Fix inference error in LlavaNextForConditionalGeneration with text-only input

[CatcherInThePy]

What does this PR do?

This simple fix addresses an issue where running inference with text-only input (a.k.a. without any image data ) causes an error on line 874, as image_features is None. The update restructures the checks for image_features and legacy_processing to properly handle cases where no image data is present, while preserving the integrity of the remaining processing steps. This is a suggested fix to the reported issue #35421

Fixes #35421

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.
  • It is discussed via a Github issue: Text Only input using LlaVa Next #35421
• 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?

Vision models: @amyeroberts, @qubvel

[qubvel]

cc @zucchini-nlp

[zucchini-nlp]

Hey, thanks for the PR but we already have merged a fix for LLaVA models today #34502. Feel free to update to the latest main branch :)

[CatcherInThePy]

Hello @zucchini-nlp,

That's fantastic news! Perfect timing as well ^-^ Thank you so much for the heads-up. I'm looking forward to more opportunities to contribute in the future. :) I am closing the PR.

Best.