fix: handle missing out_hidden_size for LLaVA models in EPD encode worker

[KrishnanPrash]

Overview:

• Fix AttributeError: 'LlavaModel' object has no attribute 'out_hidden_size' crash in EncodeWorkerHandler initialization when using LLaVA models (e.g., llava-hf/llava-1.5-7b-hf) in disaggregated EPD multimodal serving.
• Use getattr with fallback to model.config.hidden_size, then "unknown"

Fixes: DIS-1529

Summary by CodeRabbit

• Bug Fixes
  • Enhanced vision model configuration handling with improved fallback logic for better compatibility across different model types.
  • Improved logging for model configuration visibility.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 6324523 and 265bf4c.

📒 Files selected for processing (1)

• components/src/dynamo/vllm/multimodal_handlers/encode_worker_handler.py

---

Walkthrough

This pull request adds defensive programming to safely access the vision model's hidden size attribute. The change implements a fallback chain that attempts to retrieve the hidden size from multiple sources, defaulting to "unknown" if unavailable, and logs the result for transparency.

Changes

Cohort / File(s)	Summary
Multimodal Handler Enhancement components/src/dynamo/vllm/multimodal_handlers/encode_worker_handler.py	Replaces direct vision_model.out_hidden_size access with a safe lookup chain: attempts out_hidden_size, falls back to vision_model.config.hidden_size, defaults to "unknown", and logs the hidden size value for debugging.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 Bouncing through the code with care,
We've added fallbacks here and there,
When hidden sizes play hide-and-seek,
Our gentle checks are never weak!
Safe lookups make the vision clear, ✨

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the fix: handling missing out_hidden_size for LLaVA models in EPD encode worker, matching the core change.
Description check	✅ Passed	The description includes overview, problem statement, solution, and issue reference (Fixes: DIS-1529), covering all template requirements adequately.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

Tip

Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs).
Share your feedback on Discord.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}