docker-compose-nvidia-docker1.yml

version:                '3'

services:

  deeplearning:

      devices:          #this is required
        - /dev/nvidiactl
        - /dev/nvidia-uvm
        - /dev/nvidia0 #in general: /dev/nvidia# where # depends on which gpu card is wanted to be used

      networks:
        - network

      volumes:
        - "nvidia_driver_384.90:/usr/local/nvidia:ro"


## Driver Volume for CUDA Version
volumes:
  nvidia_driver_384.90:
    external:           true

networks:
  network:
    driver:             bridge