[Qwen2.5-Omni] Fix dtype of cos,sin when used with flash attention

[HarryHsing]

What does this PR do?

Fixes a dtype mismatch in Qwen2.5-Omni models when Flash Attention is enabled:
sin and cos positional embeddings are now explicitly cast to the same dtype as the working tensor.

Related issue: #38451

Thanks @zucchini-nlp for validating the fix!

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[zucchini-nlp]

Thanks! Can you update the modulear_qwen_omni.py file and then run make fix-copies?

[HarryHsing]

update the modulear_qwen_omni.py

I've updated modular_qwen_omni.py and successfully ran make fix-copies.

Here is the output from the command:

python utils/check_copies.py --fix_and_overwrite
python utils/check_modular_conversion.py --fix_and_overwrite
No models files or model tests in the diff, skipping modular checks
python utils/check_dummies.py --fix_and_overwrite
python utils/check_doctest_list.py --fix_and_overwrite
python utils/check_docstrings.py --fix_and_overwrite

No files were updated during make fix-copies.

Let me know if there's anything else needed. Thanks!

[zucchini-nlp]

Thanks a lot!

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.