version: '0'

services:
  vllm-openai:
    restart: always
    image: vllm/vllm-openai:latest
    container_name: custom_service
    shm_size: "32g"
    ports:
      - "8087:8087"
      - "8088:8088"
    volumes:
      - "/home/jeff/Custom_service/deploy:/root"
    entrypoint: /bin/bash /root/run.sh
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]