1、下载模型
hf download Qwen/Qwen3-Coder-Next-FP8 
2、docker-compose
注意修改下模型映射路径和参数，我这是Windows环境下的docker，硬件是4*4090D 
version: '3.8'

services:
  vllm:
    image: vllm/vllm-openai:v0.15.0
    container_name: vllm-qwen3-coder-next
    ports:
      - "8000:8000"
    environment:
      - NVIDIA_VISIBLE_DEVICES=all
      - HF_ENDPOINT=https://hf-mirror.com
    volumes:
      - ${USERPROFILE}/.cache/huggingfa...