttention", "ModelMultiDimensionDampedEma", "ModelPreTrainedModel", "ModelRMSNorm", "ModelRotaryRelativePositionalBias", "ModelScaleNorm", "ModelSimpleRelativePositionalBias", "None", "PreTrainedModel", "_init_weights", "_no_split_modules", "add_", "alpha", "b_param", "base_model_prefix", "bias", "class", "config", "constant_", "damping_factor", "data", "decay_factor", "def", "elif", "ema_beta_range", "ema_delta_alpha_range", "ema_expansion_matrix", "ema_gamma_oModel_range", "ema_projection_size", "fill_", "idx", "if", "index_fill_", "init", "initializer_range", "is", "isinstance", "kernel_projection_matrix", "list", "mean", "module", "nn", "no_grad", "norm_affine", "normal_", "not", "ones", "padding_idx", "qk_bias", "qk_weight", "range", "rel_pos_bias", "residual_weight", "scalar", "self", "std", "supports_gradient_checkpointing", "tensor", "torch", "val", "weight", "with", "zero_"], "deprecated/mega/modeling_mega.py:MegaModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "False", "ModelBlock", "ModelConfig", "ModelEmbeddings", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "Model_layer", "Model_outputs", "ModuleList", "None", "Optional", "Received", "Tensor", "True", "Union", "ValueError", "You", "_", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "a", "activated", "add_code_sample_docstrings", "add_cross_attention", "add_pooling_layer", "add_start_docstrings_to_model_forward", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "at", "attention_mask", "attentions", "batch_size", "be", "bool", "both", "cache", "cannot", "causal_mask", "checkpoint", "chunk", "chunk_size", "class", "config", "config_class", "create_extended_attention_mask_for_decoder", "cross_attentions", "cross_attn_weights", "current_decoder_cache", "def", "device", "dtype", "either", "elif", "else", "embedding_layer", "embedding_output", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "expected", "f", "for", "format", "forward", "get_input_embeddings", "have", "hidden_states", "i", "if", "in", "input", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "key", "last_hidden_state", "layers", "len", "length", "long", "mismatch", "multiple", "must", "next_decoder_cache", "nn", "not", "nreceived", "num_hidden_layers", "of", "ones", "or", "output_attentions", "output_hidden_states", "output_type", "past", "past_key_values", "pooled_output", "pooler", "pooler_output", "post_init", "r", "raise", "range", "received", "return", "return_dict", "same", "self", "self_attn_weights", "sequence", "sequence_length", "set_input_embeddings", "shorter", "size", "specify", "squeeze", "super", "temp_mask_for_extension", "tensor", "than", "the", "time", "to", "token_type_ids", "torch", "transpose", "tuple", "updated_cache", "use_cache", "use_chunking", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "with", "word_embeddings"], "deprecated/mega/modeling_mega.py:MegaForCausalLM": ["Cache", "CausalLMOutputWithCrossAttentions", "CrossEntropyLoss", "False", "FloatTensor", "If", "Linear", "LongTensor", "Model", "ModelConfig", "ModelForCausalLM", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Tanh", "Tensor", "True", "Union", "_CONFIG_FOR_DOC", "__init__", "_reorder_cache", "_tied_weights_keys", "a", "add", "add_lm_hidden_dense_layer", "add_pooling_layer", "add_start_docstrings_to_model_forward", "as", "attention_mask", "attentions", "batch_size", "beam_idx", "bool", "class", "config", "config_class", "contiguous", "cross_attentions", "def", "dense", "device", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "format", "forward", "hidden_activation", "hidden_size", "hidden_states", "if", "in", "index_select", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "layer_past", "lm_head", "lm_loss", "logger", "logits", "loss", "loss_fct", "model_kwargs", "new_ones", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "past_state", "post_init", "prediction_scores", "prepare_inputs_for_generation", "r", "reordered_past", "replace_return_docstrings", "return", "return_dict", "self", "sequence_length", "sequence_output", "shape", "shifted_prediction_scores", "standalone", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_cache", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "deprecated/mega/modeling_mega.py:MegaForMaskedLM": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "If", "Linear", "LongTensor", "MaskedLMOutput", "Model", "ModelConfig", "ModelForMaskedLM", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Paris", "Tanh", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_tied_weights_keys", "add_code_sample_docstrings", "add_lm_hidden_dense_layer", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention", "attention_mask", "attentions", "batch_size", "bi", "bool", "checkpoint", "class", "config", "config_class", "def", "dense", "directional", "dropout", "dropout_prob", "else", "encoder_attention_mask", "encoder_hidden_states", "expected_loss", "expected_output", "for", "format", "forward", "get_output_embeddings", "hidden_activation", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "loss", "loss_fct", "mask", "masked_lm_loss", "mlm_head", "new_embeddings", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "post_init", "prediction_scores", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "set", "set_output_embeddings", "super", "to", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "weight", "you"], "deprecated/mega/modeling_mega.py:MegaForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "False", "FloatTensor", "LongTensor", "MSELoss", "Model", "ModelClassificationHead", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "dtype", "elif", "else", "format", "forward", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_length", "sequence_output", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/mega/modeling_mega.py:MegaForMultipleChoice": ["CrossEntropyLoss", "Dropout", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "dropout", "else", "flat_attention_mask", "flat_input_ids", "flat_inputs_embeds", "flat_token_type_ids", "format", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "post_init", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/mega/modeling_mega.py:MegaForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "else", "format", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "post_init", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/mega/modeling_mega.py:MegaClassificationHead": ["Dropout", "Linear", "ModelClassificationHead", "Module", "None", "__init__", "class", "classifier_dropout", "config", "def", "dense", "dropout", "else", "features", "forward", "hidden_dropout_prob", "hidden_size", "if", "is", "kwargs", "nn", "not", "num_labels", "out_proj", "return", "self", "super", "tanh", "torch", "x"], "deprecated/mega/modeling_mega.py:MegaForQuestionAnswering": ["CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "clamp", "class", "config", "config_class", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "format", "forward", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "deprecated/retribert/modeling_retribert.py:RetriBertPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "ModelBertConfig", "ModelBertPreTrainedModel", "Modelbert", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "weight", "zero_"], "deprecated/retribert/modeling_retribert.py:RetriBertModel": ["BertModel", "CrossEntropyLoss", "Dropout", "False", "FloatTensor", "Linear", "LongTensor", "ModelBertConfig", "ModelBertModel", "ModelBertPreTrainedModel", "None", "Optional", "Tensor", "__init__", "a_reps", "append", "arange", "attention_mask", "attention_mask_doc", "attention_mask_query", "b", "b_attention_mask", "b_embedding_output", "bert_doc", "bert_query", "bias", "cat", "ce_loss", "ceil", "checkpoint", "checkpoint_batch_size", "class", "compare_scores", "config", "def", "device", "dim", "dropout", "dtype", "else", "embed_answers", "embed_questions", "embed_sentences_checkpointed", "embedding_output", "embeddings", "encoder", "encoder_outputs", "extended_attention_mask", "for", "forward", "get_extended_attention_mask", "head_mask", "hidden_dropout_prob", "hidden_size", "if", "in", "input_ids", "input_ids_doc", "input_ids_query", "input_shape", "inputs", "inputs_embeds", "int", "is", "long", "loss", "loss_aq", "loss_qa", "math", "mean", "mm", "nn", "num_hidden_layers", "or", "partial_encode", "pooled_output", "pooled_output_list", "pooler", "position_ids", "post_init", "project_doc", "project_query", "projection_dim", "q_reps", "r", "range", "reduction", "return", "self", "sent_encoder", "sequence_output", "shape", "share_encoders", "size", "super", "t", "to", "token_type_ids", "torch", "zeros"], "deprecated/nezha/modeling_nezha.py:NezhaRelativePositionsEncoding": ["False", "ModelRelativePositionsEncoding", "Module", "__init__", "append", "arange", "clamp", "class", "cos", "def", "depth", "distance_mat", "distance_mat_clipped", "div_term", "dtype", "embeddings_table", "exp", "final_mat", "flat_relative_positions_matrix", "float", "forward", "functional", "int64", "length", "list", "log", "math", "matmul", "max_relative_position", "my_shape", "nn", "num_classes", "one_hot", "one_hot_relative_positions_matrix", "persistent", "position", "positions_encoding", "range_mat", "range_vec", "register_buffer", "repeat", "return", "self", "sin", "size", "super", "t", "torch", "unsqueeze", "view", "vocab_size", "zeros"], "deprecated/nezha/modeling_nezha.py:NezhaEmbeddings": ["Dropout", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelEmbeddings", "Module", "None", "Optional", "Tensor", "__init__", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "eps", "expand", "forward", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_ids", "input_shape", "inputs_embeds", "is", "layer_norm_eps", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "persistent", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "deprecated/nezha/modeling_nezha.py:NezhaSelfAttention": ["Cache", "Dropout", "False", "FloatTensor", "Linear", "ModelRelativePositionsEncoding", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "The", "ValueError", "__init__", "a", "all_head_size", "and", "attention", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_probs_t", "attention_scores", "attentions_probs_r", "batch_size", "bool", "cat", "class", "config", "context_layer", "contiguous", "def", "deprecate_kwarg", "depth", "dim", "dropout", "elif", "else", "encoder_attention_mask", "encoder_hidden_states", "f", "forward", "from_seq_length", "functional", "head_mask", "heads", "hidden", "hidden_size", "hidden_states", "if", "int", "is", "is_cross_attention", "is_decoder", "key", "key_layer", "key_position_scores", "key_position_scores_r", "key_position_scores_r_t", "length", "math", "matmul", "max_position_embeddings", "max_relative_position", "mixed_query_layer", "multiple", "new_context_layer_shape", "new_name", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "past_key_value", "past_key_values", "permute", "query", "query_layer", "query_layer_r", "query_layer_t", "raise", "relations_keys", "relations_values", "relative_positions_encoding", "return", "self", "size", "softmax", "sqrt", "super", "the", "to_seq_length", "torch", "transpose", "transpose_for_scores", "tuple", "value", "value_layer", "value_position_scores", "value_position_scores_r", "value_position_scores_r_t", "version", "view", "x"], "deprecated/nezha/modeling_nezha.py:NezhaSelfOutput": ["Dropout", "LayerNorm", "Linear", "ModelSelfOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "deprecated/nezha/modeling_nezha.py:NezhaAttention": ["Cache", "False", "FloatTensor", "ModelAttention", "ModelSelfAttention", "ModelSelfOutput", "Module", "None", "Optional", "Tensor", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "bool", "class", "config", "def", "dense", "deprecate_kwarg", "dim", "encoder_attention_mask", "encoder_hidden_states", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "new_name", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "past_key_value", "past_key_values", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "torch", "tuple", "union", "value", "version"], "deprecated/nezha/modeling_nezha.py:NezhaIntermediate": ["ACT2FN", "Linear", "ModelIntermediate", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super", "torch"], "deprecated/nezha/modeling_nezha.py:NezhaOutput": ["Dropout", "LayerNorm", "Linear", "ModelOutput", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super", "torch"], "deprecated/nezha/modeling_nezha.py:NezhaLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "If", "ModelAttention", "ModelIntermediate", "ModelLayer", "ModelOutput", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "a", "add_cross_attention", "added", "and", "apply_chunking_to_forward", "are", "as", "attention", "attention_mask", "attention_output", "be", "bool", "by", "chunk_size_feed_forward", "class", "config", "cross", "cross_attention_outputs", "cross_attn_past_key_value", "cross_attn_present_key_value", "crossattention", "decoder", "def", "deprecate_kwarg", "else", "encoder_attention_mask", "encoder_hidden_states", "f", "feed_forward_chunk", "forward", "has", "hasattr", "head_mask", "hidden_states", "if", "instantiated", "intermediate", "intermediate_output", "is", "is_decoder", "layer_output", "layers", "model", "new_name", "not", "output", "output_attentions", "outputs", "passed", "past_key_value", "past_key_values", "present_key_value", "raise", "return", "self", "self_attention_outputs", "self_attn_past_key_value", "seq_len_dim", "setting", "should", "super", "to", "torch", "tuple", "used", "version", "with"], "deprecated/nezha/modeling_nezha.py:NezhaEncoder": ["BaseModelOutputWithPastAndCrossAttentions", "Cache", "False", "FloatTensor", "ModelEncoder", "ModelLayer", "Module", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "_", "__init__", "add_cross_attention", "all_cross_attentions", "all_hidden_states", "all_self_attentions", "and", "attention_mask", "attentions", "bool", "checkpointing", "class", "config", "cross_attentions", "def", "else", "encoder_attention_mask", "encoder_hidden_states", "enumerate", "for", "forward", "gradient", "gradient_checkpointing", "head_mask", "hidden_states", "i", "if", "in", "incompatible", "is", "last_hidden_state", "layer", "layer_head_mask", "layer_module", "layer_outputs", "logger", "next_decoder_cache", "nn", "not", "num_hidden_layers", "output_attentions", "output_hidden_states", "past_key_values", "range", "return", "return_dict", "self", "super", "torch", "training", "tuple", "use_cache", "v", "warning_once", "with"], "deprecated/nezha/modeling_nezha.py:NezhaPooler": ["Linear", "ModelPooler", "Module", "Tanh", "Tensor", "__init__", "activation", "class", "config", "def", "dense", "first_token_tensor", "forward", "hidden_size", "hidden_states", "nn", "pooled_output", "return", "self", "super", "torch"], "deprecated/nezha/modeling_nezha.py:NezhaPredictionHeadTransform": ["ACT2FN", "LayerNorm", "Linear", "ModelPredictionHeadTransform", "Module", "Tensor", "__init__", "class", "config", "def", "dense", "else", "eps", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "isinstance", "layer_norm_eps", "nn", "return", "self", "str", "super", "torch", "transform_act_fn"], "deprecated/nezha/modeling_nezha.py:NezhaLMPredictionHead": ["False", "Linear", "ModelLMPredictionHead", "ModelPredictionHeadTransform", "Module", "Parameter", "__init__", "_tie_weights", "bias", "class", "config", "decoder", "def", "forward", "hidden_size", "hidden_states", "nn", "return", "self", "super", "torch", "transform", "vocab_size", "zeros"], "deprecated/nezha/modeling_nezha.py:NezhaOnlyMLMHead": ["ModelLMPredictionHead", "ModelOnlyMLMHead", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "nn", "prediction_scores", "predictions", "return", "self", "sequence_output", "super", "torch"], "deprecated/nezha/modeling_nezha.py:NezhaOnlyNSPHead": ["Linear", "ModelOnlyNSPHead", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "return", "self", "seq_relationship", "seq_relationship_score", "super"], "deprecated/nezha/modeling_nezha.py:NezhaPreTrainingHeads": ["Linear", "ModelLMPredictionHead", "ModelPreTrainingHeads", "Module", "__init__", "class", "config", "def", "forward", "hidden_size", "nn", "pooled_output", "prediction_scores", "predictions", "return", "self", "seq_relationship", "seq_relationship_score", "sequence_output", "super"], "deprecated/nezha/modeling_nezha.py:NezhaPreTrainedModel": ["Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "mean", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "deprecated/nezha/modeling_nezha.py:NezhaForPreTrainingOutput": ["FloatTensor", "ModelForPreTrainingOutput", "ModelOutput", "None", "Optional", "attentions", "class", "hidden_states", "loss", "prediction_logits", "seq_relationship_logits", "torch", "tuple"], "deprecated/nezha/modeling_nezha.py:NezhaModel": ["BaseModelOutputWithPoolingAndCrossAttentions", "Cache", "False", "ModelEmbeddings", "ModelEncoder", "ModelModel", "ModelPooler", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Tensor", "True", "Union", "ValueError", "You", "_", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_prune_heads", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "at", "attention", "attention_mask", "attentions", "batch_size", "bool", "both", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "cannot", "checkpoint", "class", "config", "config_class", "cross_attentions", "def", "device", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_batch_size", "encoder_extended_attention_mask", "encoder_hidden_shape", "encoder_hidden_states", "encoder_outputs", "encoder_sequence_length", "expand", "extended_attention_mask", "for", "format", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "get_seq_length", "hasattr", "have", "head_mask", "heads", "heads_to_prune", "hidden_states", "if", "in", "input_ids", "input_shape", "inputs_embeds", "invert_attention_mask", "is", "is_decoder", "items", "last_hidden_state", "layer", "long", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "output_type", "past_key_values", "past_key_values_length", "pooled_output", "pooler", "pooler_output", "post_init", "prune_heads", "r", "raise", "return", "return_dict", "same", "self", "seq_length", "sequence_length", "sequence_output", "set_input_embeddings", "size", "specify", "super", "the", "time", "to", "token_type_ids", "torch", "tuple", "use_cache", "use_return_dict", "value", "warn_if_padding_and_no_attention_mask", "word_embeddings", "zeros"], "deprecated/nezha/modeling_nezha.py:NezhaForPreTraining": ["CrossEntropyLoss", "Model", "ModelForPreTraining", "ModelForPreTrainingOutput", "ModelModel", "ModelPreTrainedModel", "ModelPreTrainingHeads", "Model_INPUTS_DOCSTRING", "None", "Optional", "Tensor", "Union", "_CONFIG_FOR_DOC", "__init__", "_tied_weights_keys", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bias", "bool", "class", "cls", "config", "config_class", "decoder", "def", "else", "format", "forward", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "loss", "loss_fct", "masked_lm_loss", "new_embeddings", "next_sentence_label", "next_sentence_loss", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "post_init", "prediction_logits", "prediction_scores", "predictions", "r", "replace_return_docstrings", "return", "return_dict", "self", "seq_relationship_logits", "seq_relationship_score", "sequence_length", "sequence_output", "set_output_embeddings", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict", "view", "vocab_size"], "deprecated/nezha/modeling_nezha.py:NezhaForMaskedLM": ["CrossEntropyLoss", "False", "If", "MaskedLMOutput", "Model", "ModelForMaskedLM", "ModelModel", "ModelOnlyMLMHead", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "PAD", "Tensor", "The", "Union", "ValueError", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "_tied_weights_keys", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention", "attention_mask", "attentions", "batch_size", "be", "bi", "bias", "bool", "cat", "checkpoint", "class", "cls", "config", "config_class", "decoder", "def", "defined", "device", "dim", "directional", "dtype", "dummy_token", "effective_batch_size", "else", "encoder_attention_mask", "encoder_hidden_states", "for", "format", "forward", "full", "generation", "get_output_embeddings", "head_mask", "hidden_states", "if", "input_ids", "input_shape", "inputs_embeds", "is", "is_decoder", "labels", "logger", "logits", "long", "loss", "loss_fct", "make", "masked_lm_loss", "model_kwargs", "new_embeddings", "new_zeros", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pad_token_id", "post_init", "prediction_scores", "predictions", "prepare_inputs_for_generation", "r", "raise", "return", "return_dict", "self", "sequence_length", "sequence_output", "set_output_embeddings", "shape", "should", "super", "sure", "to", "token", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "view", "vocab_size", "want", "warning", "you"], "deprecated/nezha/modeling_nezha.py:NezhaForNextSentencePrediction": ["CrossEntropyLoss", "FutureWarning", "Model", "ModelForNextSentencePrediction", "ModelModel", "ModelOnlyNSPHead", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "NextSentencePredictorOutput", "None", "Optional", "Tensor", "The", "Union", "_CONFIG_FOR_DOC", "__init__", "a", "add_start_docstrings_to_model_forward", "and", "argument", "attention_mask", "attentions", "batch_size", "be", "bool", "class", "cls", "config", "config_class", "def", "else", "format", "forward", "future", "head_mask", "hidden_states", "if", "in", "input_ids", "inputs_embeds", "instead", "is", "kwargs", "labels", "logits", "loss", "loss_fct", "next_sentence_label", "next_sentence_loss", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "pop", "post_init", "r", "removed", "replace_return_docstrings", "return", "return_dict", "self", "seq_relationship_scores", "sequence_length", "super", "token_type_ids", "torch", "tuple", "use", "use_return_dict", "version", "view", "warn", "warnings", "will"], "deprecated/nezha/modeling_nezha.py:NezhaForSequenceClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "Dropout", "Linear", "MSELoss", "Model", "ModelForSequenceClassification", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "SequenceClassifierOutput", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "dtype", "elif", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "multi_label_classification", "nn", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "post_init", "problem_type", "r", "regression", "return", "return_dict", "self", "sequence_length", "single_label_classification", "squeeze", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/nezha/modeling_nezha.py:NezhaForMultipleChoice": ["CrossEntropyLoss", "Dropout", "Linear", "Model", "ModelForMultipleChoice", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "MultipleChoiceModelOutput", "None", "Optional", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_choices", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pooled_output", "post_init", "print", "r", "reshaped_logits", "return", "return_dict", "self", "sequence_length", "shape", "size", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/nezha/modeling_nezha.py:NezhaForTokenClassification": ["CrossEntropyLoss", "Dropout", "False", "Linear", "Model", "ModelForTokenClassification", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Tensor", "TokenClassifierOutput", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "class", "classifier", "classifier_dropout", "config", "config_class", "def", "dropout", "else", "format", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "post_init", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "super", "token_type_ids", "torch", "tuple", "use_return_dict", "view"], "deprecated/nezha/modeling_nezha.py:NezhaForQuestionAnswering": ["CrossEntropyLoss", "False", "Linear", "Model", "ModelForQuestionAnswering", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "QuestionAnsweringModelOutput", "Tensor", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "__init__", "add_code_sample_docstrings", "add_pooling_layer", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "batch_size", "bool", "checkpoint", "clamp", "class", "config", "config_class", "contiguous", "def", "dim", "else", "end_logits", "end_loss", "end_positions", "format", "forward", "head_mask", "hidden_size", "hidden_states", "if", "ignore_index", "ignored_index", "input_ids", "inputs_embeds", "is", "len", "logits", "loss", "loss_fct", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "post_init", "qa_outputs", "r", "return", "return_dict", "self", "sequence_length", "sequence_output", "size", "split", "squeeze", "start_logits", "start_loss", "start_positions", "super", "token_type_ids", "torch", "total_loss", "tuple", "use_return_dict"], "deprecated/mctct/modeling_mctct.py:MCTCTConv1dSubsampler": ["Conv1d", "Dropout", "MCTCTConfig", "Model", "Module", "ModuleList", "Need", "None", "ValueError", "__init__", "class", "config", "configuration", "constant", "contiguous", "conv", "conv_channels", "conv_dropout", "conv_glu_dim", "conv_kernel", "conv_layers", "conv_stride", "convolution", "def", "dim", "dropout", "else", "enumerate", "for", "forward", "functional", "glu", "glu_dim", "hidden_size", "hidden_states", "i", "if", "in", "in_channels", "input_channels", "input_feat_per_channel", "input_features", "is", "k", "kernel_size", "layers", "mid_channels", "multiple", "nn", "num_conv_layers", "num_layers", "out_channels", "pad", "padding", "raise", "return", "self", "size", "specify", "stride", "sum", "super", "to", "torch", "transpose", "use", "valid"], "deprecated/mctct/modeling_mctct.py:MCTCTEmbeddings": ["Dropout", "Embedding", "False", "LayerNorm", "MCTCTLayerNorm", "Model", "Module", "None", "__init__", "arange", "buffered_token_type_ids", "buffered_token_type_ids_expanded", "class", "config", "def", "device", "dropout", "dtype", "else", "embeddings", "expand", "forward", "hasattr", "hidden_dropout_prob", "hidden_size", "if", "input_features", "input_shape", "inputs_embeds", "is", "long", "max_position_embeddings", "nn", "not", "pad_token_id", "padding_idx", "past_key_values_length", "persistent", "position_embeddings", "position_ids", "register_buffer", "return", "self", "seq_length", "size", "super", "token_type_embeddings", "token_type_ids", "torch", "type_vocab_size", "vocab_size", "word_embeddings", "zeros"], "deprecated/mctct/modeling_mctct.py:MCTCTSelfAttention": ["Dropout", "Embedding", "False", "Linear", "Model", "Module", "None", "The", "ValueError", "__init__", "a", "all_head_size", "and", "attention", "attention_head_dim", "attention_head_size", "attention_mask", "attention_probs", "attention_probs_dropout_prob", "attention_scores", "batch", "bche", "bcle", "bias", "cat", "class", "config", "context_layer", "def", "device", "dim", "distance_embedding", "dropout", "einsum", "else", "embedding_size", "f", "flatten", "forward", "functional", "halfpoint", "hasattr", "head_mask", "heads", "hidden", "hidden_size", "hidden_state", "hidden_states", "if", "is", "is_decoder", "key", "key_layer", "len", "lh", "math", "matmul", "max_position_embeddings", "mixed_query_layer", "multiple", "new_x_shape", "nn", "not", "num_attention_heads", "number", "of", "output_attentions", "outputs", "permute", "positional_embedding", "query", "query_layer", "raise", "range", "relative_position_embedding_rotate", "relative_position_scores", "reshape", "reshape_fortran", "return", "reversed", "scores", "self", "seq_len", "shape", "size", "softmax", "sqrt", "start_dim", "super", "the", "torch", "transpose", "transpose_for_scores", "value", "value_layer", "view", "weight", "x", "zeros"], "deprecated/mctct/modeling_mctct.py:MCTCTLayerNorm": ["Model", "Module", "Parameter", "__init__", "class", "def", "forward", "hidden_states", "nn", "ones", "return", "self", "singleton_bias", "singleton_weight", "super", "torch", "zeros"], "deprecated/mctct/modeling_mctct.py:MCTCTSelfOutput": ["Dropout", "False", "LayerNorm", "Linear", "Model", "Module", "__init__", "bias", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "layer_norm_eps", "nn", "return", "self", "super"], "deprecated/mctct/modeling_mctct.py:MCTCTAttention": ["False", "MCTCTSelfAttention", "MCTCTSelfOutput", "Model", "Module", "None", "__init__", "all_head_size", "attention_head_size", "attention_mask", "attention_output", "class", "config", "def", "dense", "dim", "find_pruneable_heads_and_indices", "forward", "head_mask", "heads", "hidden_states", "if", "index", "key", "len", "nn", "num_attention_heads", "output", "output_attentions", "outputs", "prune_heads", "prune_linear_layer", "pruned_heads", "query", "return", "self", "self_outputs", "set", "super", "union", "value"], "deprecated/mctct/modeling_mctct.py:MCTCTIntermediate": ["ACT2FN", "False", "Linear", "Model", "Module", "__init__", "bias", "class", "config", "def", "dense", "else", "forward", "hidden_act", "hidden_size", "hidden_states", "if", "intermediate_act_fn", "intermediate_size", "isinstance", "nn", "return", "self", "str", "super"], "deprecated/mctct/modeling_mctct.py:MCTCTOutput": ["Dropout", "False", "LayerNorm", "Linear", "Model", "Module", "__init__", "bias", "class", "config", "def", "dense", "dropout", "eps", "forward", "hidden_dropout_prob", "hidden_size", "hidden_states", "input_tensor", "intermediate_size", "layer_norm_eps", "nn", "return", "self", "super"], "deprecated/mctct/modeling_mctct.py:MCTCTLayer": ["False", "GradientCheckpointingLayer", "MCTCTAttention", "MCTCTConfig", "MCTCTIntermediate", "MCTCTOutput", "Model", "None", "__init__", "apply_chunking_to_forward", "attention", "attention_mask", "attention_output", "chunk_size_feed_forward", "class", "config", "def", "feed_forward_chunk", "forward", "head_mask", "hidden_states", "intermediate", "intermediate_output", "is_decoder", "layer_output", "output", "output_attentions", "outputs", "return", "self", "self_attention_outputs", "seq_len_dim", "super"], "deprecated/mctct/modeling_mctct.py:MCTCTPreTrainedModel": ["Conv1d", "Embedding", "LayerNorm", "Linear", "LongTensor", "MCTCTConfig", "MCTCTLayerNorm", "Model", "None", "PreTrainedModel", "True", "_", "_get_feat_extract_output_lengths", "_get_feature_vector_attention_mask", "_init_weights", "arange", "attention_mask", "base_model_prefix", "bias", "bsz", "class", "config", "conv_kernel", "conv_stride", "cumsum", "data", "def", "device", "dilation", "div", "dtype", "elif", "feature_vector_length", "fill_", "flip", "for", "if", "in", "initializer_range", "input_features", "input_lengths", "is", "isinstance", "kernel_sz", "len", "long", "main_input_name", "mctct", "mean", "module", "nn", "normal_", "not", "num_conv_layers", "padding", "padding_idx", "range", "return", "rounding_mode", "self", "shape", "singleton_bias", "singleton_weight", "size", "std", "stride", "subsampled_lengths", "sum", "supports_gradient_checkpointing", "torch", "trunc", "weight", "zero_", "zeros", "zip"], "deprecated/mctct/modeling_mctct.py:MCTCTEncoder": ["BaseModelOutput", "False", "MCTCTConfig", "MCTCTConv1dSubsampler", "MCTCTLayer", "MCTCTLayerNorm", "MCTCTPreTrainedModel", "Model", "ModuleList", "None", "Tensor", "The", "True", "Union", "ValueError", "_", "__init__", "_get_feature_vector_attention_mask", "_prepare_4d_attention_mask", "all_attentions", "and", "attention_mask", "attentions", "be", "bool", "but", "class", "config", "conv", "def", "dropout", "dropout_probability", "dtype", "else", "encoder_layer", "encoder_states", "enumerate", "f", "for", "forward", "functional", "gradient_checkpointing", "head_mask", "hidden_dropout_prob", "hidden_states", "idx", "if", "in", "input_features", "inputs_embeds", "is", "is_deepspeed_zero3_enabled", "is_fsdp_managed_module", "it", "last_hidden_state", "layer_norm", "layer_outputs", "layerdrop", "layers", "len", "nn", "not", "num_hidden_layers", "or", "output_attentions", "output_hidden_states", "p", "raise", "rand", "range", "return", "return_dict", "self", "shape", "should", "size", "skip_the_layer", "specified", "super", "synced_gpus", "torch", "training", "tuple", "use_return_dict", "v"], "deprecated/mctct/modeling_mctct.py:MCTCTModel": ["BaseModelOutput", "MCTCTEncoder", "MCTCTPreTrainedModel", "MCTCT_INPUTS_DOCSTRING", "Model", "None", "Optional", "Tensor", "Union", "ValueError", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_EXPECTED_OUTPUT_SHAPE", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attention_mask", "attentions", "audio", "batch_size", "bool", "checkpoint", "class", "config", "config_class", "def", "else", "encoder", "encoder_outputs", "expected_output", "format", "forward", "have", "head_mask", "hidden_states", "if", "input_features", "is", "last_hidden_state", "modality", "not", "output_attentions", "output_hidden_states", "output_type", "post_init", "raise", "return", "return_dict", "self", "sequence_length", "sequence_output", "specify", "super", "to", "torch", "tuple", "use_return_dict"], "deprecated/mctct/modeling_mctct.py:MCTCTForCTC": ["CausalLMOutput", "False", "Label", "Linear", "LongTensor", "MCTCTModel", "MCTCTPreTrainedModel", "MCTCT_INPUTS_DOCSTRING", "Model", "None", "Optional", "Please", "Tensor", "Union", "ValueError", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_CTC_EXPECTED_LOSS", "_CTC_EXPECTED_OUTPUT", "_HIDDEN_STATES_START_POSITION", "__class__", "__init__", "_get_feat_extract_output_lengths", "a", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "and", "are", "as", "attention_mask", "attentions", "backends", "be", "blank", "bool", "checkpoint", "class", "config", "config_class", "configuration", "ctc_head", "ctc_loss", "ctc_loss_reduction", "ctc_zero_infinity", "cudnn", "def", "define", "dim", "does", "dtype", "else", "enabled", "expected_loss", "expected_output", "f", "flags", "flattened_targets", "float32", "follows", "forward", "from_pretrained", "functional", "head", "head_mask", "hidden_size", "hidden_states", "if", "input_features", "input_lengths", "instantiate", "is", "labels", "labels_mask", "language", "log_probs", "log_softmax", "logits", "long", "loss", "masked_select", "max", "mctct", "model", "must", "nn", "not", "of", "ones", "or", "output", "output_attentions", "output_hidden_size", "output_hidden_states", "output_type", "outputs", "pad_token_id", "post_init", "r", "raise", "reduction", "return", "return_dict", "s", "self", "shape", "size", "sum", "super", "target_lengths", "that", "the", "to", "torch", "transpose", "trying", "tuple", "use_return_dict", "values", "vocab_size", "vocabulary", "with", "your", "zero_infinity"], "deprecated/mmbt/modeling_mmbt.py:ModalEmbeddings": ["Dropout", "LayerNorm", "Linear", "ModelEmbeddings", "Model_hidden_size", "Module", "None", "__init__", "arange", "cat", "class", "config", "def", "device", "dim", "dropout", "dtype", "embeddings", "encoder", "end_token", "end_token_embeds", "expand", "forward", "hidden_dropout_prob", "hidden_size", "if", "input_Model", "is", "long", "nn", "not", "p", "position_embeddings", "position_ids", "proj_embeddings", "return", "self", "seq_length", "size", "start_token", "start_token_embeds", "super", "token_embeddings", "token_type_embeddings", "token_type_ids", "torch", "unsqueeze", "word_embeddings", "zeros"], "deprecated/mmbt/modeling_mmbt.py:MMBTModel": ["BaseModelOutputWithPooling", "MMBT_INPUTS_DOCSTRING", "ModalEmbeddings", "Model", "Module", "ModuleUtilsMixin", "None", "ValueError", "You", "_CONFIG_FOR_DOC", "__init__", "add_start_docstrings_to_model_forward", "and", "at", "attention_mask", "attentions", "both", "cannot", "cat", "class", "config", "config_class", "def", "device", "dim", "dtype", "either", "elif", "else", "embedding_output", "embeddings", "encoder", "encoder_attention_mask", "encoder_extended_attention_mask", "encoder_hidden_states", "encoder_outputs", "end_token", "extended_attention_mask", "forward", "get_extended_attention_mask", "get_head_mask", "get_input_embeddings", "have", "head_mask", "hidden_states", "if", "input_ids", "input_modal", "input_modal_shape", "input_shape", "input_txt_shape", "inputs_embeds", "invert_attention_mask", "is", "last_hidden_state", "long", "modal_embeddings", "modal_encoder", "modal_end_tokens", "modal_position_ids", "modal_start_tokens", "modal_token_type_ids", "nn", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "output_type", "pooled_output", "pooler", "pooler_output", "position_ids", "r", "raise", "replace_return_docstrings", "return", "return_dict", "same", "self", "sequence_output", "set_input_embeddings", "size", "specify", "start_token", "super", "the", "time", "to", "token_type_ids", "torch", "transformer", "txt_embeddings", "use_return_dict", "value", "word_embeddings"], "deprecated/mmbt/modeling_mmbt.py:MMBTForClassification": ["CrossEntropyLoss", "Dropout", "Linear", "MMBTModel", "MSELoss", "Model", "Module", "None", "SequenceClassifierOutput", "__init__", "attention_mask", "attentions", "class", "classifier", "config", "def", "dropout", "else", "encoder", "forward", "head_mask", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_ids", "input_modal", "inputs_embeds", "is", "labels", "logits", "loss", "loss_fct", "mmbt", "modal_end_tokens", "modal_position_ids", "modal_start_tokens", "modal_token_type_ids", "nn", "not", "num_labels", "output", "outputs", "pooled_output", "position_ids", "r", "return", "return_dict", "self", "super", "token_type_ids", "transformer", "use_return_dict", "view"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerPatchEmbeddings": ["BatchNorm2d", "Conv2d", "Identity", "Make", "ModelFormerConfig", "ModelFormerPatchEmbeddings", "Module", "Tensor", "True", "ValueError", "__init__", "apply_norm", "batch_norm_eps", "batch_size", "bool", "channel", "class", "config", "configuration", "def", "dimension", "downsample_pad", "downsample_patch_size", "downsample_stride", "else", "embed_dim", "embeddings", "eps", "forward", "height", "if", "in", "int", "kernel_size", "match", "nn", "norm", "num_channels", "of", "one", "padding", "pixel", "pixel_values", "projection", "raise", "return", "self", "set", "shape", "stride", "super", "sure", "that", "the", "torch", "values", "width", "with"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerSelfAttention": ["False", "Linear", "LongTensor", "ModelFormerSelfAttention", "Module", "Parameter", "Tensor", "True", "__init__", "ab", "abs", "and", "append", "attention_bias_idxs", "attention_biases", "attention_offsets", "attention_probs", "attention_ratio", "batch_size", "bool", "class", "context_layer", "def", "del", "device", "dim", "else", "expanded_key_dim", "for", "forward", "hasattr", "hidden_size", "hidden_states", "idxs", "if", "in", "int", "itertools", "key_dim", "key_layer", "len", "list", "matmul", "mode", "nn", "no_grad", "not", "num_channels", "num_heads", "num_points", "offset", "output_attentions", "outputs", "permute", "point_1", "point_2", "points", "product", "projection", "qkv", "query_layer", "range", "register_buffer", "reshape", "resolution", "return", "scale", "self", "sequence_length", "shape", "softmax", "split", "super", "to", "torch", "total_expanded_key_dim", "total_key_dim", "train", "training", "transpose", "tuple", "value_layer", "view", "zeros"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerConvStem": ["BatchNorm2d", "Conv2d", "ModelFormerConfig", "ModelFormerConvStem", "Module", "ReLU", "Tensor", "__init__", "activation", "batch_norm_eps", "batchnorm_after", "batchnorm_before", "class", "config", "convolution1", "convolution2", "def", "eps", "features", "forward", "int", "kernel_size", "nn", "num_channels", "out_channels", "padding", "pixel_values", "return", "self", "stride", "super", "torch"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerPooling": ["AvgPool2d", "False", "ModelFormerPooling", "Module", "Tensor", "__init__", "class", "count_include_pad", "def", "forward", "hidden_states", "int", "nn", "output", "padding", "pool", "pool_size", "return", "self", "stride", "super", "torch"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerDenseMlp": ["ACT2FN", "Dropout", "Linear", "ModelFormerConfig", "ModelFormerDenseMlp", "Module", "None", "Optional", "Tensor", "__init__", "activation", "class", "config", "def", "dropout", "forward", "hidden_act", "hidden_dropout_prob", "hidden_features", "hidden_states", "in_features", "int", "linear_in", "linear_out", "nn", "or", "out_features", "return", "self", "super", "torch"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerConvMlp": ["ACT2FN", "BatchNorm2d", "Conv2d", "Dropout", "ModelFormerConfig", "ModelFormerConvMlp", "Module", "None", "Optional", "Tensor", "__init__", "activation", "batch_norm_eps", "batchnorm_after", "batchnorm_before", "class", "config", "convolution1", "convolution2", "def", "drop", "dropout", "eps", "float", "forward", "hidden_act", "hidden_features", "hidden_state", "in_features", "int", "nn", "or", "out_features", "return", "self", "super", "torch"], "deprecated/efficientformer/modeling_efficientformer.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerDropPath": ["ModelFormerDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerFlat": ["ModelFormerFlat", "Module", "Tensor", "__init__", "class", "def", "flatten", "forward", "hidden_states", "nn", "return", "self", "super", "torch", "transpose", "tuple"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerMeta3D": ["False", "Identity", "LayerNorm", "ModelFormerConfig", "ModelFormerDenseMlp", "ModelFormerDropPath", "ModelFormerMeta3D", "ModelFormerSelfAttention", "Module", "Parameter", "Tensor", "True", "__init__", "attention_output", "attention_ratio", "bool", "class", "config", "def", "dim", "drop_path", "else", "eps", "float", "forward", "hidden_features", "hidden_states", "if", "in_features", "int", "key_dim", "layer_norm_eps", "layer_output", "layer_scale_1", "layer_scale_2", "layer_scale_init_value", "layernorm1", "layernorm2", "mlp", "mlp_expansion_ratio", "mlp_hidden_dim", "nn", "num_attention_heads", "num_heads", "ones", "output_attentions", "outputs", "requires_grad", "resolution", "return", "self", "self_attention_outputs", "super", "token_mixer", "torch", "tuple", "unsqueeze", "use_layer_scale"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerMeta3DLayers": ["False", "ModelFormerConfig", "ModelFormerMeta3D", "ModelFormerMeta3DLayers", "Module", "ModuleList", "None", "Tensor", "__init__", "all_attention_outputs", "block_idx", "blocks", "bool", "class", "config", "def", "depths", "drop_path", "drop_path_rate", "drop_paths", "else", "for", "forward", "hidden_sizes", "hidden_states", "if", "in", "isinstance", "layer_module", "nn", "num_meta3d_blocks", "output_attentions", "outputs", "range", "return", "self", "sum", "super", "torch", "tuple"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerMeta4D": ["Identity", "ModelFormerConfig", "ModelFormerConvMlp", "ModelFormerDropPath", "ModelFormerMeta4D", "ModelFormerPooling", "Module", "None", "Parameter", "Tensor", "True", "__init__", "class", "config", "def", "dim", "drop", "drop_path", "else", "float", "forward", "hidden_dropout_prob", "hidden_features", "hidden_states", "if", "in_features", "int", "is", "layer_output", "layer_scale_1", "layer_scale_2", "layer_scale_init_value", "mlp", "mlp_expansion_ratio", "mlp_hidden_dim", "nn", "not", "ones", "outputs", "pool_size", "requires_grad", "return", "self", "super", "token_mixer", "torch", "tuple", "unsqueeze", "use_layer_scale"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerMeta4DLayers": ["ModelFormerConfig", "ModelFormerMeta4D", "ModelFormerMeta4DLayers", "Module", "ModuleList", "Tensor", "__init__", "block_idx", "blocks", "class", "config", "def", "depths", "drop_path", "drop_path_rate", "drop_paths", "else", "for", "forward", "hidden_sizes", "hidden_states", "if", "in", "int", "layer_module", "nn", "num_layers", "num_meta3d_blocks", "range", "return", "self", "stage_idx", "sum", "super", "torch", "tuple"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerIntermediateStage": ["ModelFormerConfig", "ModelFormerIntermediateStage", "ModelFormerMeta4DLayers", "Module", "Tensor", "__init__", "class", "config", "def", "forward", "hidden_states", "index", "int", "meta4D_layers", "nn", "return", "self", "super", "torch", "tuple"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerLastStage": ["False", "ModelFormerConfig", "ModelFormerFlat", "ModelFormerLastStage", "ModelFormerMeta3DLayers", "ModelFormerMeta4DLayers", "Module", "Tensor", "__init__", "bool", "class", "config", "def", "flat", "forward", "hidden_states", "meta3D_layers", "meta4D_layers", "nn", "output_attentions", "return", "self", "super", "torch", "tuple"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerEncoder": ["BaseModelOutput", "False", "ModelFormerConfig", "ModelFormerEncoder", "ModelFormerIntermediateStage", "ModelFormerLastStage", "ModelFormerPatchEmbeddings", "Module", "ModuleList", "None", "Tensor", "True", "__init__", "all_hidden_states", "all_self_attentions", "append", "attentions", "bool", "class", "config", "def", "depths", "downsamples", "else", "for", "forward", "hidden_sizes", "hidden_states", "i", "if", "in", "intermediate_stages", "is", "last_hidden_state", "last_stage", "layer_module", "layer_output", "len", "nn", "not", "num_intermediate_stages", "or", "output_attentions", "output_hidden_states", "range", "return", "return_dict", "self", "super", "torch", "tuple", "v"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerPreTrainedModel": ["Conv2d", "False", "LayerNorm", "Linear", "ModelFormerConfig", "ModelFormerPreTrainedModel", "Modelformer", "Module", "None", "PreTrainedModel", "_init_weights", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "fill_", "if", "initializer_range", "is", "isinstance", "main_input_name", "mean", "module", "nn", "normal_", "not", "pixel_values", "self", "std", "supports_gradient_checkpointing", "weight", "zero_"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerModel": ["BaseModelOutput", "BaseModelOutputWithPooling", "LayerNorm", "ModelFORMER_INPUTS_DOCSTRING", "ModelFormerConfig", "ModelFormerConvStem", "ModelFormerEncoder", "ModelFormerMeta4D", "ModelFormerModel", "ModelFormerPreTrainedModel", "None", "Optional", "Tensor", "Union", "ValueError", "You", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_EXPECTED_OUTPUT_SHAPE", "__init__", "_no_split_modules", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attentions", "bool", "checkpoint", "class", "config", "config_class", "def", "else", "embedding_output", "encoder", "encoder_outputs", "eps", "expected_output", "forward", "have", "head_outputs", "hidden_sizes", "hidden_states", "if", "is", "last_hidden_state", "layer_norm_eps", "layernorm", "modality", "nn", "not", "output_attentions", "output_hidden_states", "output_type", "patch_embed", "pixel_values", "post_init", "raise", "return", "return_dict", "self", "sequence_output", "specify", "super", "to", "torch", "tuple", "use_return_dict", "vision"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerForImageClassification": ["Identity", "ImageClassifierOutput", "Linear", "ModelFORMER_INPUTS_DOCSTRING", "ModelFormerConfig", "ModelFormerForImageClassification", "ModelFormerModel", "ModelFormerPreTrainedModel", "Modelformer", "None", "Optional", "Tensor", "Union", "_CONFIG_FOR_DOC", "_IMAGE_CLASS_CHECKPOINT", "_IMAGE_CLASS_EXPECTED_OUTPUT", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attentions", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "else", "expected_output", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "mean", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pixel_values", "post_init", "r", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerForImageClassificationWithTeacherOutput": ["FloatTensor", "ModelFormerForImageClassificationWithTeacherOutput", "ModelOutput", "None", "Optional", "attentions", "class", "cls_logits", "distillation_logits", "hidden_states", "logits", "torch", "tuple"], "deprecated/efficientformer/modeling_efficientformer.py:EfficientFormerForImageClassificationWithTeacher": ["Identity", "Linear", "ModelFORMER_INPUTS_DOCSTRING", "ModelFormerConfig", "ModelFormerForImageClassificationWithTeacher", "ModelFormerForImageClassificationWithTeacherOutput", "ModelFormerModel", "ModelFormerPreTrainedModel", "Modelformer", "None", "Optional", "Tensor", "Union", "_CONFIG_FOR_DOC", "_IMAGE_CLASS_CHECKPOINT", "_IMAGE_CLASS_EXPECTED_OUTPUT", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "attentions", "bool", "checkpoint", "class", "classifier", "cls_logits", "config", "config_class", "def", "distillation_classifier", "distillation_logits", "else", "expected_output", "forward", "hidden_size", "hidden_states", "if", "is", "logits", "mean", "nn", "not", "num_labels", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "pixel_values", "post_init", "return", "return_dict", "self", "sequence_output", "super", "torch", "tuple", "use_return_dict"], "deprecated/van/modeling_van.py:drop_path": ["False", "Model_path", "Model_prob", "Tensor", "bool", "def", "device", "div", "dtype", "float", "floor_", "if", "input", "keep_prob", "ndim", "not", "or", "output", "rand", "random_tensor", "return", "shape", "torch", "training"], "deprecated/van/modeling_van.py:VanDropPath": ["ModelDropPath", "Module", "None", "Optional", "Tensor", "__init__", "class", "def", "drop_path", "drop_prob", "extra_repr", "f", "float", "forward", "hidden_states", "nn", "p", "return", "self", "str", "super", "torch", "training"], "deprecated/van/modeling_van.py:VanOverlappingPatchEmbedder": ["BatchNorm2d", "Conv2d", "ModelOverlappingPatchEmbedder", "Module", "Tensor", "__init__", "class", "convolution", "def", "forward", "hidden_size", "hidden_state", "in_channels", "input", "int", "kernel_size", "nn", "normalization", "padding", "patch_size", "return", "self", "stride", "super", "torch"], "deprecated/van/modeling_van.py:VanMlpLayer": ["ACT2FN", "Conv2d", "Dropout", "ModelMlpLayer", "Module", "Tensor", "__init__", "activation", "class", "def", "depth_wise", "dropout1", "dropout2", "dropout_rate", "float", "forward", "gelu", "groups", "hidden_act", "hidden_size", "hidden_state", "in_channels", "in_dense", "int", "kernel_size", "nn", "out_channels", "out_dense", "padding", "return", "self", "str", "super", "torch"], "deprecated/van/modeling_van.py:VanLargeKernelAttention": ["Conv2d", "ModelLargeKernelAttention", "Module", "Tensor", "__init__", "class", "def", "depth_wise", "depth_wise_dilated", "dilation", "forward", "groups", "hidden_size", "hidden_state", "int", "kernel_size", "nn", "padding", "point_wise", "return", "self", "super", "torch"], "deprecated/van/modeling_van.py:VanLargeKernelAttentionLayer": ["ModelLargeKernelAttention", "ModelLargeKernelAttentionLayer", "Module", "Tensor", "__init__", "attended", "attention", "class", "def", "forward", "hidden_size", "hidden_state", "int", "nn", "return", "self", "super", "torch"], "deprecated/van/modeling_van.py:VanSpatialAttentionLayer": ["ACT2FN", "Conv2d", "ModelLargeKernelAttentionLayer", "ModelSpatialAttentionLayer", "Module", "OrderedDict", "Sequential", "Tensor", "__init__", "act", "attention_layer", "class", "conv", "def", "forward", "gelu", "hidden_act", "hidden_size", "hidden_state", "int", "kernel_size", "nn", "post_projection", "pre_projection", "residual", "return", "self", "str", "super", "torch"], "deprecated/van/modeling_van.py:VanLayerScaling": ["ModelLayerScaling", "Module", "Parameter", "Tensor", "True", "__init__", "class", "def", "float", "forward", "hidden_size", "hidden_state", "initial_value", "int", "nn", "ones", "requires_grad", "return", "self", "super", "torch", "unsqueeze", "weight"], "deprecated/van/modeling_van.py:VanLayer": ["BatchNorm2d", "Identity", "ModelConfig", "ModelDropPath", "ModelLayer", "ModelLayerScaling", "ModelMlpLayer", "ModelSpatialAttentionLayer", "Module", "Tensor", "__init__", "attention", "attention_scaling", "class", "config", "def", "drop_path", "drop_path_rate", "dropout_rate", "else", "float", "forward", "hidden_act", "hidden_size", "hidden_state", "if", "int", "layer_scale_init_value", "mlp", "mlp_ratio", "mlp_scaling", "nn", "post_normalization", "pre_normomalization", "residual", "return", "self", "super", "torch"], "deprecated/van/modeling_van.py:VanStage": ["LayerNorm", "ModelConfig", "ModelLayer", "ModelOverlappingPatchEmbedder", "ModelStage", "Module", "Sequential", "Tensor", "_", "__init__", "batch_size", "class", "config", "def", "depth", "drop_path_rate", "embeddings", "eps", "flatten", "float", "for", "forward", "height", "hidden_size", "hidden_state", "in", "in_channels", "int", "layer_norm_eps", "layers", "mlp_ratio", "nn", "normalization", "patch_size", "permute", "range", "return", "self", "shape", "stride", "super", "torch", "transpose", "view", "width"], "deprecated/van/modeling_van.py:VanEncoder": ["BaseModelOutputWithNoAttention", "False", "ModelConfig", "ModelEncoder", "ModelStage", "Module", "ModuleList", "None", "Optional", "Tensor", "True", "Union", "_", "__init__", "all_hidden_states", "append", "bool", "class", "config", "cpu", "def", "depth", "depths", "device", "drop_path_rate", "drop_path_rates", "else", "enumerate", "for", "forward", "hidden_size", "hidden_sizes", "hidden_state", "hidden_states", "if", "in", "in_channels", "is", "is_first_stage", "item", "last_hidden_state", "linspace", "mlp_expansion", "mlp_ratio", "mlp_ratios", "nn", "not", "num_channels", "num_stage", "output_hidden_states", "patch_size", "patch_sizes", "return", "return_dict", "self", "stage_module", "stages", "stride", "strides", "sum", "super", "torch", "tuple", "v", "x", "zip"], "deprecated/van/modeling_van.py:VanPreTrainedModel": ["Conv2d", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "and", "base_model_prefix", "bias", "class", "config", "constant_", "data", "def", "elif", "fan_out", "groups", "if", "init", "initializer_range", "is", "isinstance", "kernel_size", "main_input_name", "math", "module", "nn", "normal_", "not", "out_channels", "pixel_values", "self", "sqrt", "std", "supports_gradient_checkpointing", "trunc_normal_", "weight", "zero_"], "deprecated/van/modeling_van.py:VanModel": ["BaseModelOutputWithPoolingAndNoAttention", "FloatTensor", "LayerNorm", "ModelEncoder", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Union", "_CHECKPOINT_FOR_DOC", "_CONFIG_FOR_DOC", "_EXPECTED_OUTPUT_SHAPE", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "bool", "checkpoint", "class", "config", "config_class", "def", "dim", "else", "encoder", "encoder_outputs", "eps", "expected_output", "forward", "hidden_sizes", "hidden_states", "if", "is", "last_hidden_state", "layer_norm_eps", "layernorm", "mean", "modality", "nn", "not", "output_hidden_states", "output_type", "pixel_values", "pooled_output", "pooler_output", "post_init", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict", "vision"], "deprecated/van/modeling_van.py:VanForImageClassification": ["FloatTensor", "Identity", "ImageClassifierOutputWithNoAttention", "Linear", "LongTensor", "Model", "ModelForImageClassification", "ModelModel", "ModelPreTrainedModel", "Model_INPUTS_DOCSTRING", "None", "Optional", "Union", "_CONFIG_FOR_DOC", "_IMAGE_CLASS_CHECKPOINT", "_IMAGE_CLASS_EXPECTED_OUTPUT", "__init__", "add_code_sample_docstrings", "add_start_docstrings_to_model_forward", "bool", "checkpoint", "class", "classifier", "config", "config_class", "def", "else", "expected_output", "forward", "hidden_sizes", "hidden_states", "if", "is", "labels", "logits", "loss", "loss_function", "nn", "not", "num_labels", "output", "output_hidden_states", "output_type", "outputs", "pixel_values", "pooled_output", "pooler_output", "post_init", "r", "return", "return_dict", "self", "super", "torch", "tuple", "use_return_dict"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaRMSNorm": ["ModelLlamaRMSNorm", "Module", "Parameter", "True", "__init__", "class", "def", "dtype", "eps", "extra_repr", "f", "float32", "forward", "hidden_size", "hidden_states", "input_dtype", "keepdim", "mean", "nn", "ones", "pow", "return", "rsqrt", "self", "shape", "super", "to", "torch", "tuple", "variance", "variance_epsilon", "weight"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaRotaryEmbedding": ["False", "ModelLlamaRotaryEmbedding", "Module", "None", "Tensor", "__init__", "_set_cos_sin_cache", "arange", "base", "cat", "class", "cos", "cos_cached", "def", "device", "dim", "dtype", "emb", "float", "forward", "freqs", "get_default_dtype", "if", "int64", "inv_freq", "max_position_embeddings", "max_seq_len_cached", "nn", "outer", "persistent", "register_buffer", "return", "self", "seq_len", "sin", "sin_cached", "super", "t", "to", "torch", "type_as", "x"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaLinearScalingRotaryEmbedding": ["False", "ModelLlamaLinearScalingRotaryEmbedding", "ModelLlamaRotaryEmbedding", "None", "__init__", "_set_cos_sin_cache", "arange", "base", "cat", "class", "cos", "cos_cached", "def", "device", "dim", "dtype", "emb", "freqs", "int64", "inv_freq", "max_position_embeddings", "max_seq_len_cached", "outer", "persistent", "register_buffer", "scaling_factor", "self", "seq_len", "sin", "sin_cached", "super", "t", "to", "torch", "type_as"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaDynamicNTKScalingRotaryEmbedding": ["False", "ModelLlamaDynamicNTKScalingRotaryEmbedding", "ModelLlamaRotaryEmbedding", "None", "__init__", "_set_cos_sin_cache", "arange", "base", "cat", "class", "cos", "cos_cached", "def", "device", "dim", "dtype", "emb", "float", "freqs", "if", "int64", "inv_freq", "max_position_embeddings", "max_seq_len_cached", "outer", "persistent", "register_buffer", "scaling_factor", "self", "seq_len", "sin", "sin_cached", "super", "t", "to", "torch", "type_as"], "deprecated/open_llama/modeling_open_llama.py:rotate_half": ["Model_half", "cat", "def", "dim", "return", "shape", "torch", "x", "x1", "x2"], "deprecated/open_llama/modeling_open_llama.py:apply_rotary_pos_emb": ["Model_rotary_pos_emb", "cos", "def", "k", "k_embed", "position_ids", "q", "q_embed", "return", "rotate_half", "sin", "unsqueeze", "unsqueeze_dim"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaMLP": ["ACT2FN", "Dropout", "False", "Linear", "ModelLlamaMLP", "Module", "__init__", "act_fn", "bias", "class", "def", "down_proj", "dropout", "dropout_prob", "float", "forward", "gate_proj", "hidden_act", "hidden_size", "int", "intermediate_size", "nn", "out", "return", "self", "str", "super", "up_proj", "x"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaAttention": ["Attention", "Cache", "False", "Linear", "LongTensor", "LowerTriangularMask", "ModelLlamaAttention", "ModelLlamaConfig", "ModelLlamaDynamicNTKScalingRotaryEmbedding", "ModelLlamaLinearScalingRotaryEmbedding", "ModelLlamaRotaryEmbedding", "Module", "None", "Optional", "RoPE", "Tensor", "Unknown", "ValueError", "_", "__init__", "_init_rope", "_shape", "and", "apply_rotary_pos_emb", "attention_dropout_prob", "attention_mask", "attn_bias", "attn_output", "attn_weights", "base", "be", "bias", "bool", "bsz", "but", "by", "cat", "class", "config", "contiguous", "cos", "def", "deprecate_kwarg", "device", "dim", "divisible", "dropout_prob", "dtype", "dynamic", "elif", "else", "f", "factor", "finfo", "float32", "forward", "functional", "got", "head_dim", "hidden_size", "hidden_states", "if", "int", "is", "k_proj", "key_states", "kv_seq_len", "linear", "mask", "math", "matmul", "max", "max_position_embeddings", "memory_efficient_attention", "min", "must", "new_name", "nn", "not", "num_attention_heads", "num_heads", "o_proj", "of", "output_attentions", "p", "past_key_value", "past_key_values", "position_ids", "q_len", "q_proj", "query_states", "raise", "reshape", "return", "rope_scaling", "rope_theta", "rotary_emb", "scaling", "scaling_factor", "scaling_type", "self", "seq_len", "shape", "should", "sin", "size", "softmax", "sqrt", "super", "tensor", "to", "torch", "training", "transpose", "tuple", "type", "use_cache", "use_memory_efficient_attention", "v_proj", "value_states", "version", "view", "weights", "xops"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaDecoderLayer": ["Cache", "False", "FloatTensor", "GradientCheckpointingLayer", "LongTensor", "ModelLlamaAttention", "ModelLlamaConfig", "ModelLlamaDecoderLayer", "ModelLlamaMLP", "ModelLlamaRMSNorm", "None", "Optional", "Tensor", "__init__", "attention_mask", "bool", "class", "config", "def", "deprecate_kwarg", "dropout_prob", "eps", "forward", "hidden_act", "hidden_dropout_prob", "hidden_size", "hidden_states", "if", "input_layernorm", "intermediate_size", "mlp", "new_name", "output_attentions", "outputs", "past_key_value", "past_key_values", "position_ids", "post_attention_layernorm", "present_key_value", "residual", "return", "rms_norm_eps", "self", "self_attn", "self_attn_weights", "super", "torch", "tuple", "use_cache", "version"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaPreTrainedModel": ["Embedding", "Linear", "ModelLlamaConfig", "ModelLlamaDecoderLayer", "ModelLlamaPreTrainedModel", "None", "PreTrainedModel", "True", "_init_weights", "_no_split_modules", "base_model_prefix", "bias", "class", "config", "data", "def", "elif", "else", "if", "init", "initializer_range", "is", "isinstance", "mean", "model", "module", "nn", "normal_", "not", "padding_idx", "self", "std", "supports_gradient_checkpointing", "torch", "use_stable_embedding", "weight", "xavier_normal_", "zero_"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaModel": ["BaseModelOutputWithPast", "Cache", "Embedding", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelLlamaConfig", "ModelLlamaDecoderLayer", "ModelLlamaModel", "ModelLlamaPreTrainedModel", "ModelLlamaRMSNorm", "Model_LLAMA_INPUTS_DOCSTRING", "ModuleList", "None", "Optional", "Setting", "Tensor", "True", "Union", "ValueError", "You", "_", "__init__", "_prepare_4d_causal_attention_mask", "add_start_docstrings_to_model_forward", "all_hidden_states", "all_self_attns", "and", "arange", "at", "attention_mask", "attentions", "batch_size", "bool", "both", "cannot", "checkpointing", "class", "config", "decoder_input_ids", "decoder_inputs_embeds", "decoder_layer", "def", "device", "dtype", "either", "elif", "else", "embed_layer_norm", "embed_tokens", "enumerate", "eps", "for", "forward", "get_seq_length", "gradient", "gradient_checkpointing", "have", "hidden_size", "hidden_states", "idx", "if", "in", "incompatible", "input_ids", "input_shape", "inputs_embeds", "is", "last_hidden_state", "layer_outputs", "layers", "logger", "long", "next_cache", "next_decoder_cache", "nn", "norm", "not", "num_hidden_layers", "ones", "or", "output_attentions", "output_hidden_states", "pad_token_id", "padding_idx", "past_key_values", "past_key_values_length", "position_ids", "post_init", "raise", "range", "return", "return_dict", "rms_norm_eps", "same", "self", "seq_length", "seq_length_with_past", "shape", "specify", "super", "the", "time", "to", "torch", "training", "tuple", "unsqueeze", "use_cache", "use_memory_efficient_attention", "use_return_dict", "use_stable_embedding", "v", "vocab_size", "warning_once", "with"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaForCausalLM": ["Cache", "CausalLMOutputWithPast", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "ModelLlamaForCausalLM", "ModelLlamaModel", "ModelLlamaPreTrainedModel", "Model_LLAMA_INPUTS_DOCSTRING", "None", "Optional", "Tensor", "Union", "_CONFIG_FOR_DOC", "__init__", "_reorder_cache", "add_start_docstrings_to_model_forward", "and", "attention_mask", "attentions", "beam_idx", "bias", "blh", "blv", "bool", "class", "config", "config_class", "contiguous", "cumsum", "def", "device", "einsum", "else", "embed_tokens", "for", "forward", "get", "get_seq_length", "hidden_size", "hidden_states", "if", "in", "index_select", "input_ids", "inputs_embeds", "is", "kwargs", "labels", "layer_past", "lm_head", "logits", "long", "loss", "loss_fct", "masked_fill_", "model", "model_inputs", "nn", "not", "output", "output_attentions", "output_hidden_states", "output_type", "outputs", "past_key_values", "past_length", "past_state", "position_ids", "post_init", "prepare_inputs_for_generation", "r", "remove_prefix_length", "reordered_past", "replace_return_docstrings", "return", "return_dict", "self", "shape", "shared_input_output_embedding", "shift_labels", "shift_logits", "staticmethod", "super", "to", "torch", "tuple", "update", "use_cache", "use_return_dict", "vh", "view", "vocab_size", "weight"], "deprecated/open_llama/modeling_open_llama.py:OpenLlamaForSequenceClassification": ["BCEWithLogitsLoss", "Cache", "Cannot", "CrossEntropyLoss", "False", "FloatTensor", "Linear", "LongTensor", "MSELoss", "ModelLlamaForSequenceClassification", "ModelLlamaModel", "ModelLlamaPreTrainedModel", "Model_LLAMA_INPUTS_DOCSTRING", "None", "Optional", "SequenceClassifierOutputWithPast", "Tensor", "Union", "ValueError", "__init__", "add_start_docstrings_to_model_forward", "and", "arange", "argmax", "attention_mask", "attentions", "batch", "batch_size", "bias", "bool", "class", "config", "def", "defined", "device", "dtype", "elif", "else", "eq", "forward", "handle", "hidden_size", "hidden_states", "if", "input_ids", "inputs_embeds", "int", "is", "labels", "logits", "long", "loss", "loss_fct", "model", "multi_label_classification", "nn", "no", "not", "num_labels", "or", "output", "output_attentions", "output_hidden_states", "pad_token_id", "padding", "past_key_values", "pooled_logits", "position_ids", "post_init", "problem_type", "r", "raise", "regression", "return", "return_dict", "score", "self", "sequence_lengths", "shape", "single_label_classification", "sizes", "squeeze", "super", "to", "token", "torch", "transformer_outputs", "tuple", "use_cache", "use_return_dict", "view"], "deprecated/trajectory_transformer/modeling_trajectory_transformer.py:TrajectoryTransformerOutput": ["Cache", "FloatTensor", "ModelOutput", "ModelTransformerOutput", "None", "Optional", "attentions", "class", "hidden_states", "logits", "loss", "past_key_values", "torch", "tuple"], "deprecated/trajectory_transformer/modeling_trajectory_transformer.py:TrajectoryTransformerPreTrainedModel": ["EinLinear", "Embedding", "LayerNorm", "Linear", "ModelTransformerConfig", "ModelTransformerPreTrainedModel", "Model_transformer", "None", "PreTrainedModel", "True", "_", "_calculate_fan_in_and_fan_out", "_init_weights", "a", "and", "base_model_prefix", "bias", "bound", "class", "config", "data", "def", "elif", "fan_in", "fill_", "for", "i", "if", "in", "init", "initializer_range", "is", "isinstance", "kaiming_initializer_range", "kaiming_uniform_", "main_input_name", "math", "mean", "module", "n_models", "nn", "normal_", "not", "range", "self", "sqrt", "std", "supports_gradient_checkpointing", "trajectories", "uniform_", "weight", "zero_"], "deprecated/trajectory_transformer/modeling_trajectory_transformer.py:EinLinear": ["ModelLinear", "Modelsum", "Module", "None", "Parameter", "RuntimeError", "Tensor", "_", "__init__", "_calculate_fan_in_and_fan_out", "a", "bei", "beo", "bias", "bound", "class", "def", "else", "eoi", "fan_in", "for", "forward", "i", "if", "in", "in_features", "init", "input", "is", "kaiming_uniform_", "math", "n_models", "nn", "not", "out_features", "output", "raise", "range", "register_parameter", "reset_parameters", "return", "self", "sqrt", "super", "torch", "uniform_", "weight"], "deprecated/trajectory_transformer/modeling_trajectory_transformer.py:CausalSelfAttention": ["Dropout", "F", "False", "FloatTensor", "Linear", "ModelSelfAttention", "Module", "None", "Optional", "Tensor", "True", "ValueError", "__init__", "_attn_map", "a", "action_dim", "attn_drop", "attn_pdrop", "attn_weights", "batch_size", "be", "block_size", "bool", "cat", "class", "clone", "config", "contiguous", "def", "dim", "divisor", "dtype", "else", "embedding_dim", "f", "finfo", "forward", "hidden_states", "if", "is", "joined_dim", "key", "layer_past", "mask", "masked_fill", "math", "matmul", "min", "n_embd", "n_head", "nn", "not", "observation_dim", "of", "ones", "output", "output_attentions", "outputs", "past_key", "past_value", "persistent", "present", "proj", "query", "raise", "register_buffer", "resid_drop", "resid_pdrop", "return", "self", "sequence_length", "should", "size", "softmax", "sqrt", "squeeze", "super", "torch", "transpose", "tril", "tuple", "use_cache", "value", "view"], "deprecated/trajectory_transformer/modeling_trajectory_transformer.py:Block": ["CausalSelfAttention", "Dropout", "False", "FloatTensor", "GELU", "GradientCheckpointingLayer", "LayerNorm", "Linear", "Model", "None", "Optional", "Tensor", "__init__", "act", "attn", "attn_output", "attn_outputs", "bool", "class", "config", "def", "drop", "else", "forward", "hidden_states", "if", "l1", "l2", "layer_past", "ln1", "ln2", "n_embd", "nn", "output_attentions", "outputs", "resid_pdrop", "residual", "return", "self", "super", "torch", "tuple", "use_cache"], "deprecated/trajectory_transformer/modeling_trajectory_transformer.py:TrajectoryTransformerModel": ["Block", "Cache", "Cannot", "Dropout", "EinLinear", "Embedding", "F", "False", "FloatTensor", "LayerNorm", "LongTensor", "ModelTransformerModel", "ModelTransformerOutput", "ModelTransformerPreTrainedModel", "Model_TRANSFORMER_INPUTS_DOCSTRING", "ModuleList", "None", "Optional", "Parameter", "Setting", "Tensor", "True", "Union", "ValueError", "_", "_CONFIG_FOR_DOC", "__init__", "action_dim", "action_weight", "add_start_docstrings_to_model_forward", "all_hidden_states", "all_self_attentions", "and", "arange", "attention_mask", "attentions", "batch_size", "bias", "block", "block_size", "blocks", "bool", "cat", "ceil", "checkpointing", "class", "config", "config_class", "cross_entropy", "def", "device", "dim", "drop", "else", "embd_pdrop", "embedding_dim", "enumerate", "exhausted", "for", "format", "forward", "get_block_size", "gradient", "gradient_checkpointing", "head", "hidden_state", "hidden_states", "hidden_states_pad", "i", "if", "in", "incompatible", "int", "is", "layer_past", "len", "ln_f", "logger", "logits", "loss", "mean", "model", "n_embd", "n_layer", "n_pad", "n_states", "nn", "none", "not", "np", "observation_dim", "offset_tokens", "offset_trajectories", "offsets", "ones", "or", "output_attentions", "output_hidden_states", "output_type", "outputs", "pad_to_full_observation", "padding", "past_key_values", "pos_emb", "position_embeddings", "post_init", "presents", "r", "raise", "range", "reduction", "repeat", "replace_return_docstrings", "reshape", "return", "return_dict", "reward_weight", "self", "sequence_length", "shape", "size", "stop_token", "super", "targets", "to", "tok_emb", "token_embeddings", "torch", "training", "trajectories", "transition_dim", "tuple", "use_cache", "v", "value_weight", "view", "vocab_size", "warning_once", "weights", "with", "zeros", "zip"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:router_z_loss_func": ["Model_logits", "Model_z_loss_func", "Tensor", "_", "def", "dim", "float", "log_z", "logsumexp", "num_groups", "r", "return", "shape", "sum", "tokens_per_group", "torch", "z_loss"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:load_balancing_loss_func": ["Model_balancing_loss_func", "Tensor", "axis", "def", "dtype", "expert_indices", "expert_mask", "float", "float32", "functional", "if", "int64", "len", "max", "mean", "nn", "num_experts", "one_hot", "r", "return", "router_prob_per_group_and_expert", "router_probs", "shape", "to", "tokens_per_group_and_expert", "torch", "unsqueeze", "values"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseDenseActDense": ["ACT2FN", "Dropout", "False", "GPTSanJapaneseConfig", "Identity", "Linear", "Model", "Module", "__init__", "act", "bias", "class", "config", "d_ext", "d_ff", "d_inter", "d_model", "def", "dropout", "dropout_rate", "else", "ext_layer", "forward", "hidden_states", "if", "nn", "r", "relu", "return", "self", "super", "swish", "wi", "wo"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseTop1Router": ["CB", "GPTSanJapaneseConfig", "Linear", "Model", "Module", "SCB", "Tensor", "__init__", "_cast_classifier", "_compute_router_probabilities", "and", "argmax", "bias", "class", "classifier", "config", "cumsum", "def", "dim", "dtype", "empty_like", "expert_capacity", "expert_capacity_mask", "expert_index", "forward", "functional", "getattr", "hasattr", "hidden_size", "hidden_states", "if", "ignore_padding_tokens", "input_dtype", "jitter_noise", "max", "nn", "not", "num_classes", "num_experts", "one_hot", "or", "r", "return", "router_bias", "router_dtype", "router_ignore_padding_tokens", "router_jitter_noise", "router_logits", "router_probabilities", "router_probs", "self", "softmax", "super", "to", "token_priority", "torch", "training", "tuple", "uniform_", "unsqueeze", "values"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseSparseMLP": ["GPTSanJapaneseConfig", "GPTSanJapaneseDenseActDense", "GPTSanJapaneseTop1Router", "Model", "Module", "ModuleDict", "__init__", "argmax", "bool", "class", "clone", "config", "def", "dim", "dtype", "enumerate", "expert", "expert_", "expert_class", "expert_index", "experts", "f", "for", "forward", "hidden_states", "idx", "in", "next_states", "nn", "num_experts", "r", "range", "return", "router", "router_logits", "router_mask", "router_probs", "self", "super", "to", "token_indices", "torch", "values"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseLayerSparseFF": ["False", "GPTSanJapaneseConfig", "GPTSanJapaneseSparseMLP", "LayerNorm", "Linear", "Model", "Module", "None", "__init__", "and", "bias", "class", "config", "d_model", "def", "else", "eps", "forward", "forwarded_states", "hidden_states", "if", "is", "layer_norm_epsilon", "mlp", "nn", "norm", "not", "output", "output_router_logits", "r", "return", "router_tuple", "self", "soft_bypass_mlp", "super", "tanh", "torch"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseLayerDenseFF": ["GPTSanJapaneseConfig", "GPTSanJapaneseDenseActDense", "LayerNorm", "Model", "Module", "True", "__init__", "class", "config", "d_model", "def", "eps", "ext_layer", "forward", "forwarded_states", "hidden_states", "layer_norm_epsilon", "mlp", "nn", "norm", "output", "r", "return", "self", "super"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseAttention": ["Attention", "Cache", "False", "GPTSanJapaneseConfig", "Head", "Linear", "Model", "Module", "None", "Optional", "Tensor", "True", "ValueError", "_", "__init__", "_shape", "a", "and", "attention_mask", "attn_output", "attn_probs", "attn_weights", "attn_weights_reshaped", "be", "bias", "bmm", "bool", "bsz", "but", "by", "cat", "class", "config", "contiguous", "def", "deprecate_kwarg", "dim", "divisible", "dropout", "elif", "else", "embed_dim", "f", "float", "for", "forward", "functional", "got", "head_dim", "hidden_states", "if", "int", "is", "is_causal", "is_cross_attention", "is_decoder", "k_proj", "key_states", "key_value_states", "layer", "layer_head_mask", "mask", "must", "new_name", "nn", "not", "num_heads", "of", "out_proj", "output_attentions", "p", "past_key_value", "past_key_values", "proj_shape", "q_proj", "query_states", "raise", "reshape", "return", "scaling", "self", "seq_len", "shape", "should", "single", "size", "softmax", "src_len", "super", "tensor", "tgt_len", "torch", "training", "transpose", "tuple", "v_proj", "value_states", "version", "view", "weights"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseLayerSelfAttention": ["Cache", "False", "FloatTensor", "GPTSanJapaneseAttention", "LayerNorm", "Model", "Module", "None", "Optional", "Tensor", "True", "Union", "__init__", "atten_out", "attention_mask", "attention_output", "attn_weights", "bias", "bool", "class", "config", "d_model", "def", "deprecate_kwarg", "dtype", "else", "embed_dim", "eps", "finfo", "forward", "has_relative_attention_bias", "head_mask", "hidden", "hidden_states", "if", "is", "is_decoder", "layer_head_mask", "layer_norm_epsilon", "min", "new_name", "nn", "norm", "not", "num_heads", "output_attentions", "outputs", "past_key_value", "past_key_values", "r", "return", "self", "self_attn", "self_attn_past_key_value", "super", "torch", "tuple", "use_cache", "version"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseBlock": ["Cache", "False", "FloatTensor", "GPTSanJapaneseLayerDenseFF", "GPTSanJapaneseLayerSelfAttention", "GPTSanJapaneseLayerSparseFF", "Model", "Module", "None", "Optional", "Tensor", "Union", "__init__", "and", "atten_out", "attention_mask", "attention_output", "bool", "class", "config", "def", "deprecate_kwarg", "else", "ext_layer", "feed_forward", "forward", "head_mask", "hidden", "hidden_states", "if", "isinstance", "new_name", "nn", "output_attentions", "output_router_tuple", "outputs", "past_key_value", "past_key_values", "r", "return", "router_tuple", "self", "self_attn", "sparse_out", "super", "torch", "tuple", "use_cache", "version"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapanesePreTrainedModel": ["DUMMY_INPUTS", "DUMMY_MASK", "Embedding", "False", "GPTSanJapaneseAttention", "GPTSanJapaneseBlock", "GPTSanJapaneseConfig", "GPTSanJapaneseDenseActDense", "GPTSanJapaneseForConditionalGeneration", "GPTSanJapaneseModel", "GPTSanJapaneseSparseMLP", "In", "LayerNorm", "Linear", "Model", "None", "PreTrainedModel", "See", "T5", "ValueError", "_init_weights", "_no_split_modules", "_shift_right", "_skip_keys_device_placement", "and", "attention_mask", "base_model_prefix", "be", "bias", "cat", "class", "classifier", "clone", "config", "d_ff", "d_model", "data", "decoder_start_token_id", "def", "defined", "dim", "docs", "dummy_inputs", "elif", "else", "embed_tokens", "expert_", "experts", "extra_position_embeddings", "f", "factor", "fill_", "final_logits_bias", "for", "full", "gptsan_japanese", "has", "hasattr", "idx", "if", "in", "information", "initializer_factor", "input_ids", "input_mask", "is", "is_torch_fx_proxy", "isinstance", "it", "k_proj", "key_value_proj_dim", "lm_head", "masked_fill_", "mean", "model", "module", "more", "n_heads", "new_zeros", "nn", "normal_", "not", "num_experts", "num_heads", "out_proj", "pad_token_id", "past_key_values", "position_embeddings", "property", "q_proj", "raise", "range", "return", "router", "self", "set", "shape", "shifted_input_ids", "std", "supports_gradient_checkpointing", "tensor", "the", "tie_word_embeddings", "to", "torch", "usually", "v_proj", "weight", "wi", "wo", "zero_"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseModel": ["ACT2FN", "Cache", "Embedding", "False", "FloatTensor", "GPTSAN_JAPANESE_INPUTS_DOCSTRING", "GPTSanJapaneseBlock", "GPTSanJapaneseConfig", "GPTSanJapanesePreTrainedModel", "Linear", "LongTensor", "Make", "MoEModelOutputWithPastAndCrossAttentions", "Model", "ModuleList", "None", "NotImplementedError", "Optional", "Sequential", "Tanh", "True", "Union", "ValueError", "_", "__init__", "a", "act", "add_start_docstrings_to_model_forward", "all_attentions", "all_hidden_states", "all_router_probs", "and", "append", "arange", "argmin", "attention_mask", "attention_mask_with_spout", "attention_probs", "attentions", "b", "batch", "be", "bias", "block_output", "blocks", "bool", "causal_mask", "class", "clip", "config", "d_model", "d_spout", "decoder_inputs_embeds", "def", "device", "dim", "does", "dtype", "elif", "else", "embed_tokens", "enumerate", "ext_layer", "extended_attention_mask", "extra_position_embeddings", "float", "for", "forward", "gather", "gather_position", "get_head_mask", "get_seq_length", "head_mask", "hidden_states", "i", "if", "in", "index", "input_ids", "inputs_embeds", "instead", "int", "is", "last_hidden_state", "last_project", "layer", "len", "long", "max_position_embeddings", "new_embeddings", "nn", "not", "num_batch", "num_ext_layers", "num_heads", "num_input_contexts", "num_layers", "num_output_contexts", "num_pasts_contexts", "num_precontext", "num_switch_layers", "ones", "or", "outpos", "output_attentions", "output_hidden_states", "output_router_logits", "output_router_tuple", "pass", "past", "past_key_values", "pasts_or_spout_value", "position_embeddings", "post_init", "prefix_lm_mask", "present", "present_key_value_states", "r", "raise", "range", "reshape", "return", "return_dict", "router_probs", "router_tuple", "self", "set_input_embeddings", "shape", "should", "size", "split", "spout", "spouts", "squeeze", "super", "sure", "swish", "to", "token_position", "token_type_ids", "torch", "transpose", "tril", "tuple", "uint8", "unsqueeze", "use", "use_cache", "use_return_dict", "v", "view", "vocab_size", "weight", "zeros"], "deprecated/gptsan_japanese/modeling_gptsan_japanese.py:GPTSanJapaneseForConditionalGeneration": ["Cache", "CrossEntropyLoss", "Embedding", "False", "FloatTensor", "GPTSAN_JAPANESE_INPUTS_DOCSTRING", "GPTSanJapaneseConfig", "GPTSanJapaneseModel", "GPTSanJapanesePreTrainedModel", "Linear", "LongTensor", "MoECausalLMOutputWithPast", "Model", "None", "Optional", "SEG_TOKEN", "Softmax", "Tensor", "True", "Union", "__init__", "_resize_final_logits_bias", "_shift_right", "_tied_weights_keys", "_unpack_router_logits", "add_start_docstrings_to_model_forward", "append", "attention_mask", "attentions", "aux_loss", "bias", "bool", "cat", "class", "config", "d_model", "decoder_inputs_embeds", "def", "device", "dim", "else", "embed_tokens", "expert_indexes", "extra_bias", "final_logits_bias", "float", "for", "forward", "get_input_embeddings", "head_mask", "hidden_states", "if", "ignore_index", "in", "input_ids", "inputs_embeds", "int", "is", "isinstance", "kwargs", "labels", "len", "list", "lm_head", "lm_logits", "load_balancing_loss_func", "logits", "loss", "loss_fct", "mean_resizing", "model", "model_return_dict", "new_bias", "new_embeddings", "new_num_tokens", "nn", "not", "num_batch", "num_precontext", "old_num_tokens", "or", "output_attentions", "output_hidden_states", "output_router_logits", "outputs", "pad_to_multiple_of", "past_key_values", "prepare_decoder_input_ids_from_labels", "prepare_inputs_for_generation", "r", "register_buffer", "resize_token_embeddings", "return", "return_dict", "router_logits", "router_output", "router_outputs", "router_probs", "router_z_loss_func", "self", "separator_token_id", "set_input_embeddings", "shape", "size", "spout", "super", "tensor", "to", "token_type_ids", "torch", "torchscript", "total_expert_indexes", "total_router_logits", "tuple", "unsqueeze", "use_cache", "use_return_dict", "v", "view", "vocab_size", "weight", "where", "where_separators", "z_loss", "zeros"], "deprecated/graphormer/modeling_graphormer.py:quant_noise": ["AssertionError", "Conv2d", "Embedding", "Input", "Kernel", "Linear", "Model_noise", "Module", "NotImplementedError", "_forward_pre_hook", "a", "be", "bernoulli_", "block", "block_size", "bool", "channels", "data", "def", "device", "else", "features", "float", "for", "if", "in_channels", "in_features", "input", "int", "is_conv", "isinstance", "k", "kernel_size", "mask", "masked_fill", "mod", "module", "multiple", "must", "ndim", "nn", "not", "of", "out_channels", "out_features", "p", "raise", "register_forward_pre_hook", "repeat", "repeat_interleave", "return", "s", "size", "sizes", "to", "torch", "training", "unsqueeze", "unsupported", "view", "weight", "zeros"], "deprecated/graphormer/modeling_graphormer.py:LayerDropModuleList": ["Iterable", "Iterator", "ModelDropModuleList", "Module", "ModuleList", "None", "Optional", "__init__", "__iter__", "class", "def", "dropout_probs", "empty", "enumerate", "float", "for", "i", "if", "in", "len", "m", "modules", "nn", "not", "or", "p", "self", "super", "torch", "training", "uniform_", "yield"], "deprecated/graphormer/modeling_graphormer.py:GraphormerGraphNodeFeature": ["Embedding", "LongTensor", "ModelConfig", "ModelGraphNodeFeature", "Module", "Tensor", "__init__", "atom_encoder", "cat", "class", "config", "def", "dim", "forward", "graph_node_feature", "graph_token", "graph_token_feature", "hidden_size", "in_degree", "in_degree_encoder", "input_nodes", "n_graph", "n_node", "nn", "node_feature", "num_atoms", "num_attention_heads", "num_heads", "num_in_degree", "num_out_degree", "out_degree", "out_degree_encoder", "pad_token_id", "padding_idx", "repeat", "return", "self", "size", "sum", "super", "torch", "unsqueeze", "weight"], "deprecated/graphormer/modeling_graphormer.py:GraphormerGraphAttnBias": ["Embedding", "LongTensor", "ModelConfig", "ModelGraphAttnBias", "Module", "Tensor", "__init__", "attn_bias", "attn_edge_type", "bmm", "clamp", "class", "clone", "config", "def", "edge_dis_encoder", "edge_encoder", "edge_input_flat", "edge_type", "else", "float", "forward", "graph_attn_bias", "graph_token_virtual_distance", "if", "input_edges", "input_nodes", "max_dist", "mean", "multi_hop", "multi_hop_max_dist", "n_graph", "n_node", "nn", "num_attention_heads", "num_edge_dis", "num_edges", "num_heads", "num_spatial", "padding_idx", "permute", "repeat", "reshape", "return", "self", "size", "spatial_pos", "spatial_pos_", "spatial_pos_bias", "spatial_pos_encoder", "sum", "super", "t", "torch", "unsqueeze", "view", "weight", "where"], "deprecated/graphormer/modeling_graphormer.py:GraphormerMultiheadAttention": ["AssertionError", "Dropout", "False", "Linear", "LongTensor", "Model", "ModelConfig", "ModelMultiheadAttention", "Module", "No", "None", "NotImplementedError", "Optional", "Query", "Self", "Tensor", "The", "True", "_", "__init__", "and", "apply_sparse_mask", "attention", "attention_dropout", "attention_dropout_module", "attn", "attn_bias", "attn_mask", "attn_probs", "attn_weights", "attn_weights_float", "batch", "be", "before_softmax", "bias", "bmm", "bool", "bsz", "by", "class", "compared", "config", "constant_", "contiguous", "def", "dim", "dimension", "dimensions", "divisible", "do", "does", "else", "embedding", "embedding_dim", "equal", "expected", "f", "float", "for", "forward", "functional", "gain", "generated", "head_dim", "if", "in", "incorrect", "inf", "init", "inplace", "int", "is", "is_scripting", "jit", "k", "k_proj", "kdim", "key", "key_bsz", "key_padding_mask", "list", "mask", "masked_fill", "match", "math", "mean", "model", "must", "need_head_weights", "need_weights", "nn", "not", "now", "num_attention_heads", "num_heads", "of", "only", "onnx_trace", "or", "out_proj", "p", "padding", "provided", "q", "q_noise", "q_proj", "qkv_same_dim", "qn_block_size", "quant_noise", "query", "raise", "requires", "reset_parameters", "return", "same", "scaling", "self", "self_attention", "shape", "shapes", "size", "softmax", "sqrt", "src_len", "super", "supports", "tgt_len", "the", "to", "torch", "transpose", "tuple", "type_as", "unsqueeze", "v", "v_proj", "value", "vdim", "view", "weight", "weights", "xavier_uniform_"], "deprecated/graphormer/modeling_graphormer.py:GraphormerGraphEncoderLayer": ["ACT2FN", "Conv2d", "Dropout", "Embedding", "False", "LayerNorm", "Linear", "ModelConfig", "ModelGraphEncoderLayer", "ModelMultiheadAttention", "Module", "None", "Optional", "Tensor", "Union", "__init__", "activation_dropout", "activation_dropout_module", "activation_fn", "attn", "attn_bias", "attn_mask", "build_fc", "class", "config", "def", "dropout", "dropout_module", "embedding_dim", "fc1", "fc2", "ffn_embedding_dim", "final_layer_norm", "float", "forward", "if", "inplace", "input_dim", "input_nodes", "int", "key", "key_padding_mask", "need_weights", "nn", "not", "num_attention_heads", "output_dim", "p", "pre_layernorm", "q_noise", "qn_block_size", "quant_noise", "query", "residual", "return", "self", "self_attn", "self_attn_bias", "self_attn_layer_norm", "self_attn_mask", "self_attn_padding_mask", "super", "torch", "tuple", "value"], "deprecated/graphormer/modeling_graphormer.py:GraphormerGraphEncoder": ["Dropout", "False", "Freezing", "LayerDropModuleList", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelGraphAttnBias", "ModelGraphEncoder", "ModelGraphEncoderLayer", "ModelGraphNodeFeature", "Module", "ModuleList", "None", "NotImplementedError", "Optional", "Tensor", "Union", "_", "__init__", "append", "apply_Model_init", "attn_bias", "attn_edge_type", "attn_mask", "bias", "bool", "cat", "class", "config", "data_x", "def", "device", "dim", "dropout", "dropout_module", "dtype", "else", "emb_layer_norm", "embed_scale", "embedding_dim", "embeddings", "encoder_normalize_before", "eq", "extend", "final_layer_norm", "for", "forward", "freeze_embeddings", "graph_attn_bias", "graph_node_feature", "graph_rep", "if", "implemented", "in", "in_degree", "inner_states", "inplace", "input_edges", "input_nodes", "is", "last_state_only", "layer", "layerdrop", "layers", "list", "m", "n_graph", "n_node", "nn", "not", "num_hidden_layers", "num_trans_layers_to_freeze", "out_degree", "p", "padding_mask", "padding_mask_cls", "parameters", "perturb", "pre_layernorm", "q_noise", "qn_block_size", "quant_noise", "raise", "range", "requires_grad", "return", "self", "self_attn_bias", "self_attn_mask", "self_attn_padding_mask", "size", "spatial_pos", "stack", "super", "token_embeddings", "torch", "traceable", "transpose", "tuple", "yet", "zeros"], "deprecated/graphormer/modeling_graphormer.py:GraphormerDecoderHead": ["False", "Linear", "ModelDecoderHead", "Module", "Parameter", "Tensor", "__init__", "bias", "class", "classifier", "def", "embedding_dim", "forward", "input_nodes", "int", "lm_output_learned_bias", "nn", "num_classes", "return", "self", "super", "torch", "unused", "zeros"], "deprecated/graphormer/modeling_graphormer.py:GraphormerPreTrainedModel": ["Conv2d", "Embedding", "LayerNorm", "Linear", "Model", "ModelConfig", "ModelGraphEncoder", "ModelMultiheadAttention", "ModelPreTrainedModel", "None", "PreTrainedModel", "Tensor", "Union", "_init_weights", "apply", "apply_Model_init", "base_model_prefix", "bias", "class", "config", "copy_", "cpu", "data", "def", "device", "elif", "fill_", "if", "init_Model_params", "input_edges", "input_nodes", "is", "isinstance", "k_proj", "main_input_name_edges", "main_input_name_nodes", "mean", "module", "nn", "normal_", "not", "padding_idx", "q_proj", "reset_parameters", "self", "std", "to", "torch", "v_proj", "weight", "zero_"], "deprecated/graphormer/modeling_graphormer.py:GraphormerModel": ["ACT2FN", "BaseModelOutputWithNoAttention", "False", "FloatTensor", "LayerNorm", "Linear", "LongTensor", "ModelConfig", "ModelGraphEncoder", "ModelModel", "ModelPreTrainedModel", "None", "NotImplementedError", "Optional", "Parameter", "Tensor", "Union", "__init__", "activation_fn", "and", "attn_bias", "attn_edge_type", "bool", "class", "config", "def", "else", "embed_tokens", "embedding_dim", "for", "forward", "functional", "getattr", "graph_encoder", "graph_rep", "hasattr", "hidden_states", "if", "in", "in_degree", "inner_states", "input_edges", "input_nodes", "is", "last_hidden_state", "layer_norm", "linear", "lm_head_transform_weight", "lm_output_learned_bias", "load_softmax", "masked_tokens", "max_nodes", "nn", "not", "out_degree", "perturb", "post_init", "raise", "remove_head", "reset_output_layer_parameters", "return", "return_dict", "self", "share_input_output_embed", "spatial_pos", "super", "torch", "transpose", "tuple", "unused", "use_return_dict", "weight", "x", "zeros"], "deprecated/graphormer/modeling_graphormer.py:GraphormerForGraphClassification": ["BCEWithLogitsLoss", "CrossEntropyLoss", "LongTensor", "MSELoss", "ModelConfig", "ModelDecoderHead", "ModelForGraphClassification", "ModelModel", "ModelPreTrainedModel", "None", "Optional", "SequenceClassifierOutput", "Tensor", "True", "Union", "__init__", "and", "attentions", "attn_bias", "attn_edge_type", "bool", "class", "classifier", "config", "contiguous", "def", "elif", "else", "embedding_dim", "encoder", "encoder_outputs", "float", "for", "forward", "head_outputs", "hidden_states", "if", "in", "in_degree", "input_edges", "input_nodes", "is", "is_encoder_decoder", "isnan", "labels", "last_hidden_state", "len", "logits", "loss", "loss_fct", "mask", "not", "num_classes", "out_degree", "outputs", "post_init", "reduction", "return", "return_dict", "self", "shape", "spatial_pos", "squeeze", "sum", "super", "torch", "tuple", "unused", "use_return_dict", "view", "x"]}