Can vLLM be supported?

[ruo-tong]

Hi, in local evaluation, to ensure vLLM runs correctly in Docker, the --ipc=host parameter needs to be added, according to the link. I would like to ask if using vLLM for evaluation is supported?

[davin_wu]

The same questions, it may be helpful for faster inference

[yandong_yang]

Hope this suggestion can be considered