services:
  vllm:
    build:
      context: .
      dockerfile: Dockerfile
    image: rocm-vllm-inference:latest
    container_name: rocm-vllm-openai
    entrypoint: ["python"]
    command: ["-m", "app.start_openai"]
    ports:
      - "8001:8001"
    volumes:
      - /opt/model:/opt/model:ro
      - ./config.json:/workspace/config.json:ro
    environment:
      HF_HOME: /opt/model
      # 严格离线模式，禁止任何网络下载
      HF_HUB_OFFLINE: "1"
      TRANSFORMERS_OFFLINE: "1"
      HF_DATASETS_OFFLINE: "1"
      # vLLM 相关配置
      VLLM_RPC_TIMEOUT: "300"
      VLLM_WORKER_MULTIPROC_METHOD: "spawn"
      PYTHONUNBUFFERED: "1"
      # 启用 AITER 加速（R9700 属于 gfx1201 架构）
      VLLM_ROCM_USE_AITER: "1"
      VLLM_ROCM_USE_AITER_MHA: "0"
      # 增强 All-Reduce 性能（双卡通信优化）
      VLLM_ROCM_QUICK_REDUCE_QUANTIZATION: "INT4"
    devices:
      - /dev/kfd
      - /dev/dri
    group_add:
      - video
    ipc: host
    shm_size: 32g
    cap_add:
      - SYS_PTRACE
    security_opt:
      - seccomp=unconfined
    restart: unless-stopped