llm_vector_server/docker-compose.yml

version: '3.8'

services:
  triton-server:
    build:
      context: .
      dockerfile: Dockerfile
    image: my-tritonserver:25.05-vllm-python-py3
    ports:
      - "8000:8000"
    environment:
      - TZ=Asia/Shanghai
      - PYTHONDONTWRITEBYTECODE=1
      - PYTHONUNBUFFERED=1
      - PIP_DISABLE_PIP_VERSION_CHECK=on
    # 如果需要GPU支持，取消下面的注释
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    # 如果需要挂载模型数据卷，取消下面的注释
    # volumes:
    #   - ./models:/app/models
    # 如果需要设置资源限制，取消下面的注释
    # mem_limit: 8g
    # cpus: 4.0