Cannot export Deberta to TorchScript #20815
Description
System Info
transformers-cli env
- `transformers` version: 4.10.2
- Platform: Linux-3.10.0-1127.18.2.el7.x86_64-x86_64-with-glibc2.23
- Python version: 3.9.13
- PyTorch version (GPU?): 1.9.0 (True)
- Tensorflow version (GPU?): not installed (NA)
- Flax version (CPU?/GPU?/TPU?): not installed (NA)
- Jax version: not installed
- JaxLib version: not installed
- Using GPU in script?: No
- Using distributed or parallel set-up in script?: No
Who can help?
Information
- The official example scripts
- My own modified scripts
Tasks
- An officially supported task in the
examplesfolder (such as GLUE/SQuAD, ...) - My own task or dataset (give details below)
Reproduction
I am trying to convert the Deberta Model to TorchScript using the instructions provided in the HF tutorial.
Code:
import torch
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("microsoft/deberta-base")
model = AutoModel.from_pretrained("microsoft/deberta-base", torchscript=True)
tokenized_dict = tokenizer(
["Is this working",], ["Not yet",],
return_tensors="pt"
)
input_tuple = (tokenized_dict['input_ids'], tokenized_dict['attention_mask'])
traced_model = torch.jit.trace(model, input_tuple)
torch.jit.save(traced_model, "compiled_deberta.pt")
Error Message:
From torch.jit.save:
Could not export Python function call 'XSoftmax'. Remove calls to Python functions before export. Did you forget to add @script or @script_method annotation? If this is a nn.ModuleList, add it to __constants__:
Expected behavior
The Traced model should be successfully saved. After loading, it should have the same functional behavior as the model it was traced from.