1444141859

### Describe the issue Use vllm to launch a local large model, in the style of openai，but it won't work ### Steps to reproduce step1:python -m vllm.entrypoints.openai.api_server --max-model-len 6144 --gpu-memory-utilization...