fix Adafactor optim on torch2.5 and fix compatibility

[tenacioustommy]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers. By the way, if you're not familiar with how to use pre-commit to fix lint issues or add unit tests, please refer to Contributing to OpenMMLab.

Motivation

The Adafactor optimizer in PyTorch 2.5 is not working correctly.

Modification

Added hardcode to rename PyTorch's Adafactor optimizer to torch_Adafactor and added a warning for users using Adafactor.

BC-breaking (Optional)

Compatible with PyTorch and versions below 2.4.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
3. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDetection or MMPretrain.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[CLAassistant]

All committers have signed the CLA.

[tonyhoo]

Thank you for your contribution! The AutoGluon team also encountered the same issue when upgrading to Torch 2.5. Do we have a timeline for the release of the change?