[huggingface/transformers]mxmax/Chinese_Chat_T5_Base模型怎么用torch.jit.trace追踪

我使用下面的代码进行模型转换

tokenizer = AutoTokenizer.from_pretrained('./outputs/model_files/')
model = AutoModelForSeq2SeqLM.from_pretrained('./outputs/model_files/')
device = torch.device("cpu")
model.to(device)
model.eval()

tokenized_dict = tokenizer(
    ["please answer the following question: what is the boiling point of nitrogen",], ["-320.4F",], 
    return_tensors="pt"
)

input_tuple = (tokenized_dict['input_ids'], tokenized_dict['attention_mask'], torch.Tensor([[2]]).long())
traced_model = torch.jit.trace(model, input_tuple)
traced_model.save("./model.pt")

但是得到了这样的错误信息

   D:\Program Files\Python310\lib\site-packages\transformers\modeling_utils.py:701: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
   if causal_mask.shape[1] < attention_mask.shape[1]:
   Traceback (most recent call last):
   File "E:\Python\project\Chinese_Chat_T5_Base-main\convertModel.py", line 25, in <module>
traced_model = torch.jit.trace(model, input_tuple)
  File "D:\Program Files\Python310\lib\site-packages\torch\jit\_trace.py", line 759, in trace
return trace_module(
  File "D:\Program Files\Python310\lib\site-packages\torch\jit\_trace.py", line 976, in trace_module
module._c._create_method_from_trace(
RuntimeError: Tracer cannot infer type of Seq2SeqLMOutput(loss=None, logits=tensor([[[-10.4197,   6.3242,   8.7392,  ..., -10.0839,  -7.8809,  -8.4109]]],
   grad_fn=<UnsafeViewBackward0>), past_key_values=((tensor([[[[-9.3662e-02, -2.6494e-01,  2.7725e-01,  3.5019e-01,  5.3944e-01,
       -2.6313e-01, -5.9071e-01,  5.1579e-01, -5.2901e-01, -5.9420e-01,
       -9.2730e-02,  1.2436e-03, -8.6124e-01, -1.4801e-01, -6.9207e-01,
   ......

[ 2.7600e-02, -2.4005e-02, -7.1618e-02,  ...,  1.9455e-01,
       1.0591e-02, -8.1877e-02],
     [ 5.6630e-02, -2.8372e-03,  3.5540e-02,  ...,  1.0443e-01,
       3.7175e-02, -5.7037e-02],
     [-5.6965e-04,  1.0548e-04,  9.4504e-04,  ..., -1.7588e-04,
       8.6722e-04, -8.3949e-04]]], grad_fn=<MulBackward0>), encoder_hidden_states=None, encoder_attentions=None)
:Dictionary inputs to traced functions must have consistent type. Found Tensor and Tuple[Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor], Tuple[Tensor, Tensor, Tensor, Tensor]]