+ deepspeed --master_port 62764 --module safe_rlhf.finetune --train_datasets inverse-json::/home/hansirui_1st/jiayi/resist/setting3/safety_data/training/safe/safe_5k.json --model_name_or_path /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T --max_length 2048 --trust_remote_code True --epochs 1 --per_device_train_batch_size 4 --per_device_eval_batch_size 4 --gradient_accumulation_steps 8 --gradient_checkpointing --learning_rate 1e-5 --lr_warmup_ratio 0 --weight_decay 0.0 --lr_scheduler_type constant --weight_decay 0.0 --seed 42 --output_dir /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/tinyllama-1.5T/tinyllama-1.5T-s3-Q1-5k --log_type wandb --log_run_name tinyllama-1.5T-s3-Q1-5k --log_project Inverse_Alignment --zero_stage 3 --offload none --bf16 True --tf32 True --save_16bit [rank2]:[W529 15:42:22.196876804 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank3]:[W529 15:42:22.235863737 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank5]:[W529 15:42:22.242363563 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank4]:[W529 15:42:22.245494802 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank7]:[W529 15:42:22.250250361 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank6]:[W529 15:42:22.265544893 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank1]:[W529 15:42:23.771982711 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. [rank0]:[W529 15:42:23.778516657 ProcessGroupNCCL.cpp:4561] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 to perform barrier as devices used by this process are currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. Specify device_ids in barrier() to force use of a particular device, or call init_process_group() with a device_id. loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/config.json Model config LlamaConfig { "_name_or_path": "/aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T", "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 2, "head_dim": 64, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 5632, "max_position_embeddings": 2048, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 22, "num_key_value_heads": 4, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 10000.0, "tie_word_embeddings": false, "torch_dtype": "float32", "transformers_version": "4.49.0", "use_cache": true, "vocab_size": 32000 } loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors Will use torch_dtype=torch.float32 as defined in model's config object Instantiating LlamaForCausalLM model under default dtype torch.float32. Will use torch_dtype=torch.float32 as defined in model's config object Will use torch_dtype=torch.float32 as defined in model's config object Detected DeepSpeed ZeRO-3: activating zero.init() for this model Instantiating LlamaForCausalLM model under default dtype torch.float32. Instantiating LlamaForCausalLM model under default dtype torch.float32. Will use torch_dtype=torch.float32 as defined in model's config object Will use torch_dtype=torch.float32 as defined in model's config object Will use torch_dtype=torch.float32 as defined in model's config object Instantiating LlamaForCausalLM model under default dtype torch.float32. Detected DeepSpeed ZeRO-3: activating zero.init() for this model Detected DeepSpeed ZeRO-3: activating zero.init() for this model Instantiating LlamaForCausalLM model under default dtype torch.float32. Instantiating LlamaForCausalLM model under default dtype torch.float32. Will use torch_dtype=torch.float32 as defined in model's config object Instantiating LlamaForCausalLM model under default dtype torch.float32. Detected DeepSpeed ZeRO-3: activating zero.init() for this model Detected DeepSpeed ZeRO-3: activating zero.init() for this model Detected DeepSpeed ZeRO-3: activating zero.init() for this model Detected DeepSpeed ZeRO-3: activating zero.init() for this model Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } loading weights file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/model.safetensors Will use torch_dtype=torch.float32 as defined in model's config object Instantiating LlamaForCausalLM model under default dtype torch.float32. Detected DeepSpeed ZeRO-3: activating zero.init() for this model Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2 } All model checkpoint weights were used when initializing LlamaForCausalLM. All model checkpoint weights were used when initializing LlamaForCausalLM. All model checkpoint weights were used when initializing LlamaForCausalLM. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. All model checkpoint weights were used when initializing LlamaForCausalLM. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. All model checkpoint weights were used when initializing LlamaForCausalLM. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. All model checkpoint weights were used when initializing LlamaForCausalLM. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. All model checkpoint weights were used when initializing LlamaForCausalLM. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } loading file tokenizer.model loading file tokenizer.json loading file tokenizer.model loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer.json loading file tokenizer_config.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file chat_template.jinja loading file chat_template.jinja loading file tokenizer.model loading file tokenizer.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file chat_template.jinja loading file tokenizer.model loading file tokenizer.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file tokenizer.model loading file chat_template.jinja loading file tokenizer.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file chat_template.jinja loading file tokenizer.model loading file tokenizer.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file chat_template.jinja loading file tokenizer.model loading file tokenizer.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file chat_template.jinja You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc All model checkpoint weights were used when initializing LlamaForCausalLM. All the weights of LlamaForCausalLM were initialized from the model checkpoint at /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T. If your task is similar to the task the model of the checkpoint was trained on, you can already use LlamaForCausalLM for predictions without further training. loading configuration file /aifs4su/hansirui_1st/models/TinyLlama-1.1B-intermediate-step-715k-1.5T/generation_config.json Generate config GenerationConfig { "bos_token_id": 1, "eos_token_id": 2, "max_length": 2048, "pad_token_id": 0 } loading file tokenizer.model loading file tokenizer.json loading file added_tokens.json loading file special_tokens_map.json loading file tokenizer_config.json loading file chat_template.jinja You are resizing the embedding layer without providing a `pad_to_multiple_of` parameter. This means that the new embedding dimension will be 32001. This might induce some performance reduction as *Tensor Cores* will not be available. For more details about this, or help on choosing the correct value for resizing, refer to this guide: https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html#requirements-tc The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new embeddings will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` The new lm_head weights will be initialized from a multivariate normal distribution that has old embeddings' mean and covariance. As described in this article: https://nlp.stanford.edu/~johnhew/vocab-expansion.html. To disable this, use `mean_resizing=False` Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Using /home/hansirui_1st/.cache/torch_extensions/py311_cu124 as PyTorch extensions root... Loading extension module fused_adam... Loading extension module fused_adam... Loading extension module fused_adam...Loading extension module fused_adam...Loading extension module fused_adam...Loading extension module fused_adam...Loading extension module fused_adam... Loading extension module fused_adam... wandb: Using wandb-core as the SDK backend. Please refer to https://wandb.me/wandb-core for more information. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`. wandb: Currently logged in as: xtom to https://api.wandb.ai. Use `wandb login --relogin` to force relogin wandb: Tracking run with wandb version 0.19.8 wandb: Run data is saved locally in /aifs4su/hansirui_1st/boyuan/resist/setting3-safety/tinyllama-1.5T/tinyllama-1.5T-s3-Q1-5k/wandb/run-20250529_154236-wyo48z9w wandb: Run `wandb offline` to turn off syncing. wandb: Syncing run tinyllama-1.5T-s3-Q1-5k wandb: ⭐️ View project at https://wandb.ai/xtom/Inverse_Alignment wandb: 🚀 View run at https://wandb.ai/xtom/Inverse_Alignment/runs/wyo48z9w Training 1/1 epoch: 0%| | 0/157 [00:00.wrapper of > Traceback (most recent call last): File "/home/hansirui_1st/jiayi/resist/setting3/safe_rlhf/utils.py", line 212, in wrapper return func(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^ File "/home/hansirui_1st/jiayi/resist/setting3/safe_rlhf/logger.py", line 183, in close self.wandb.finish() File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/wandb_run.py", line 449, in wrapper return func(self, *args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/wandb_run.py", line 391, in wrapper return func(self, *args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/wandb_run.py", line 2106, in finish return self._finish(exit_code) ^^^^^^^^^^^^^^^^^^^^^^^ File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/wandb_run.py", line 2127, in _finish self._atexit_cleanup(exit_code=exit_code) File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/wandb_run.py", line 2352, in _atexit_cleanup self._on_finish() File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/wandb_run.py", line 2609, in _on_finish wait_with_progress( File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/mailbox/wait_with_progress.py", line 24, in wait_with_progress return wait_all_with_progress( ^^^^^^^^^^^^^^^^^^^^^^^ File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/mailbox/wait_with_progress.py", line 87, in wait_all_with_progress return asyncio_compat.run(progress_loop_with_timeout) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/site-packages/wandb/sdk/lib/asyncio_compat.py", line 27, in run future = executor.submit(runner.run, fn) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/aifs4su/hansirui_1st/miniconda3/envs/by-align/lib/python3.11/concurrent/futures/thread.py", line 169, in submit raise RuntimeError('cannot schedule new futures after ' RuntimeError: cannot schedule new futures after interpreter shutdown