services: 
  ml-service:
    build: ./ml-service
    shm_size: '2gb'     # Limita a 2 GB de memoria compartida
    ports:
      - 7860:7860
    deploy:
      resources:
        limits:
          cpus: '0.80'  # Limita a 80% de la CPU
          memory: 6G    # Limita a 6 GB de RAM

      devices:
      - "/dev/nvidia0:/dev/nvidia0"