reopen: llava-next fails to consider padding_side during Training

[jp1924]

What does this PR do?

reopen pr: #32386

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, look good to me!

We recently merged the PR to refactor vlm code a bit and around next release we should have an option to bypass merge and pad path and merge images directly by scatter. Current warnings will be useful for those who didn't update their model on the hub

[amyeroberts]

Looks great - thanks for adding!