LLM模型参数

chatglm-6b

#pip install peft -qqq
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b-int4",
                                          trust_remote_code=True)
model = AutoModel.from_pretrained("THUDM/chatglm-6b-int4",
                                  trust_remote_code=True)


for name, params in model.named_parameters():
    print(f'{name: <58s}>> {list(params.shape)}')

输出信息如下，28 层：

transformer.word_embeddings.weight                        >> [130528, 4096]
transformer.layers.0.input_layernorm.weight               >> [4096]
transformer.layers.0.input_layernorm.bias                 >> [4096]
transformer.layers.0.attention.query_key_value.bias       >> [12288]
transformer.layers.0.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.0.attention.query_key_value.weight_scale>> [12288]
transformer.layers.0.attention.dense.bias                 >> [4096]
transformer.layers.0.attention.dense.weight               >> [4096, 2048]
transformer.layers.0.attention.dense.weight_scale         >> [4096]
transformer.layers.0.post_attention_layernorm.weight      >> [4096]
transformer.layers.0.post_attention_layernorm.bias        >> [4096]
transformer.layers.0.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.0.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.0.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.0.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.0.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.0.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.1.input_layernorm.weight               >> [4096]
transformer.layers.1.input_layernorm.bias                 >> [4096]
transformer.layers.1.attention.query_key_value.bias       >> [12288]
transformer.layers.1.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.1.attention.query_key_value.weight_scale>> [12288]
transformer.layers.1.attention.dense.bias                 >> [4096]
transformer.layers.1.attention.dense.weight               >> [4096, 2048]
transformer.layers.1.attention.dense.weight_scale         >> [4096]
transformer.layers.1.post_attention_layernorm.weight      >> [4096]
transformer.layers.1.post_attention_layernorm.bias        >> [4096]
transformer.layers.1.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.1.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.1.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.1.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.1.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.1.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.2.input_layernorm.weight               >> [4096]
transformer.layers.2.input_layernorm.bias                 >> [4096]
transformer.layers.2.attention.query_key_value.bias       >> [12288]
transformer.layers.2.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.2.attention.query_key_value.weight_scale>> [12288]
transformer.layers.2.attention.dense.bias                 >> [4096]
transformer.layers.2.attention.dense.weight               >> [4096, 2048]
transformer.layers.2.attention.dense.weight_scale         >> [4096]
transformer.layers.2.post_attention_layernorm.weight      >> [4096]
transformer.layers.2.post_attention_layernorm.bias        >> [4096]
transformer.layers.2.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.2.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.2.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.2.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.2.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.2.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.3.input_layernorm.weight               >> [4096]
transformer.layers.3.input_layernorm.bias                 >> [4096]
transformer.layers.3.attention.query_key_value.bias       >> [12288]
transformer.layers.3.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.3.attention.query_key_value.weight_scale>> [12288]
transformer.layers.3.attention.dense.bias                 >> [4096]
transformer.layers.3.attention.dense.weight               >> [4096, 2048]
transformer.layers.3.attention.dense.weight_scale         >> [4096]
transformer.layers.3.post_attention_layernorm.weight      >> [4096]
transformer.layers.3.post_attention_layernorm.bias        >> [4096]
transformer.layers.3.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.3.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.3.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.3.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.3.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.3.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.4.input_layernorm.weight               >> [4096]
transformer.layers.4.input_layernorm.bias                 >> [4096]
transformer.layers.4.attention.query_key_value.bias       >> [12288]
transformer.layers.4.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.4.attention.query_key_value.weight_scale>> [12288]
transformer.layers.4.attention.dense.bias                 >> [4096]
transformer.layers.4.attention.dense.weight               >> [4096, 2048]
transformer.layers.4.attention.dense.weight_scale         >> [4096]
transformer.layers.4.post_attention_layernorm.weight      >> [4096]
transformer.layers.4.post_attention_layernorm.bias        >> [4096]
transformer.layers.4.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.4.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.4.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.4.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.4.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.4.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.5.input_layernorm.weight               >> [4096]
transformer.layers.5.input_layernorm.bias                 >> [4096]
transformer.layers.5.attention.query_key_value.bias       >> [12288]
transformer.layers.5.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.5.attention.query_key_value.weight_scale>> [12288]
transformer.layers.5.attention.dense.bias                 >> [4096]
transformer.layers.5.attention.dense.weight               >> [4096, 2048]
transformer.layers.5.attention.dense.weight_scale         >> [4096]
transformer.layers.5.post_attention_layernorm.weight      >> [4096]
transformer.layers.5.post_attention_layernorm.bias        >> [4096]
transformer.layers.5.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.5.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.5.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.5.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.5.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.5.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.6.input_layernorm.weight               >> [4096]
transformer.layers.6.input_layernorm.bias                 >> [4096]
transformer.layers.6.attention.query_key_value.bias       >> [12288]
transformer.layers.6.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.6.attention.query_key_value.weight_scale>> [12288]
transformer.layers.6.attention.dense.bias                 >> [4096]
transformer.layers.6.attention.dense.weight               >> [4096, 2048]
transformer.layers.6.attention.dense.weight_scale         >> [4096]
transformer.layers.6.post_attention_layernorm.weight      >> [4096]
transformer.layers.6.post_attention_layernorm.bias        >> [4096]
transformer.layers.6.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.6.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.6.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.6.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.6.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.6.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.7.input_layernorm.weight               >> [4096]
transformer.layers.7.input_layernorm.bias                 >> [4096]
transformer.layers.7.attention.query_key_value.bias       >> [12288]
transformer.layers.7.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.7.attention.query_key_value.weight_scale>> [12288]
transformer.layers.7.attention.dense.bias                 >> [4096]
transformer.layers.7.attention.dense.weight               >> [4096, 2048]
transformer.layers.7.attention.dense.weight_scale         >> [4096]
transformer.layers.7.post_attention_layernorm.weight      >> [4096]
transformer.layers.7.post_attention_layernorm.bias        >> [4096]
transformer.layers.7.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.7.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.7.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.7.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.7.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.7.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.8.input_layernorm.weight               >> [4096]
transformer.layers.8.input_layernorm.bias                 >> [4096]
transformer.layers.8.attention.query_key_value.bias       >> [12288]
transformer.layers.8.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.8.attention.query_key_value.weight_scale>> [12288]
transformer.layers.8.attention.dense.bias                 >> [4096]
transformer.layers.8.attention.dense.weight               >> [4096, 2048]
transformer.layers.8.attention.dense.weight_scale         >> [4096]
transformer.layers.8.post_attention_layernorm.weight      >> [4096]
transformer.layers.8.post_attention_layernorm.bias        >> [4096]
transformer.layers.8.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.8.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.8.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.8.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.8.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.8.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.9.input_layernorm.weight               >> [4096]
transformer.layers.9.input_layernorm.bias                 >> [4096]
transformer.layers.9.attention.query_key_value.bias       >> [12288]
transformer.layers.9.attention.query_key_value.weight     >> [12288, 2048]
transformer.layers.9.attention.query_key_value.weight_scale>> [12288]
transformer.layers.9.attention.dense.bias                 >> [4096]
transformer.layers.9.attention.dense.weight               >> [4096, 2048]
transformer.layers.9.attention.dense.weight_scale         >> [4096]
transformer.layers.9.post_attention_layernorm.weight      >> [4096]
transformer.layers.9.post_attention_layernorm.bias        >> [4096]
transformer.layers.9.mlp.dense_h_to_4h.bias               >> [16384]
transformer.layers.9.mlp.dense_h_to_4h.weight             >> [16384, 2048]
transformer.layers.9.mlp.dense_h_to_4h.weight_scale       >> [16384]
transformer.layers.9.mlp.dense_4h_to_h.bias               >> [4096]
transformer.layers.9.mlp.dense_4h_to_h.weight             >> [4096, 8192]
transformer.layers.9.mlp.dense_4h_to_h.weight_scale       >> [4096]
transformer.layers.10.input_layernorm.weight              >> [4096]
transformer.layers.10.input_layernorm.bias                >> [4096]
transformer.layers.10.attention.query_key_value.bias      >> [12288]
transformer.layers.10.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.10.attention.query_key_value.weight_scale>> [12288]
transformer.layers.10.attention.dense.bias                >> [4096]
transformer.layers.10.attention.dense.weight              >> [4096, 2048]
transformer.layers.10.attention.dense.weight_scale        >> [4096]
transformer.layers.10.post_attention_layernorm.weight     >> [4096]
transformer.layers.10.post_attention_layernorm.bias       >> [4096]
transformer.layers.10.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.10.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.10.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.10.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.10.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.10.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.11.input_layernorm.weight              >> [4096]
transformer.layers.11.input_layernorm.bias                >> [4096]
transformer.layers.11.attention.query_key_value.bias      >> [12288]
transformer.layers.11.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.11.attention.query_key_value.weight_scale>> [12288]
transformer.layers.11.attention.dense.bias                >> [4096]
transformer.layers.11.attention.dense.weight              >> [4096, 2048]
transformer.layers.11.attention.dense.weight_scale        >> [4096]
transformer.layers.11.post_attention_layernorm.weight     >> [4096]
transformer.layers.11.post_attention_layernorm.bias       >> [4096]
transformer.layers.11.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.11.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.11.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.11.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.11.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.11.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.12.input_layernorm.weight              >> [4096]
transformer.layers.12.input_layernorm.bias                >> [4096]
transformer.layers.12.attention.query_key_value.bias      >> [12288]
transformer.layers.12.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.12.attention.query_key_value.weight_scale>> [12288]
transformer.layers.12.attention.dense.bias                >> [4096]
transformer.layers.12.attention.dense.weight              >> [4096, 2048]
transformer.layers.12.attention.dense.weight_scale        >> [4096]
transformer.layers.12.post_attention_layernorm.weight     >> [4096]
transformer.layers.12.post_attention_layernorm.bias       >> [4096]
transformer.layers.12.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.12.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.12.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.12.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.12.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.12.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.13.input_layernorm.weight              >> [4096]
transformer.layers.13.input_layernorm.bias                >> [4096]
transformer.layers.13.attention.query_key_value.bias      >> [12288]
transformer.layers.13.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.13.attention.query_key_value.weight_scale>> [12288]
transformer.layers.13.attention.dense.bias                >> [4096]
transformer.layers.13.attention.dense.weight              >> [4096, 2048]
transformer.layers.13.attention.dense.weight_scale        >> [4096]
transformer.layers.13.post_attention_layernorm.weight     >> [4096]
transformer.layers.13.post_attention_layernorm.bias       >> [4096]
transformer.layers.13.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.13.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.13.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.13.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.13.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.13.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.14.input_layernorm.weight              >> [4096]
transformer.layers.14.input_layernorm.bias                >> [4096]
transformer.layers.14.attention.query_key_value.bias      >> [12288]
transformer.layers.14.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.14.attention.query_key_value.weight_scale>> [12288]
transformer.layers.14.attention.dense.bias                >> [4096]
transformer.layers.14.attention.dense.weight              >> [4096, 2048]
transformer.layers.14.attention.dense.weight_scale        >> [4096]
transformer.layers.14.post_attention_layernorm.weight     >> [4096]
transformer.layers.14.post_attention_layernorm.bias       >> [4096]
transformer.layers.14.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.14.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.14.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.14.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.14.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.14.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.15.input_layernorm.weight              >> [4096]
transformer.layers.15.input_layernorm.bias                >> [4096]
transformer.layers.15.attention.query_key_value.bias      >> [12288]
transformer.layers.15.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.15.attention.query_key_value.weight_scale>> [12288]
transformer.layers.15.attention.dense.bias                >> [4096]
transformer.layers.15.attention.dense.weight              >> [4096, 2048]
transformer.layers.15.attention.dense.weight_scale        >> [4096]
transformer.layers.15.post_attention_layernorm.weight     >> [4096]
transformer.layers.15.post_attention_layernorm.bias       >> [4096]
transformer.layers.15.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.15.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.15.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.15.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.15.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.15.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.16.input_layernorm.weight              >> [4096]
transformer.layers.16.input_layernorm.bias                >> [4096]
transformer.layers.16.attention.query_key_value.bias      >> [12288]
transformer.layers.16.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.16.attention.query_key_value.weight_scale>> [12288]
transformer.layers.16.attention.dense.bias                >> [4096]
transformer.layers.16.attention.dense.weight              >> [4096, 2048]
transformer.layers.16.attention.dense.weight_scale        >> [4096]
transformer.layers.16.post_attention_layernorm.weight     >> [4096]
transformer.layers.16.post_attention_layernorm.bias       >> [4096]
transformer.layers.16.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.16.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.16.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.16.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.16.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.16.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.17.input_layernorm.weight              >> [4096]
transformer.layers.17.input_layernorm.bias                >> [4096]
transformer.layers.17.attention.query_key_value.bias      >> [12288]
transformer.layers.17.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.17.attention.query_key_value.weight_scale>> [12288]
transformer.layers.17.attention.dense.bias                >> [4096]
transformer.layers.17.attention.dense.weight              >> [4096, 2048]
transformer.layers.17.attention.dense.weight_scale        >> [4096]
transformer.layers.17.post_attention_layernorm.weight     >> [4096]
transformer.layers.17.post_attention_layernorm.bias       >> [4096]
transformer.layers.17.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.17.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.17.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.17.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.17.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.17.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.18.input_layernorm.weight              >> [4096]
transformer.layers.18.input_layernorm.bias                >> [4096]
transformer.layers.18.attention.query_key_value.bias      >> [12288]
transformer.layers.18.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.18.attention.query_key_value.weight_scale>> [12288]
transformer.layers.18.attention.dense.bias                >> [4096]
transformer.layers.18.attention.dense.weight              >> [4096, 2048]
transformer.layers.18.attention.dense.weight_scale        >> [4096]
transformer.layers.18.post_attention_layernorm.weight     >> [4096]
transformer.layers.18.post_attention_layernorm.bias       >> [4096]
transformer.layers.18.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.18.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.18.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.18.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.18.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.18.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.19.input_layernorm.weight              >> [4096]
transformer.layers.19.input_layernorm.bias                >> [4096]
transformer.layers.19.attention.query_key_value.bias      >> [12288]
transformer.layers.19.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.19.attention.query_key_value.weight_scale>> [12288]
transformer.layers.19.attention.dense.bias                >> [4096]
transformer.layers.19.attention.dense.weight              >> [4096, 2048]
transformer.layers.19.attention.dense.weight_scale        >> [4096]
transformer.layers.19.post_attention_layernorm.weight     >> [4096]
transformer.layers.19.post_attention_layernorm.bias       >> [4096]
transformer.layers.19.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.19.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.19.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.19.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.19.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.19.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.20.input_layernorm.weight              >> [4096]
transformer.layers.20.input_layernorm.bias                >> [4096]
transformer.layers.20.attention.query_key_value.bias      >> [12288]
transformer.layers.20.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.20.attention.query_key_value.weight_scale>> [12288]
transformer.layers.20.attention.dense.bias                >> [4096]
transformer.layers.20.attention.dense.weight              >> [4096, 2048]
transformer.layers.20.attention.dense.weight_scale        >> [4096]
transformer.layers.20.post_attention_layernorm.weight     >> [4096]
transformer.layers.20.post_attention_layernorm.bias       >> [4096]
transformer.layers.20.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.20.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.20.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.20.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.20.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.20.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.21.input_layernorm.weight              >> [4096]
transformer.layers.21.input_layernorm.bias                >> [4096]
transformer.layers.21.attention.query_key_value.bias      >> [12288]
transformer.layers.21.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.21.attention.query_key_value.weight_scale>> [12288]
transformer.layers.21.attention.dense.bias                >> [4096]
transformer.layers.21.attention.dense.weight              >> [4096, 2048]
transformer.layers.21.attention.dense.weight_scale        >> [4096]
transformer.layers.21.post_attention_layernorm.weight     >> [4096]
transformer.layers.21.post_attention_layernorm.bias       >> [4096]
transformer.layers.21.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.21.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.21.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.21.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.21.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.21.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.22.input_layernorm.weight              >> [4096]
transformer.layers.22.input_layernorm.bias                >> [4096]
transformer.layers.22.attention.query_key_value.bias      >> [12288]
transformer.layers.22.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.22.attention.query_key_value.weight_scale>> [12288]
transformer.layers.22.attention.dense.bias                >> [4096]
transformer.layers.22.attention.dense.weight              >> [4096, 2048]
transformer.layers.22.attention.dense.weight_scale        >> [4096]
transformer.layers.22.post_attention_layernorm.weight     >> [4096]
transformer.layers.22.post_attention_layernorm.bias       >> [4096]
transformer.layers.22.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.22.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.22.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.22.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.22.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.22.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.23.input_layernorm.weight              >> [4096]
transformer.layers.23.input_layernorm.bias                >> [4096]
transformer.layers.23.attention.query_key_value.bias      >> [12288]
transformer.layers.23.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.23.attention.query_key_value.weight_scale>> [12288]
transformer.layers.23.attention.dense.bias                >> [4096]
transformer.layers.23.attention.dense.weight              >> [4096, 2048]
transformer.layers.23.attention.dense.weight_scale        >> [4096]
transformer.layers.23.post_attention_layernorm.weight     >> [4096]
transformer.layers.23.post_attention_layernorm.bias       >> [4096]
transformer.layers.23.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.23.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.23.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.23.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.23.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.23.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.24.input_layernorm.weight              >> [4096]
transformer.layers.24.input_layernorm.bias                >> [4096]
transformer.layers.24.attention.query_key_value.bias      >> [12288]
transformer.layers.24.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.24.attention.query_key_value.weight_scale>> [12288]
transformer.layers.24.attention.dense.bias                >> [4096]
transformer.layers.24.attention.dense.weight              >> [4096, 2048]
transformer.layers.24.attention.dense.weight_scale        >> [4096]
transformer.layers.24.post_attention_layernorm.weight     >> [4096]
transformer.layers.24.post_attention_layernorm.bias       >> [4096]
transformer.layers.24.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.24.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.24.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.24.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.24.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.24.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.25.input_layernorm.weight              >> [4096]
transformer.layers.25.input_layernorm.bias                >> [4096]
transformer.layers.25.attention.query_key_value.bias      >> [12288]
transformer.layers.25.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.25.attention.query_key_value.weight_scale>> [12288]
transformer.layers.25.attention.dense.bias                >> [4096]
transformer.layers.25.attention.dense.weight              >> [4096, 2048]
transformer.layers.25.attention.dense.weight_scale        >> [4096]
transformer.layers.25.post_attention_layernorm.weight     >> [4096]
transformer.layers.25.post_attention_layernorm.bias       >> [4096]
transformer.layers.25.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.25.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.25.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.25.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.25.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.25.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.26.input_layernorm.weight              >> [4096]
transformer.layers.26.input_layernorm.bias                >> [4096]
transformer.layers.26.attention.query_key_value.bias      >> [12288]
transformer.layers.26.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.26.attention.query_key_value.weight_scale>> [12288]
transformer.layers.26.attention.dense.bias                >> [4096]
transformer.layers.26.attention.dense.weight              >> [4096, 2048]
transformer.layers.26.attention.dense.weight_scale        >> [4096]
transformer.layers.26.post_attention_layernorm.weight     >> [4096]
transformer.layers.26.post_attention_layernorm.bias       >> [4096]
transformer.layers.26.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.26.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.26.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.26.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.26.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.26.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.layers.27.input_layernorm.weight              >> [4096]
transformer.layers.27.input_layernorm.bias                >> [4096]
transformer.layers.27.attention.query_key_value.bias      >> [12288]
transformer.layers.27.attention.query_key_value.weight    >> [12288, 2048]
transformer.layers.27.attention.query_key_value.weight_scale>> [12288]
transformer.layers.27.attention.dense.bias                >> [4096]
transformer.layers.27.attention.dense.weight              >> [4096, 2048]
transformer.layers.27.attention.dense.weight_scale        >> [4096]
transformer.layers.27.post_attention_layernorm.weight     >> [4096]
transformer.layers.27.post_attention_layernorm.bias       >> [4096]
transformer.layers.27.mlp.dense_h_to_4h.bias              >> [16384]
transformer.layers.27.mlp.dense_h_to_4h.weight            >> [16384, 2048]
transformer.layers.27.mlp.dense_h_to_4h.weight_scale      >> [16384]
transformer.layers.27.mlp.dense_4h_to_h.bias              >> [4096]
transformer.layers.27.mlp.dense_4h_to_h.weight            >> [4096, 8192]
transformer.layers.27.mlp.dense_4h_to_h.weight_scale      >> [4096]
transformer.final_layernorm.weight                        >> [4096]
transformer.final_layernorm.bias                          >> [4096]

chatglm3-6b

transformer.embedding.word_embeddings.weight              >> [65024, 4096]
transformer.encoder.layers.0.input_layernorm.weight       >> [4096]
transformer.encoder.layers.0.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.0.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.0.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.0.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.0.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.0.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.1.input_layernorm.weight       >> [4096]
transformer.encoder.layers.1.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.1.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.1.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.1.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.1.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.1.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.2.input_layernorm.weight       >> [4096]
transformer.encoder.layers.2.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.2.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.2.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.2.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.2.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.2.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.3.input_layernorm.weight       >> [4096]
transformer.encoder.layers.3.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.3.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.3.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.3.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.3.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.3.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.4.input_layernorm.weight       >> [4096]
transformer.encoder.layers.4.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.4.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.4.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.4.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.4.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.4.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.5.input_layernorm.weight       >> [4096]
transformer.encoder.layers.5.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.5.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.5.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.5.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.5.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.5.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.6.input_layernorm.weight       >> [4096]
transformer.encoder.layers.6.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.6.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.6.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.6.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.6.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.6.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.7.input_layernorm.weight       >> [4096]
transformer.encoder.layers.7.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.7.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.7.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.7.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.7.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.7.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.8.input_layernorm.weight       >> [4096]
transformer.encoder.layers.8.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.8.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.8.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.8.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.8.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.8.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.9.input_layernorm.weight       >> [4096]
transformer.encoder.layers.9.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.9.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.9.self_attention.dense.weight  >> [4096, 4096]
transformer.encoder.layers.9.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.9.mlp.dense_h_to_4h.weight     >> [27392, 4096]
transformer.encoder.layers.9.mlp.dense_4h_to_h.weight     >> [4096, 13696]
transformer.encoder.layers.10.input_layernorm.weight      >> [4096]
transformer.encoder.layers.10.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.10.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.10.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.10.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.10.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.10.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.11.input_layernorm.weight      >> [4096]
transformer.encoder.layers.11.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.11.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.11.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.11.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.11.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.11.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.12.input_layernorm.weight      >> [4096]
transformer.encoder.layers.12.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.12.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.12.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.12.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.12.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.12.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.13.input_layernorm.weight      >> [4096]
transformer.encoder.layers.13.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.13.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.13.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.13.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.13.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.13.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.14.input_layernorm.weight      >> [4096]
transformer.encoder.layers.14.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.14.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.14.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.14.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.14.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.14.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.15.input_layernorm.weight      >> [4096]
transformer.encoder.layers.15.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.15.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.15.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.15.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.15.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.15.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.16.input_layernorm.weight      >> [4096]
transformer.encoder.layers.16.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.16.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.16.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.16.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.16.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.16.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.17.input_layernorm.weight      >> [4096]
transformer.encoder.layers.17.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.17.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.17.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.17.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.17.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.17.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.18.input_layernorm.weight      >> [4096]
transformer.encoder.layers.18.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.18.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.18.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.18.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.18.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.18.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.19.input_layernorm.weight      >> [4096]
transformer.encoder.layers.19.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.19.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.19.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.19.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.19.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.19.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.20.input_layernorm.weight      >> [4096]
transformer.encoder.layers.20.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.20.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.20.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.20.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.20.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.20.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.21.input_layernorm.weight      >> [4096]
transformer.encoder.layers.21.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.21.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.21.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.21.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.21.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.21.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.22.input_layernorm.weight      >> [4096]
transformer.encoder.layers.22.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.22.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.22.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.22.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.22.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.22.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.23.input_layernorm.weight      >> [4096]
transformer.encoder.layers.23.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.23.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.23.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.23.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.23.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.23.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.24.input_layernorm.weight      >> [4096]
transformer.encoder.layers.24.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.24.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.24.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.24.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.24.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.24.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.25.input_layernorm.weight      >> [4096]
transformer.encoder.layers.25.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.25.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.25.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.25.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.25.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.25.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.26.input_layernorm.weight      >> [4096]
transformer.encoder.layers.26.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.26.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.26.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.26.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.26.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.26.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.layers.27.input_layernorm.weight      >> [4096]
transformer.encoder.layers.27.self_attention.query_key_value.weight>> [4608, 4096]
transformer.encoder.layers.27.self_attention.query_key_value.bias>> [4608]
transformer.encoder.layers.27.self_attention.dense.weight >> [4096, 4096]
transformer.encoder.layers.27.post_attention_layernorm.weight>> [4096]
transformer.encoder.layers.27.mlp.dense_h_to_4h.weight    >> [27392, 4096]
transformer.encoder.layers.27.mlp.dense_4h_to_h.weight    >> [4096, 13696]
transformer.encoder.final_layernorm.weight                >> [4096]
transformer.output_layer.weight                           >> [65024, 4096]

Baichuan2-7B-base

model.embed_tokens.weight                                 >> [125696, 4096]
model.layers.0.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.0.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.0.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.0.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.0.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.0.input_layernorm.weight                     >> [4096]
model.layers.0.post_attention_layernorm.weight            >> [4096]
model.layers.1.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.1.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.1.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.1.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.1.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.1.input_layernorm.weight                     >> [4096]
model.layers.1.post_attention_layernorm.weight            >> [4096]
model.layers.2.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.2.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.2.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.2.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.2.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.2.input_layernorm.weight                     >> [4096]
model.layers.2.post_attention_layernorm.weight            >> [4096]
model.layers.3.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.3.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.3.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.3.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.3.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.3.input_layernorm.weight                     >> [4096]
model.layers.3.post_attention_layernorm.weight            >> [4096]
model.layers.4.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.4.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.4.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.4.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.4.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.4.input_layernorm.weight                     >> [4096]
model.layers.4.post_attention_layernorm.weight            >> [4096]
model.layers.5.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.5.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.5.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.5.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.5.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.5.input_layernorm.weight                     >> [4096]
model.layers.5.post_attention_layernorm.weight            >> [4096]
model.layers.6.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.6.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.6.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.6.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.6.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.6.input_layernorm.weight                     >> [4096]
model.layers.6.post_attention_layernorm.weight            >> [4096]
model.layers.7.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.7.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.7.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.7.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.7.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.7.input_layernorm.weight                     >> [4096]
model.layers.7.post_attention_layernorm.weight            >> [4096]
model.layers.8.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.8.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.8.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.8.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.8.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.8.input_layernorm.weight                     >> [4096]
model.layers.8.post_attention_layernorm.weight            >> [4096]
model.layers.9.self_attn.W_pack.weight                    >> [12288, 4096]
model.layers.9.self_attn.o_proj.weight                    >> [4096, 4096]
model.layers.9.mlp.gate_proj.weight                       >> [11008, 4096]
model.layers.9.mlp.down_proj.weight                       >> [4096, 11008]
model.layers.9.mlp.up_proj.weight                         >> [11008, 4096]
model.layers.9.input_layernorm.weight                     >> [4096]
model.layers.9.post_attention_layernorm.weight            >> [4096]
model.layers.10.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.10.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.10.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.10.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.10.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.10.input_layernorm.weight                    >> [4096]
model.layers.10.post_attention_layernorm.weight           >> [4096]
model.layers.11.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.11.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.11.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.11.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.11.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.11.input_layernorm.weight                    >> [4096]
model.layers.11.post_attention_layernorm.weight           >> [4096]
model.layers.12.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.12.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.12.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.12.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.12.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.12.input_layernorm.weight                    >> [4096]
model.layers.12.post_attention_layernorm.weight           >> [4096]
model.layers.13.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.13.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.13.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.13.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.13.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.13.input_layernorm.weight                    >> [4096]
model.layers.13.post_attention_layernorm.weight           >> [4096]
model.layers.14.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.14.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.14.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.14.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.14.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.14.input_layernorm.weight                    >> [4096]
model.layers.14.post_attention_layernorm.weight           >> [4096]
model.layers.15.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.15.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.15.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.15.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.15.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.15.input_layernorm.weight                    >> [4096]
model.layers.15.post_attention_layernorm.weight           >> [4096]
model.layers.16.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.16.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.16.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.16.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.16.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.16.input_layernorm.weight                    >> [4096]
model.layers.16.post_attention_layernorm.weight           >> [4096]
model.layers.17.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.17.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.17.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.17.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.17.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.17.input_layernorm.weight                    >> [4096]
model.layers.17.post_attention_layernorm.weight           >> [4096]
model.layers.18.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.18.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.18.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.18.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.18.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.18.input_layernorm.weight                    >> [4096]
model.layers.18.post_attention_layernorm.weight           >> [4096]
model.layers.19.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.19.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.19.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.19.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.19.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.19.input_layernorm.weight                    >> [4096]
model.layers.19.post_attention_layernorm.weight           >> [4096]
model.layers.20.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.20.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.20.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.20.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.20.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.20.input_layernorm.weight                    >> [4096]
model.layers.20.post_attention_layernorm.weight           >> [4096]
model.layers.21.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.21.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.21.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.21.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.21.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.21.input_layernorm.weight                    >> [4096]
model.layers.21.post_attention_layernorm.weight           >> [4096]
model.layers.22.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.22.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.22.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.22.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.22.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.22.input_layernorm.weight                    >> [4096]
model.layers.22.post_attention_layernorm.weight           >> [4096]
model.layers.23.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.23.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.23.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.23.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.23.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.23.input_layernorm.weight                    >> [4096]
model.layers.23.post_attention_layernorm.weight           >> [4096]
model.layers.24.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.24.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.24.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.24.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.24.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.24.input_layernorm.weight                    >> [4096]
model.layers.24.post_attention_layernorm.weight           >> [4096]
model.layers.25.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.25.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.25.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.25.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.25.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.25.input_layernorm.weight                    >> [4096]
model.layers.25.post_attention_layernorm.weight           >> [4096]
model.layers.26.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.26.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.26.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.26.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.26.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.26.input_layernorm.weight                    >> [4096]
model.layers.26.post_attention_layernorm.weight           >> [4096]
model.layers.27.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.27.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.27.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.27.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.27.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.27.input_layernorm.weight                    >> [4096]
model.layers.27.post_attention_layernorm.weight           >> [4096]
model.layers.28.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.28.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.28.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.28.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.28.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.28.input_layernorm.weight                    >> [4096]
model.layers.28.post_attention_layernorm.weight           >> [4096]
model.layers.29.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.29.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.29.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.29.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.29.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.29.input_layernorm.weight                    >> [4096]
model.layers.29.post_attention_layernorm.weight           >> [4096]
model.layers.30.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.30.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.30.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.30.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.30.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.30.input_layernorm.weight                    >> [4096]
model.layers.30.post_attention_layernorm.weight           >> [4096]
model.layers.31.self_attn.W_pack.weight                   >> [12288, 4096]
model.layers.31.self_attn.o_proj.weight                   >> [4096, 4096]
model.layers.31.mlp.gate_proj.weight                      >> [11008, 4096]
model.layers.31.mlp.down_proj.weight                      >> [4096, 11008]
model.layers.31.mlp.up_proj.weight                        >> [11008, 4096]
model.layers.31.input_layernorm.weight                    >> [4096]
model.layers.31.post_attention_layernorm.weight           >> [4096]
model.norm.weight                                         >> [4096]
lm_head.weight                                            >> [125696, 4096]

Qwen 7B

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    trust_remote_code=True,
    torch_dtype=torch.float16,
    device_map={"":0}
)
for name, params in model.named_parameters():
    print(f'{name: <58s}>> {list(params.shape)}')

transformer.wte.weight                                    >> [151936, 4096]
transformer.h.0.ln_1.weight                               >> [4096]
transformer.h.0.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.0.attn.c_attn.bias                          >> [12288]
transformer.h.0.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.0.ln_2.weight                               >> [4096]
transformer.h.0.mlp.w1.weight                             >> [11008, 4096]
transformer.h.0.mlp.w2.weight                             >> [11008, 4096]
transformer.h.0.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.1.ln_1.weight                               >> [4096]
transformer.h.1.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.1.attn.c_attn.bias                          >> [12288]
transformer.h.1.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.1.ln_2.weight                               >> [4096]
transformer.h.1.mlp.w1.weight                             >> [11008, 4096]
transformer.h.1.mlp.w2.weight                             >> [11008, 4096]
transformer.h.1.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.2.ln_1.weight                               >> [4096]
transformer.h.2.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.2.attn.c_attn.bias                          >> [12288]
transformer.h.2.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.2.ln_2.weight                               >> [4096]
transformer.h.2.mlp.w1.weight                             >> [11008, 4096]
transformer.h.2.mlp.w2.weight                             >> [11008, 4096]
transformer.h.2.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.3.ln_1.weight                               >> [4096]
transformer.h.3.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.3.attn.c_attn.bias                          >> [12288]
transformer.h.3.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.3.ln_2.weight                               >> [4096]
transformer.h.3.mlp.w1.weight                             >> [11008, 4096]
transformer.h.3.mlp.w2.weight                             >> [11008, 4096]
transformer.h.3.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.4.ln_1.weight                               >> [4096]
transformer.h.4.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.4.attn.c_attn.bias                          >> [12288]
transformer.h.4.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.4.ln_2.weight                               >> [4096]
transformer.h.4.mlp.w1.weight                             >> [11008, 4096]
transformer.h.4.mlp.w2.weight                             >> [11008, 4096]
transformer.h.4.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.5.ln_1.weight                               >> [4096]
transformer.h.5.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.5.attn.c_attn.bias                          >> [12288]
transformer.h.5.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.5.ln_2.weight                               >> [4096]
transformer.h.5.mlp.w1.weight                             >> [11008, 4096]
transformer.h.5.mlp.w2.weight                             >> [11008, 4096]
transformer.h.5.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.6.ln_1.weight                               >> [4096]
transformer.h.6.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.6.attn.c_attn.bias                          >> [12288]
transformer.h.6.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.6.ln_2.weight                               >> [4096]
transformer.h.6.mlp.w1.weight                             >> [11008, 4096]
transformer.h.6.mlp.w2.weight                             >> [11008, 4096]
transformer.h.6.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.7.ln_1.weight                               >> [4096]
transformer.h.7.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.7.attn.c_attn.bias                          >> [12288]
transformer.h.7.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.7.ln_2.weight                               >> [4096]
transformer.h.7.mlp.w1.weight                             >> [11008, 4096]
transformer.h.7.mlp.w2.weight                             >> [11008, 4096]
transformer.h.7.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.8.ln_1.weight                               >> [4096]
transformer.h.8.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.8.attn.c_attn.bias                          >> [12288]
transformer.h.8.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.8.ln_2.weight                               >> [4096]
transformer.h.8.mlp.w1.weight                             >> [11008, 4096]
transformer.h.8.mlp.w2.weight                             >> [11008, 4096]
transformer.h.8.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.9.ln_1.weight                               >> [4096]
transformer.h.9.attn.c_attn.weight                        >> [12288, 4096]
transformer.h.9.attn.c_attn.bias                          >> [12288]
transformer.h.9.attn.c_proj.weight                        >> [4096, 4096]
transformer.h.9.ln_2.weight                               >> [4096]
transformer.h.9.mlp.w1.weight                             >> [11008, 4096]
transformer.h.9.mlp.w2.weight                             >> [11008, 4096]
transformer.h.9.mlp.c_proj.weight                         >> [4096, 11008]
transformer.h.10.ln_1.weight                              >> [4096]
transformer.h.10.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.10.attn.c_attn.bias                         >> [12288]
transformer.h.10.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.10.ln_2.weight                              >> [4096]
transformer.h.10.mlp.w1.weight                            >> [11008, 4096]
transformer.h.10.mlp.w2.weight                            >> [11008, 4096]
transformer.h.10.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.11.ln_1.weight                              >> [4096]
transformer.h.11.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.11.attn.c_attn.bias                         >> [12288]
transformer.h.11.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.11.ln_2.weight                              >> [4096]
transformer.h.11.mlp.w1.weight                            >> [11008, 4096]
transformer.h.11.mlp.w2.weight                            >> [11008, 4096]
transformer.h.11.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.12.ln_1.weight                              >> [4096]
transformer.h.12.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.12.attn.c_attn.bias                         >> [12288]
transformer.h.12.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.12.ln_2.weight                              >> [4096]
transformer.h.12.mlp.w1.weight                            >> [11008, 4096]
transformer.h.12.mlp.w2.weight                            >> [11008, 4096]
transformer.h.12.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.13.ln_1.weight                              >> [4096]
transformer.h.13.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.13.attn.c_attn.bias                         >> [12288]
transformer.h.13.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.13.ln_2.weight                              >> [4096]
transformer.h.13.mlp.w1.weight                            >> [11008, 4096]
transformer.h.13.mlp.w2.weight                            >> [11008, 4096]
transformer.h.13.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.14.ln_1.weight                              >> [4096]
transformer.h.14.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.14.attn.c_attn.bias                         >> [12288]
transformer.h.14.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.14.ln_2.weight                              >> [4096]
transformer.h.14.mlp.w1.weight                            >> [11008, 4096]
transformer.h.14.mlp.w2.weight                            >> [11008, 4096]
transformer.h.14.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.15.ln_1.weight                              >> [4096]
transformer.h.15.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.15.attn.c_attn.bias                         >> [12288]
transformer.h.15.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.15.ln_2.weight                              >> [4096]
transformer.h.15.mlp.w1.weight                            >> [11008, 4096]
transformer.h.15.mlp.w2.weight                            >> [11008, 4096]
transformer.h.15.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.16.ln_1.weight                              >> [4096]
transformer.h.16.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.16.attn.c_attn.bias                         >> [12288]
transformer.h.16.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.16.ln_2.weight                              >> [4096]
transformer.h.16.mlp.w1.weight                            >> [11008, 4096]
transformer.h.16.mlp.w2.weight                            >> [11008, 4096]
transformer.h.16.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.17.ln_1.weight                              >> [4096]
transformer.h.17.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.17.attn.c_attn.bias                         >> [12288]
transformer.h.17.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.17.ln_2.weight                              >> [4096]
transformer.h.17.mlp.w1.weight                            >> [11008, 4096]
transformer.h.17.mlp.w2.weight                            >> [11008, 4096]
transformer.h.17.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.18.ln_1.weight                              >> [4096]
transformer.h.18.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.18.attn.c_attn.bias                         >> [12288]
transformer.h.18.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.18.ln_2.weight                              >> [4096]
transformer.h.18.mlp.w1.weight                            >> [11008, 4096]
transformer.h.18.mlp.w2.weight                            >> [11008, 4096]
transformer.h.18.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.19.ln_1.weight                              >> [4096]
transformer.h.19.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.19.attn.c_attn.bias                         >> [12288]
transformer.h.19.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.19.ln_2.weight                              >> [4096]
transformer.h.19.mlp.w1.weight                            >> [11008, 4096]
transformer.h.19.mlp.w2.weight                            >> [11008, 4096]
transformer.h.19.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.20.ln_1.weight                              >> [4096]
transformer.h.20.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.20.attn.c_attn.bias                         >> [12288]
transformer.h.20.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.20.ln_2.weight                              >> [4096]
transformer.h.20.mlp.w1.weight                            >> [11008, 4096]
transformer.h.20.mlp.w2.weight                            >> [11008, 4096]
transformer.h.20.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.21.ln_1.weight                              >> [4096]
transformer.h.21.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.21.attn.c_attn.bias                         >> [12288]
transformer.h.21.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.21.ln_2.weight                              >> [4096]
transformer.h.21.mlp.w1.weight                            >> [11008, 4096]
transformer.h.21.mlp.w2.weight                            >> [11008, 4096]
transformer.h.21.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.22.ln_1.weight                              >> [4096]
transformer.h.22.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.22.attn.c_attn.bias                         >> [12288]
transformer.h.22.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.22.ln_2.weight                              >> [4096]
transformer.h.22.mlp.w1.weight                            >> [11008, 4096]
transformer.h.22.mlp.w2.weight                            >> [11008, 4096]
transformer.h.22.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.23.ln_1.weight                              >> [4096]
transformer.h.23.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.23.attn.c_attn.bias                         >> [12288]
transformer.h.23.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.23.ln_2.weight                              >> [4096]
transformer.h.23.mlp.w1.weight                            >> [11008, 4096]
transformer.h.23.mlp.w2.weight                            >> [11008, 4096]
transformer.h.23.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.24.ln_1.weight                              >> [4096]
transformer.h.24.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.24.attn.c_attn.bias                         >> [12288]
transformer.h.24.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.24.ln_2.weight                              >> [4096]
transformer.h.24.mlp.w1.weight                            >> [11008, 4096]
transformer.h.24.mlp.w2.weight                            >> [11008, 4096]
transformer.h.24.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.25.ln_1.weight                              >> [4096]
transformer.h.25.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.25.attn.c_attn.bias                         >> [12288]
transformer.h.25.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.25.ln_2.weight                              >> [4096]
transformer.h.25.mlp.w1.weight                            >> [11008, 4096]
transformer.h.25.mlp.w2.weight                            >> [11008, 4096]
transformer.h.25.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.26.ln_1.weight                              >> [4096]
transformer.h.26.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.26.attn.c_attn.bias                         >> [12288]
transformer.h.26.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.26.ln_2.weight                              >> [4096]
transformer.h.26.mlp.w1.weight                            >> [11008, 4096]
transformer.h.26.mlp.w2.weight                            >> [11008, 4096]
transformer.h.26.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.27.ln_1.weight                              >> [4096]
transformer.h.27.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.27.attn.c_attn.bias                         >> [12288]
transformer.h.27.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.27.ln_2.weight                              >> [4096]
transformer.h.27.mlp.w1.weight                            >> [11008, 4096]
transformer.h.27.mlp.w2.weight                            >> [11008, 4096]
transformer.h.27.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.28.ln_1.weight                              >> [4096]
transformer.h.28.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.28.attn.c_attn.bias                         >> [12288]
transformer.h.28.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.28.ln_2.weight                              >> [4096]
transformer.h.28.mlp.w1.weight                            >> [11008, 4096]
transformer.h.28.mlp.w2.weight                            >> [11008, 4096]
transformer.h.28.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.29.ln_1.weight                              >> [4096]
transformer.h.29.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.29.attn.c_attn.bias                         >> [12288]
transformer.h.29.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.29.ln_2.weight                              >> [4096]
transformer.h.29.mlp.w1.weight                            >> [11008, 4096]
transformer.h.29.mlp.w2.weight                            >> [11008, 4096]
transformer.h.29.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.30.ln_1.weight                              >> [4096]
transformer.h.30.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.30.attn.c_attn.bias                         >> [12288]
transformer.h.30.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.30.ln_2.weight                              >> [4096]
transformer.h.30.mlp.w1.weight                            >> [11008, 4096]
transformer.h.30.mlp.w2.weight                            >> [11008, 4096]
transformer.h.30.mlp.c_proj.weight                        >> [4096, 11008]
transformer.h.31.ln_1.weight                              >> [4096]
transformer.h.31.attn.c_attn.weight                       >> [12288, 4096]
transformer.h.31.attn.c_attn.bias                         >> [12288]
transformer.h.31.attn.c_proj.weight                       >> [4096, 4096]
transformer.h.31.ln_2.weight                              >> [4096]
transformer.h.31.mlp.w1.weight                            >> [11008, 4096]
transformer.h.31.mlp.w2.weight                            >> [11008, 4096]
transformer.h.31.mlp.c_proj.weight                        >> [4096, 11008]
transformer.ln_f.weight                                   >> [4096]
lm_head.weight                                            >> [151936, 4096]