fix(vllm): do not set videos if we don't have any

[mudler]

Description

This pull request to backend/python/vllm/backend.py includes changes to improve the handling of multi-modal data and error management in the load_image method.

Improvements to multi-modal data handling:

• async def _predict(self, request, context, streaming=False): Refactored the handling of multi_modal_data to use a dictionary that is conditionally populated with image and video data, simplifying the code and making it more readable.

Error management improvements:

• def load_image(self, image_path: str): Modified the method to return None instead of calling load_video when an error occurs, ensuring that the method's behavior is more predictable and consistent.

Notes for Reviewers

This fixes a bug where if a model didn't support both video and image understanding would fail. Seems vLLM is sensible and if there is a "video" key in the args it will check if the model capabilities include video understanding and fail otherwise.

Signed commits

• Yes, I signed my commits.

[netlify]

✅ Deploy Preview for localai ready!

Name	Link
🔨 Latest commit	abadd47
🔍 Latest deploy log	https://app.netlify.com/sites/localai/deploys/6714d02ad26e540008c757ba
😎 Deploy Preview	https://deploy-preview-3885--localai.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.