vllm 部署 Qwen/Qwen2.5-Coder-7B-Instruct openai接口

modelscope下载模型：

bash
conda create -n modelscope python=3.10
conda activate modelscope

bash
pip install modelscope -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple

bash
modelscope download --model 'Qwen/Qwen2.5-Coder-7B-Instruct' --local_dir '/ssd/xiedong/Qwen2.5-Coder-7B-Instruct'

部署为api接口：

bash
docker run -d --gpus device=3 \
    -v /ssd/xiedong/Qwen2.5-Coder-7B-Instruct:/ssd/xiedong/Qwen2.5-Coder-7B-Instruct \
    -p 8005:8000 \
    --ipc=host \
    dockerpull.org/vllm/vllm-openai:latest \
    --model /ssd/xiedong/Qwen2.5-Coder-7B-Instruct --gpu_memory_utilization=0.4