[Docs] Complete missing Llama4 configuration docs

[udaymehta]

What does this PR do?

This PR addresses the TODO placeholders left in src/transformers/models/llama4/configuration_llama4.py.

The Llama4VisionConfig and Llama4TextConfig classes contained several "TODO" markers in their docstrings (specifically for MoE parameters and Vision Projector settings). I have replaced these with standard parameter descriptions consistent with the MoE architecture and similar models.

I also ran make style, which automatically reformatted a few lines in the file.

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 @Cyrilvallez @stevhliu

[Copilot]

Pull request overview

This PR completes missing documentation by replacing TODO placeholders in the Llama4 configuration file with proper parameter descriptions. The changes add documentation for MoE (Mixture of Experts) parameters in Llama4TextConfig and Vision Projector settings in Llama4VisionConfig.

Changes:

• Added documentation for vision projector parameters (vision_feature_select_strategy, pixel_shuffle_ratio, projector dimensions, dropout settings)
• Added documentation for MoE parameters (num_experts_per_tok, num_local_experts, moe_layers, router settings)
• Added documentation for attention-related parameters (attention_dropout, use_qk_norm, attention_chunk_size, floor_scale, attn_scale)
• Applied automatic code formatting via make style to several lines

[udaymehta]

Update: I did fixed type hints (from int to float where appropriate) and moe_layers type based on copilot suggestions.

However, I noticed that tests failing due to PIL.UnidentifiedImageError and timeout errors. I think it seems unrelated to the docstring changes in this PR.

[stevhliu]

thanks for filling in the todo's!

i think it'd be even more helpful if you could add to the argument description what effect changing a specific arg has. for example, what does increasing or decreasing pixel_shuffle_ratio do?

[udaymehta]

for example, what does increasing or decreasing pixel_shuffle_ratio do?

Thanks for the review, Steven!

I've updated the vague parameters (including pixel_shuffle_ratio) to explicitly explain how changing the values impacts the behavior.

I rebased on main, so it's ready for a re-review!

[stevhliu]

super nice! made some suggestions to make them a bit more concise :)

[udaymehta]

Thanks for the better wording! I've applied the necessary suggestions to each one of them and amended to previous commit to keep the commit history clean.

[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.

[stevhliu]

lgtm, thanks!

[stevhliu]

@ydshieh, can you help merge this please? this pr adds missing arg descriptions to docstrings and the failing tests are unrelated 🙏

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: llama4