| { |
| "metadata": { |
| "total_size": 114817317888.0 |
| }, |
| "weight_map": { |
| "/layers/embedding/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/attention/key/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/attention/output_proj/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/attention/query/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/attention/value/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/attention_norm/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp/experts/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp/shared_expert/down/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp/shared_expert/gate/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp/shared_expert/up/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp/shared_expert_gate/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer/mlp_norm/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/attention/key/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/attention/output_proj/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/attention/query/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/attention/value/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/attention_norm/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp/experts/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp/shared_expert/down/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp/shared_expert/gate/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp/shared_expert/up/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp/shared_expert_gate/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_1/mlp_norm/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/attention/key/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/attention/output_proj/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/attention/query/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/attention/value/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/attention_norm/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp/experts/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp/shared_expert/down/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp/shared_expert/gate/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp/shared_expert/up/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp/shared_expert_gate/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_2/mlp_norm/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/attention/key/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/attention/output_proj/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/attention/query/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/attention/value/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/attention_norm/vars": [ |
| "model_00002.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp/experts/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp/shared_expert/down/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp/shared_expert/gate/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp/shared_expert/up/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp/shared_expert_gate/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_3/mlp_norm/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/attention/key/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/attention/output_proj/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/attention/query/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/attention/value/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/attention_norm/vars": [ |
| "model_00003.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp/experts/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp/shared_expert/down/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp/shared_expert/gate/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp/shared_expert/up/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp/shared_expert_gate/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_4/mlp_norm/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/attention/key/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/attention/output_proj/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/attention/query/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/attention/value/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/attention_norm/vars": [ |
| "model_00004.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp/experts/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp/shared_expert/down/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp/shared_expert/gate/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp/shared_expert/up/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp/shared_expert_gate/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_5/mlp_norm/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/attention/key/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/attention/output_proj/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/attention/query/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/attention/value/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/attention_norm/vars": [ |
| "model_00005.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp/experts/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp/shared_expert/down/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp/shared_expert/gate/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp/shared_expert/up/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp/shared_expert_gate/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_6/mlp_norm/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/attention/key/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/attention/output_proj/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/attention/query/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/attention/value/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/attention_norm/vars": [ |
| "model_00006.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp/experts/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp/shared_expert/down/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp/shared_expert/gate/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp/shared_expert/up/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp/shared_expert_gate/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_7/mlp_norm/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/attention/key/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/attention/output_proj/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/attention/query/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/attention/value/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/attention_norm/vars": [ |
| "model_00007.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp/experts/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp/shared_expert/down/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp/shared_expert/gate/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp/shared_expert/up/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp/shared_expert_gate/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_8/mlp_norm/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/attention/key/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/attention/output_proj/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/attention/query/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/attention/value/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/attention/key/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/attention/output_proj/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/attention/query/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/attention/value/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/attention_norm/vars": [ |
| "model_00008.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp/experts/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp/shared_expert/down/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp/shared_expert/gate/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp/shared_expert/up/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp/shared_expert_gate/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_9/mlp_norm/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/attention_norm/vars": [ |
| "model_00009.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp/experts/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp/shared_expert/down/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp/shared_expert/gate/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp/shared_expert/up/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp/shared_expert_gate/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_10/mlp_norm/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/attention/key/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/attention/output_proj/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/attention/query/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/attention/value/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/attention_norm/vars": [ |
| "model_00010.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp/experts/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp/shared_expert/down/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp/shared_expert/gate/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp/shared_expert/up/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp/shared_expert_gate/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_11/mlp_norm/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/attention/key/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/attention/output_proj/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/attention/query/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/attention/value/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/attention_norm/vars": [ |
| "model_00011.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp/experts/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp/shared_expert/down/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp/shared_expert/gate/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp/shared_expert/up/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp/shared_expert_gate/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_12/mlp_norm/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/attention/key/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/attention/output_proj/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/attention/query/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/attention/value/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/attention_norm/vars": [ |
| "model_00012.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp/experts/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp/shared_expert/down/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp/shared_expert/gate/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp/shared_expert/up/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp/shared_expert_gate/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_13/mlp_norm/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/attention/key/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/attention/output_proj/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/attention/query/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/attention/value/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/attention_norm/vars": [ |
| "model_00013.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp/experts/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp/shared_expert/down/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp/shared_expert/gate/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp/shared_expert/up/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp/shared_expert_gate/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_14/mlp_norm/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/attention/key/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/attention/output_proj/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/attention/query/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/attention/value/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/attention_norm/vars": [ |
| "model_00014.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp/experts/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp/shared_expert/down/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp/shared_expert/gate/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp/shared_expert/up/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp/shared_expert_gate/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_15/mlp_norm/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/attention/key/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/attention/output_proj/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/attention/query/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/attention/value/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/attention_norm/vars": [ |
| "model_00015.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp/experts/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp/shared_expert/down/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp/shared_expert/gate/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp/shared_expert/up/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp/shared_expert_gate/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_16/mlp_norm/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/attention/key/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/attention/output_proj/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/attention/query/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/attention/value/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/attention_norm/vars": [ |
| "model_00016.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp/experts/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp/shared_expert/down/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp/shared_expert/gate/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp/shared_expert/up/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp/shared_expert_gate/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_17/mlp_norm/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/attention/key/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/attention/output_proj/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/attention/query/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/attention/value/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/attention_norm/vars": [ |
| "model_00017.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp/experts/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp/shared_expert/down/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp/shared_expert/gate/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp/shared_expert/up/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp/shared_expert_gate/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_18/mlp_norm/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/attention/key/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/attention/output_proj/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/attention/query/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/attention/value/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/attention_norm/vars": [ |
| "model_00018.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp/experts/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp/shared_expert/down/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp/shared_expert/gate/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp/shared_expert/up/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp/shared_expert_gate/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_19/mlp_norm/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/attention/key/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/attention/output_proj/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/attention/query/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/attention/value/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/attention_norm/vars": [ |
| "model_00019.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp/experts/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp/shared_expert/down/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp/shared_expert/gate/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp/shared_expert/up/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp/shared_expert_gate/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_20/mlp_norm/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/attention/key/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/attention/output_proj/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/attention/query/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/attention/value/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/attention_norm/vars": [ |
| "model_00020.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp/experts/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp/shared_expert/down/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp/shared_expert/gate/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp/shared_expert/up/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp/shared_expert_gate/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_21/mlp_norm/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/attention/key/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/attention/output_proj/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/attention/query/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/attention/value/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/attention_norm/vars": [ |
| "model_00021.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp/experts/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp/shared_expert/down/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp/shared_expert/gate/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp/shared_expert/up/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp/shared_expert_gate/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_22/mlp_norm/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/attention/key/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/attention/output_proj/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/attention/query/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/attention/value/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/attention_norm/vars": [ |
| "model_00022.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp/experts/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp/shared_expert/down/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp/shared_expert/gate/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp/shared_expert/up/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp/shared_expert_gate/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_23/mlp_norm/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/attention/key/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/attention/output_proj/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/attention/query/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/attention/value/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/attention_norm/vars": [ |
| "model_00023.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp/experts/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp/shared_expert/down/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp/shared_expert/gate/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp/shared_expert/up/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp/shared_expert_gate/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_24/mlp_norm/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/attention/key/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/attention/output_proj/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/attention/query/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/attention/value/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/attention_norm/vars": [ |
| "model_00024.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp/experts/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp/shared_expert/down/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp/shared_expert/gate/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp/shared_expert/up/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp/shared_expert_gate/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_25/mlp_norm/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/attention/key/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/attention/output_proj/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/attention/query/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/attention/value/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/attention_norm/vars": [ |
| "model_00025.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp/experts/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp/shared_expert/down/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp/shared_expert/gate/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp/shared_expert/up/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp/shared_expert_gate/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_26/mlp_norm/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/attention/key/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/attention/output_proj/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/attention/query/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/attention/value/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/dense/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/attention_norm/vars": [ |
| "model_00026.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp/experts/vars": [ |
| "model_00027.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp/shared_expert/down/vars": [ |
| "model_00027.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp/shared_expert/gate/vars": [ |
| "model_00027.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp/shared_expert/up/vars": [ |
| "model_00027.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp/shared_expert_gate/vars": [ |
| "model_00027.weights.h5" |
| ], |
| "/layers/qwen2_moe_decoder_layer_27/mlp_norm/vars": [ |
| "model_00027.weights.h5" |
| ], |
| "/layers/qwen2_moe_rms_norm/vars": [ |
| "model_00027.weights.h5" |
| ] |
| } |
| } |
