Fix load torch.

[ZHUI]

Before submitting

• Lint code. If there are lint issues, please format the code first.

# Install and register `pre-commit` in the project folder
pip install pre-commit && pre-commit install

# Process previous code files separately
pre-commit run --file XXXX.py

• Add test cases into tests folder. If there are codecov issues, please add tests cases first.

PR types

PR changes

Description

[paddle-bot]

Thanks for your contribution!

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull Request Overview

This PR introduces support for loading Torch weight files in two different formats within the PaddleNLP framework. It adds a new function, load_torch, to handle both standard PyTorch weight files (by converting tensors via DLPack) and safe weight files (using deserialization to convert to Paddle tensors).

[Copilot]

The function signature includes **pickle_load_args, but these arguments are not passed into torch.load. Consider passing **pickle_load_args to torch.load (e.g., state_dict = torch.load(path, map_location="cpu", weights_only=False, **pickle_load_args)) or remove the unused parameter.

Suggested change

	state_dict = torch.load(path, map_location="cpu", weights_only=False)
	state_dict = torch.load(path, map_location="cpu", weights_only=False, **pickle_load_args)