TypeError: _unmask_unattended_patched() missing 1 required positional argument: 'unmasked_value' during ONNX export LLaMa

[nvluxiaoz]

System Info

version: latest main or release
system: any

Who can help?

No response

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)

Take any LLaMa model,
from optimum.exporters.onnx import main_export
main_export(args.model_dir,task="text-generation-with-past",output=args.output_dir,opset=17,dtype=args.dtype,device="cuda",framework="pt")

Expected behavior

ONNX exported, but the above error arises, due to HF removed the argument, but optimum did not.

[MatthewChang]

I am also seeing this.

[parsa-pico]

same here when running this code :

from optimum.onnxruntime import ORTModelForCausalLM
ORTModelForCausalLM.from_pretrained("./llama3-q", export=True, provider="CUDAExecutionProvider",)

[fxmarty]

Thank you, here is a fix #1963