ImportError with Torch 2.9.0 / CUDA 12.8 when running ONNX export

[ayan4m1]

System Info

optimum==2.0.0
optimum-onnx==0.0.1
torch==2.9.0+cu128
torchvision==0.24.0+cu128

Windows 11 25H2

Python 3.12

Using a virtualenv

Who can help?

@echarlaix @IlyasMoutawwakil

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction (minimal, reproducible, runnable)

I created a venv, activated it, and did:

pip install "optimum-onnx[onnxruntime]"

optimum-cli export onnx --model prithivMLmods/DeepCaption-VLA-V2.0-7B --task visual-question-answering DeepCaption-VLA-V2.0-7B-ONNX/

And got the error:

Traceback (most recent call last):
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/bin/optimum-cli", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/optimum/commands/optimum_cli.py", line 219, in main
    service.run()
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/optimum/commands/export/onnx.py", line 256, in run
    from optimum.exporters.onnx import main_export
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/transformers/utils/import_utils.py", line 2292, in __getattr__
    module = self._get_module(self._class_to_module[name])
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/transformers/utils/import_utils.py", line 2322, in _get_module
    raise e
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/transformers/utils/import_utils.py", line 2320, in _get_module
    return importlib.import_module("." + module_name, self.__name__)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.12/importlib/__init__.py", line 90, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/optimum/exporters/onnx/__main__.py", line 28, in <module>
    from optimum.exporters.onnx.convert import onnx_export_from_model
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/optimum/exporters/onnx/convert.py", line 35, in <module>
    from optimum.exporters.onnx.base import OnnxConfig
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/optimum/exporters/onnx/base.py", line 39, in <module>
    from optimum.exporters.onnx.model_patcher import DecoderModelPatcher, ModelPatcher, Seq2SeqModelPatcher
  File "/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/optimum/exporters/onnx/model_patcher.py", line 25, in <module>
    from torch.onnx.symbolic_opset14 import (
ImportError: cannot import name '_attention_scale' from 'torch.onnx.symbolic_opset14' (/media/ayan4m1/StableDiffusion/Clarissa-Reasoning-20B/venv/lib/python3.12/site-packages/torch/onnx/symbolic_opset14.py)

Expected behavior

No error.

[IlyasMoutawwakil]

Thanks for the catch ! Torch 2.9 was just released with major changes, I'm adding support in #82
Will release a patch once merged 🤗

[ayan4m1]

Figured as much, really appreciate the quick turnaround. Just wanted to make sure it wasn't something else. Thanks!