Qwen2 模型剪枝保存后，无法加载模型

[pennyLuo-hub]

非常感谢您的开源，我在qwen2-7B剪枝和精度验证过程中遇见以下问题：
----------------- After Pruning -----------------
Qwen2ForCausalLM(
(model): Qwen2Model(
(embed_tokens): Embedding(152064, 1792)
(layers): ModuleList(
(0-27): 28 x Qwen2DecoderLayer(
(self_attn): Qwen2Attention(
(q_proj): Linear(in_features=1792, out_features=2048, bias=True)
(k_proj): Linear(in_features=1792, out_features=512, bias=True)
(v_proj): Linear(in_features=1792, out_features=512, bias=True)
(o_proj): Linear(in_features=2048, out_features=1792, bias=False)
(rotary_emb): Qwen2RotaryEmbedding()
)
(mlp): Qwen2MLP(
(gate_proj): Linear(in_features=1792, out_features=9472, bias=False)
(up_proj): Linear(in_features=1792, out_features=9472, bias=False)
(down_proj): Linear(in_features=9472, out_features=1792, bias=False)
(act_fn): SiLU()
)
(input_layernorm): Qwen2RMSNorm((1792,), eps=1e-06)
(post_attention_layernorm): Qwen2RMSNorm((1792,), eps=1e-06)
)
)
(norm): Qwen2RMSNorm((1792,), eps=1e-06)
(rotary_emb): Qwen2RotaryEmbedding()
)
(lm_head): Linear(in_features=1792, out_features=152064, bias=False)
)
Qwen2Config {
"_attn_implementation_autoset": true,
"_name_or_path": "/data/yaotong/Torch-Pruning/examples/LLMs/Qwen/Qwen2-7B",
"architectures": [
"Qwen2ForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 1792,
"initializer_range": 0.02,
"intermediate_size": 9472,
"max_position_embeddings": 131072,
"max_window_layers": 28,
"model_type": "qwen2",
"num_attention_heads": 16,
"num_hidden_layers": 28,
"num_key_value_heads": 4,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000.0,
"sliding_window": null,
"tie_word_embeddings": false,
"torch_dtype": "float16",
"transformers_version": "4.46.3",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 152064
}

1. 加载模型时报错如下：
Traceback (most recent call last):
File "/data/Torch-Pruning/examples/LLMs/eval_ppl.py", line 272, in
main()
File "/data/Torch-Pruning/examples/LLMs/eval_ppl.py", line 237, in main
model = get_llm(args.model, args.cache_dir)
File "/data/Torch-Pruning/examples/LLMs/eval_ppl.py", line 206, in get_llm
model = AutoModelForCausalLM.from_pretrained(
File "/usr/local/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py", line 564, in from_pretrained
return model_class.from_pretrained(
File "/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py", line 4225, in from_pretrained
) = cls._load_pretrained_model(
File "/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py", line 4728, in _load_pretrained_model
new_error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model(
File "/usr/local/lib/python3.10/site-packages/transformers/modeling_utils.py", line 993, in _load_state_dict_into_meta_model
set_module_tensor_to_device(model, param_name, param_device, **set_module_kwargs)
File "/usr/local/lib/python3.10/site-packages/accelerate/utils/modeling.py", line 358, in set_module_tensor_to_device
raise ValueError(
ValueError: Trying to set a tensor of shape torch.Size([1792, 2048]) in "weight" (which has shape torch.Size([1792, 1792])), this look incorrect.