fix(Phi4Multimodal): Fix incorrect default vision/audio config initialization in Phi4MultimodalConfig

[charlieJ107]

🐛 Bug Fix: Phi4MultimodalConfig default sub-config initialization

This PR fixes two issues in Phi4MultimodalConfig.__init__ related to default initialization of multimodal sub-configs.

Rations in Phi4MultimodalConfig

What does this PR do?

❌ Problems fixed

1. When vision_config is None, a default Phi4MultimodalVisionConfig() was instantiated but not assigned, leaving self.vision_config as None.
2. The audio_config default initialization incorrectly checked vision_config is None instead of audio_config is None.

✅ Changes

if isinstance(vision_config, dict):
    vision_config = Phi4MultimodalVisionConfig(**vision_config)
elif vision_config is None:
    vision_config = Phi4MultimodalVisionConfig()
self.vision_config = vision_config

if isinstance(audio_config, dict):
    audio_config = Phi4MultimodalAudioConfig(**audio_config)
elif audio_config is None:
    audio_config = Phi4MultimodalAudioConfig()
self.audio_config = audio_config

🎯 Impact

• Ensures vision_config and audio_config are always valid config objects when omitted

• Prevents downstream errors caused by unexpected None values

• Improves consistency with other multimodal config implementations

No behavior change for users explicitly passing configs.

Let me know if you'd like tests added or additional adjustments.
This pull request fixes initialization logic for the vision and audio configuration objects in the Phi4MultimodalModel constructor. The changes ensure that default configuration objects are correctly assigned when none are provided.

Configuration initialization fixes:

• Corrected the assignment of default vision_config and audio_config objects in the __init__ method of modular_phi4_multimodal.py, ensuring that they are properly instantiated when not provided.

Fixes #43479

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.
  Fixing the issue: Phi4MultimodalConfig incorrectly initializes default vision/audio configs when passed as None #43479
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
  I don't think this PR need to update the documentation. Please correct me and I'll add additional submits to this PR.
• Did you write any new necessary tests?
  I don't think any additional test is needed. Please let me know if it's reuqired and I'll add additional submits to this PR.

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.
@Cyrilvallez

Note: This PR description is refined by AI, I have review this description and report and taken the full responsibility for this PR.

[Rocketknight1]

Wow, yes, this is clearly incorrect. I guess it wasn't spotted because vision/audio configs are explicitly passed in from_pretrained() most of the time. Thank you for the fix!

[github-actions]

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

run-slow: phi4_multimodal

[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.