Docker and GPU agent

6

I fixed my problem.
I had to install nvidia-container-toolkit

I googled: “datalore Failed to start the machine: Status 500: {“message”:“could not select device driver "" with capabilities: [[gpu]]”}”

Solution came from this could not select device driver "" with capabilities: [[gpu]]. · Issue #1034 · NVIDIA/nvidia-docker · GitHub

current agents-config.yaml

docker:
  network: datalore-agents-network
  dataloreHost: datalore
  instances:
    - id: basic-agent
      default: true
      label: "docker-base"
      description: "docker-base"
      image: docker.io/jetbrains/datalore-agent:2023.3
    - id: gpu-agent
      default: false
      label: "gpu-agent"
      description: "gpu-agent"
      image: docker.io/jetbrains/datalore-agent:2023.3
      command: "docker"
      additionalOptions: "--gpus all"
      deviceRequests:
        #- driver: "gpu-agent"
        - capabilities: [ [ "gpu" ] ]

seems to work is the name of the game at the moment.
Docs should include nvidia-container-toolkit as a dependency.

Anders

1 Like