manual head_dim for mixtral model

[wavy-jung]

What does this PR do?

• manual head_dim setting for mixtral model
• same as llama and mistral implementation
• worked well when I tested with sparse upcycled llama model

Fixes #34261

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?

@ArthurZucker @gante

[ArthurZucker]

Hey! Sounds good, could you make sure to fix the cis and rebase? 🤗

[wavy-jung]

@ArthurZucker I literally have no idea why the CI check fails at glm modeling code. I did not make any changes to the glm models. Can you help me with this or can you give me a brief guideline?

[ArthurZucker]

These were fixed on main! You can now rebase and should be good!

[ArthurZucker]

make fix-copies might be needed as mistral is copied here and there

[wavy-jung]

@ArthurZucker I did fix-copies & rebase. It seems all the checks are passed.

[ArthurZucker]

Thanks 🤗

[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.