version: "3" services: eval: container_name: eval image: eval build: dockerfile: Dockerfile context: . volumes: # if you want to decrease your model download time, use this. - ../.cache/huggingface/:/root/.cache/huggingface/ - ./static/:/app/static/ ports: - "7500:7500" - "7501:7501" - "8000:8000" # eval port env_file: - .env deploy: resources: reservations: devices: - driver: nvidia device_ids: ["3"] # You can choose which GPU to use capabilities: [gpu]