Fix mask creations of GPTNeoX and GPT2

[vasqu]

What does this PR do?

Fixes #31943

SDPA implementations done by me have introduced a careless mistake in the mask creation, e.g. SDPA has to create a mask even when the attention mask is None, for example when cached input is passed as demonstrated in the issue mentioned above.

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.

@ArthurZucker @amyeroberts

[vasqu]

The timeouts are back 😢

[ArthurZucker]

Indeed! Can you add what was reported as an issue to the tests? 🤗

[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.

[vasqu]

@ArthurZucker I've added the tests to GPTNeoX and GPT2. Not sure if the reference to the issue is a good idea or not, just went with it.

[ArthurZucker]

Thanks for updating, indeed the order makes sense as for full attention you gotta create the mask even if None