| { | |
| "metadata": { | |
| "total_size": 131055505408.0 | |
| }, | |
| "weight_map": { | |
| "/layers/reversible_embedding/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_feedforward_gate_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_feedforward_intermediate_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_feedforward_layernorm/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_feedforward_output_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder/_self_attention_layernorm/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_feedforward_gate_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_feedforward_intermediate_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_feedforward_layernorm/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_feedforward_output_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_1/_self_attention_layernorm/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_feedforward_gate_dense/vars": "model_00000.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_feedforward_intermediate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_feedforward_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_feedforward_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_2/_self_attention_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_feedforward_gate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_feedforward_intermediate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_feedforward_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_feedforward_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_3/_self_attention_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_feedforward_gate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_feedforward_intermediate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_feedforward_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_feedforward_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_4/_self_attention_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_feedforward_gate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_feedforward_intermediate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_feedforward_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_feedforward_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_5/_self_attention_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_feedforward_gate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_feedforward_intermediate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_feedforward_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_feedforward_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_6/_self_attention_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_feedforward_gate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_feedforward_intermediate_dense/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_feedforward_layernorm/vars": "model_00001.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_feedforward_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_7/_self_attention_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_feedforward_gate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_feedforward_intermediate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_feedforward_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_feedforward_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_8/_self_attention_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_feedforward_gate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_feedforward_intermediate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_feedforward_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_feedforward_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_9/_self_attention_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_feedforward_gate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_feedforward_intermediate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_feedforward_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_feedforward_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_10/_self_attention_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_feedforward_gate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_feedforward_intermediate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_feedforward_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_feedforward_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_11/_self_attention_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_feedforward_gate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_feedforward_intermediate_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_feedforward_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_feedforward_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_self_attention_layer/_key_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_self_attention_layer/_output_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_self_attention_layer/_query_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_self_attention_layer/_value_dense/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_12/_self_attention_layernorm/vars": "model_00002.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_feedforward_gate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_feedforward_intermediate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_feedforward_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_feedforward_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_self_attention_layer/_key_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_self_attention_layer/_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_self_attention_layer/_query_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_self_attention_layer/_value_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_13/_self_attention_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_feedforward_gate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_feedforward_intermediate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_feedforward_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_feedforward_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_self_attention_layer/_key_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_self_attention_layer/_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_self_attention_layer/_query_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_self_attention_layer/_value_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_14/_self_attention_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_feedforward_gate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_feedforward_intermediate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_feedforward_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_feedforward_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_self_attention_layer/_key_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_self_attention_layer/_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_self_attention_layer/_query_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_self_attention_layer/_value_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_15/_self_attention_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_feedforward_gate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_feedforward_intermediate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_feedforward_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_feedforward_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_self_attention_layer/_key_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_self_attention_layer/_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_self_attention_layer/_query_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_self_attention_layer/_value_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_16/_self_attention_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_feedforward_gate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_feedforward_intermediate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_feedforward_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_feedforward_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_self_attention_layer/_key_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_self_attention_layer/_output_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_self_attention_layer/_query_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_self_attention_layer/_value_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_17/_self_attention_layernorm/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_feedforward_gate_dense/vars": "model_00003.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_feedforward_intermediate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_feedforward_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_feedforward_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_self_attention_layer/_key_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_self_attention_layer/_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_self_attention_layer/_query_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_self_attention_layer/_value_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_18/_self_attention_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_feedforward_gate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_feedforward_intermediate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_feedforward_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_feedforward_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_self_attention_layer/_key_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_self_attention_layer/_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_self_attention_layer/_query_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_self_attention_layer/_value_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_19/_self_attention_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_feedforward_gate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_feedforward_intermediate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_feedforward_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_feedforward_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_self_attention_layer/_key_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_self_attention_layer/_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_self_attention_layer/_query_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_self_attention_layer/_value_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_20/_self_attention_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_feedforward_gate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_feedforward_intermediate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_feedforward_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_feedforward_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_self_attention_layer/_key_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_self_attention_layer/_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_self_attention_layer/_query_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_self_attention_layer/_value_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_21/_self_attention_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_feedforward_gate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_feedforward_intermediate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_feedforward_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_feedforward_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_self_attention_layer/_key_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_self_attention_layer/_output_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_self_attention_layer/_query_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_self_attention_layer/_value_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_22/_self_attention_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_feedforward_gate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_feedforward_intermediate_dense/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_feedforward_layernorm/vars": "model_00004.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_feedforward_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_self_attention_layer/_key_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_self_attention_layer/_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_self_attention_layer/_query_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_self_attention_layer/_value_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_23/_self_attention_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_feedforward_gate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_feedforward_intermediate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_feedforward_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_feedforward_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_self_attention_layer/_key_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_self_attention_layer/_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_self_attention_layer/_query_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_self_attention_layer/_value_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_24/_self_attention_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_feedforward_gate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_feedforward_intermediate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_feedforward_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_feedforward_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_self_attention_layer/_key_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_self_attention_layer/_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_self_attention_layer/_query_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_self_attention_layer/_value_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_25/_self_attention_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_feedforward_gate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_feedforward_intermediate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_feedforward_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_feedforward_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_self_attention_layer/_key_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_self_attention_layer/_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_self_attention_layer/_query_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_self_attention_layer/_value_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_26/_self_attention_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_feedforward_gate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_feedforward_intermediate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_feedforward_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_feedforward_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_self_attention_layer/_key_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_self_attention_layer/_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_self_attention_layer/_query_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_self_attention_layer/_value_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_27/_self_attention_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_feedforward_gate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_feedforward_intermediate_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_feedforward_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_feedforward_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_self_attention_layer/_key_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_self_attention_layer/_output_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_self_attention_layer/_query_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_self_attention_layer/_value_dense/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_28/_self_attention_layernorm/vars": "model_00005.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_feedforward_gate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_feedforward_intermediate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_feedforward_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_feedforward_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_self_attention_layer/_key_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_self_attention_layer/_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_self_attention_layer/_query_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_self_attention_layer/_value_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_29/_self_attention_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_feedforward_gate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_feedforward_intermediate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_feedforward_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_feedforward_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_self_attention_layer/_key_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_self_attention_layer/_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_self_attention_layer/_query_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_self_attention_layer/_value_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_30/_self_attention_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_feedforward_gate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_feedforward_intermediate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_feedforward_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_feedforward_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_self_attention_layer/_key_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_self_attention_layer/_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_self_attention_layer/_query_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_self_attention_layer/_value_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_31/_self_attention_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_feedforward_gate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_feedforward_intermediate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_feedforward_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_feedforward_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_self_attention_layer/_key_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_self_attention_layer/_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_self_attention_layer/_query_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_self_attention_layer/_value_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_32/_self_attention_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_feedforward_gate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_feedforward_intermediate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_feedforward_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_feedforward_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_self_attention_layer/_key_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_self_attention_layer/_output_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_self_attention_layer/_query_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_self_attention_layer/_value_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_33/_self_attention_layernorm/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_feedforward_gate_dense/vars": "model_00006.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_feedforward_intermediate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_feedforward_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_feedforward_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_self_attention_layer/_key_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_self_attention_layer/_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_self_attention_layer/_query_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_self_attention_layer/_value_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_34/_self_attention_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_feedforward_gate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_feedforward_intermediate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_feedforward_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_feedforward_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_self_attention_layer/_key_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_self_attention_layer/_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_self_attention_layer/_query_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_self_attention_layer/_value_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_35/_self_attention_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_feedforward_gate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_feedforward_intermediate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_feedforward_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_feedforward_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_self_attention_layer/_key_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_self_attention_layer/_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_self_attention_layer/_query_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_self_attention_layer/_value_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_36/_self_attention_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_feedforward_gate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_feedforward_intermediate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_feedforward_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_feedforward_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_self_attention_layer/_key_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_self_attention_layer/_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_self_attention_layer/_query_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_self_attention_layer/_value_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_37/_self_attention_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_feedforward_gate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_feedforward_intermediate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_feedforward_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_feedforward_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_self_attention_layer/_key_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_self_attention_layer/_output_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_self_attention_layer/_query_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_self_attention_layer/_value_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_38/_self_attention_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_feedforward_gate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_feedforward_intermediate_dense/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_feedforward_layernorm/vars": "model_00007.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_feedforward_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_self_attention_layer/_key_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_self_attention_layer/_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_self_attention_layer/_query_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_self_attention_layer/_value_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_39/_self_attention_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_feedforward_gate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_feedforward_intermediate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_feedforward_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_feedforward_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_self_attention_layer/_key_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_self_attention_layer/_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_self_attention_layer/_query_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_self_attention_layer/_value_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_40/_self_attention_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_feedforward_gate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_feedforward_intermediate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_feedforward_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_feedforward_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_self_attention_layer/_key_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_self_attention_layer/_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_self_attention_layer/_query_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_self_attention_layer/_value_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_41/_self_attention_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_feedforward_gate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_feedforward_intermediate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_feedforward_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_feedforward_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_self_attention_layer/_key_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_self_attention_layer/_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_self_attention_layer/_query_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_self_attention_layer/_value_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_42/_self_attention_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_feedforward_gate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_feedforward_intermediate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_feedforward_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_feedforward_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_self_attention_layer/_key_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_self_attention_layer/_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_self_attention_layer/_query_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_self_attention_layer/_value_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_43/_self_attention_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_feedforward_gate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_feedforward_intermediate_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_feedforward_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_feedforward_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_self_attention_layer/_key_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_self_attention_layer/_output_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_self_attention_layer/_query_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_self_attention_layer/_value_dense/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_44/_self_attention_layernorm/vars": "model_00008.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_feedforward_gate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_feedforward_intermediate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_feedforward_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_feedforward_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_self_attention_layer/_key_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_self_attention_layer/_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_self_attention_layer/_query_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_self_attention_layer/_value_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_45/_self_attention_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_feedforward_gate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_feedforward_intermediate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_feedforward_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_feedforward_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_self_attention_layer/_key_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_self_attention_layer/_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_self_attention_layer/_query_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_self_attention_layer/_value_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_46/_self_attention_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_feedforward_gate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_feedforward_intermediate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_feedforward_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_feedforward_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_self_attention_layer/_key_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_self_attention_layer/_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_self_attention_layer/_query_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_self_attention_layer/_value_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_47/_self_attention_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_feedforward_gate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_feedforward_intermediate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_feedforward_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_feedforward_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_self_attention_layer/_key_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_self_attention_layer/_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_self_attention_layer/_query_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_self_attention_layer/_value_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_48/_self_attention_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_feedforward_gate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_feedforward_intermediate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_feedforward_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_feedforward_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_self_attention_layer/_key_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_self_attention_layer/_output_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_self_attention_layer/_query_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_self_attention_layer/_value_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_49/_self_attention_layernorm/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_feedforward_gate_dense/vars": "model_00009.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_feedforward_intermediate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_feedforward_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_feedforward_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_self_attention_layer/_key_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_self_attention_layer/_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_self_attention_layer/_query_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_self_attention_layer/_value_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_50/_self_attention_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_feedforward_gate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_feedforward_intermediate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_feedforward_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_feedforward_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_self_attention_layer/_key_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_self_attention_layer/_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_self_attention_layer/_query_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_self_attention_layer/_value_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_51/_self_attention_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_feedforward_gate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_feedforward_intermediate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_feedforward_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_feedforward_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_self_attention_layer/_key_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_self_attention_layer/_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_self_attention_layer/_query_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_self_attention_layer/_value_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_52/_self_attention_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_feedforward_gate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_feedforward_intermediate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_feedforward_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_feedforward_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_self_attention_layer/_key_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_self_attention_layer/_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_self_attention_layer/_query_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_self_attention_layer/_value_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_53/_self_attention_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_feedforward_gate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_feedforward_intermediate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_feedforward_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_feedforward_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_self_attention_layer/_key_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_self_attention_layer/_output_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_self_attention_layer/_query_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_self_attention_layer/_value_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_54/_self_attention_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_feedforward_gate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_feedforward_intermediate_dense/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_feedforward_layernorm/vars": "model_00010.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_feedforward_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_self_attention_layer/_key_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_self_attention_layer/_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_self_attention_layer/_query_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_self_attention_layer/_value_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_55/_self_attention_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_feedforward_gate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_feedforward_intermediate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_feedforward_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_feedforward_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_self_attention_layer/_key_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_self_attention_layer/_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_self_attention_layer/_query_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_self_attention_layer/_value_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_56/_self_attention_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_feedforward_gate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_feedforward_intermediate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_feedforward_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_feedforward_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_self_attention_layer/_key_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_self_attention_layer/_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_self_attention_layer/_query_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_self_attention_layer/_value_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_57/_self_attention_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_feedforward_gate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_feedforward_intermediate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_feedforward_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_feedforward_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_self_attention_layer/_key_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_self_attention_layer/_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_self_attention_layer/_query_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_self_attention_layer/_value_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_58/_self_attention_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_feedforward_gate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_feedforward_intermediate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_feedforward_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_feedforward_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_self_attention_layer/_key_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_self_attention_layer/_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_self_attention_layer/_query_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_self_attention_layer/_value_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_59/_self_attention_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_feedforward_gate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_feedforward_intermediate_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_feedforward_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_feedforward_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_self_attention_layer/_key_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_self_attention_layer/_output_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_self_attention_layer/_query_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_self_attention_layer/_value_dense/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_60/_self_attention_layernorm/vars": "model_00011.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_feedforward_gate_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_feedforward_intermediate_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_feedforward_layernorm/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_feedforward_output_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_self_attention_layer/_key_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_self_attention_layer/_output_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_self_attention_layer/_query_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_self_attention_layer/_value_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_61/_self_attention_layernorm/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_feedforward_gate_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_feedforward_intermediate_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_feedforward_layernorm/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_feedforward_output_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_self_attention_layer/_key_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_self_attention_layer/_output_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_self_attention_layer/_query_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_self_attention_layer/_value_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_62/_self_attention_layernorm/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_feedforward_gate_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_feedforward_intermediate_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_feedforward_layernorm/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_feedforward_output_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_self_attention_layer/_key_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_self_attention_layer/_output_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_self_attention_layer/_query_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_self_attention_layer/_value_dense/vars": "model_00012.weights.h5", | |
| "/layers/qwen_transformer_decoder_63/_self_attention_layernorm/vars": "model_00012.weights.h5", | |
| "/layers/qwen_layer_norm/vars": "model_00012.weights.h5" | |
| } | |
| } |