Mllama fsdp

[blbadger]

What does this PR do?

Fixes # 35880

During the forward pass pixel values are sent do the device and datatype assigned to the patch embedding's weight rather than the device and datatype of the model. This prevents misassignment issues when training via FSDP, as the patch embedding's weight parameters are dispatched to the proper device and datatype before reference.

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?

@qubvel

[qubvel]

Thanks for the PR!

[blbadger]

@qubvel my pleasure, thanks for the suggestion! I have committed it and updated the branch.

[qubvel]

Thanks 👍

[ArthurZucker]

Thanks!