multimodal support

[qiuhuiGithub]

Hi, Can vllm support multimodal LLM? such as Qwen-vl, yi-vl-6b, mplug-owl2, etc

[ruifengma]

I have the same question, currently the multimodal LLM is not supported by the framework?

[GallonDeng]

+1

[DarkLight1337]

With the merge of #6121, we will have a dev API that is relatively stable. Please refer to #4194 for more details.