Skip to content

Docker nvidia

Nvidia NGC Catalog
alt text alt text alt text

Jetpack

1
docker pull nvcr.io/nvidia/l4t-jetpack:r36.4.0

Nvidia container runtime

1
sudo apt install nvidia-container-toolkit
usage
1
sudo docker run --gpus all --runtime=nvidia -it --rm nvcr.io/nvidia/l4t-jetpack:r36.4.0 nvidia-smi

Using compose

Using compose override to run nvidia docker on pc and jetson

nvidia runtime on jetson

  • jetPack includes NVIDIA Container Runtime for Jetson, which is based on nvidia-docker2, but it does not auto-assign runtime: nvidia.
  • You must explicitly request the nvidia runtime in your Docker Compose file to enable GPU access.

What Does runtime: nvidia Do?:

  • Enables GPU passthrough into the container.
  • Loads required GPU drivers and libraries into container's environment.
  • Makes nvidia-smi, CUDA, cuDNN, TensorRT available in the container.

Demo: using vscode devcontainer with docker compose

  • use nvidia/cuda:12.6.0-cudnn-runtime-ubuntu22.04 as a base image
  • use compose override to support pc and jetson running
  • check running using nvidia-smi
1
FROM nvidia/cuda:12.6.0-cudnn-runtime-ubuntu22.04
docker-compose.yaml
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
services:
  dev:
    build:
      context: .
      dockerfile: .devcontainer/Dockerfile
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: [gpu]
    volumes:
      - .:/workspace:cached
    hostname: dev
    network_mode: host
    stdin_open: true
    tty: true
docker-compose.jetson.yaml
1
2
3
4
5
6
services:
  dev:
    runtime: nvidia
    environment:
      - NVIDIA_VISIBLE_DEVICES=all
      - NVIDIA_DRIVER_CAPABILITIES=compute,utility

VScode devcontainer

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
{
    "name": "opencv_cuda",
    "dockerComposeFile": [
      "../docker-compose.yaml",
      "../docker-compose.jetson.yaml"
    ],
    "service": "dev",
    "shutdownAction": "stopCompose",
    "workspaceFolder": "/workspace",
    "customizations": {
      "vscode": {
        "extensions": [],
        "settings": {}
      }
    }
  }