How to add a multimodal model into vllm？

[944264716]

I want to add megrezo into vllm,but how to write here?
such as image and audio.

and my problem is: