[Feature]: support Qwen2 embedding about vllm HOT 4 OPEN

@DavidPeleg6 - do you have capacity to add this feature? I can help guide you. It should not be too hard

from vllm.

@DavidPeleg6 Here is my quick sketch at adding support for Qwen2 embedding models #5611

This is not sufficient though since that model you shared still says it is for causal language modeling, rather than just Qwen2Model

from vllm.

@DavidPeleg6 Here is my quick sketch at adding support for Qwen2 embedding models #5611 这是我添加对 Qwen2 嵌入模型的支持的快速草图 #5611

This is not sufficient though since that model you shared still says it is for causal language modeling, rather than just Qwen2Model但这还不够，因为您共享的模型仍然表示它用于因果语言建模，而不仅仅是 Qwen2Model

I test the vllm openai server. And I found embeddings and normal chat api cannot use together.

When I use embeddings model start server, I cannot use the same chat model

from vllm.

the same bug~

from vllm.