Skip to content

"nvidia-container only" configuration for JetPack 5.x

If you are on JetPack 5.0.2 and above, and if you plan to run your application(s) inside a docker container, you can choose to install jest the nvidia-container package to be able to run containers on your Jetson.

sudo apt install nvidia-container

Disk space used for JetPack Runtime configuration

JetPack 5.0.2
(Rel 35.1.0)
Jetson AGX Orin
Developer Kit
JetPack 5.0.2
(Rel 35.1.0)
Jetson AGX Xavier
Developer Kit
JetPack 5.0.2
(Rel 35.1.0)
Jetson Xavier NX
Developer Kit
Base L4T 5.6 GB nnn GB 5.6 GB
nvidia-container on L4T 5.9 GB nnn GB 5.9 GB
JetPack Full 16.6 GB nnn GB nnn GB

Step 1. Flash L4T

Flash your Jetson with just the regular L4T (Jetson Linux).

See this section of the former page for the detailed steps.

Step 2. Install nvidia-container

After flashing is done, boot your Jetson, complete the initial setup (OEM-config) flow if necessary.

Once ready, with Internet connection secured, execute the following.

sudo apt update
sudo apt install nvidia-container
sudo systemctl restart docker
sudo usermod -aG docker $USER
newgrp docker

Done - Verification

Check Verification (JP5) page.

How to build Docker container with CUDA and other SDKs

Refer to this Dockerfile of l4t-jetpack container (NGC link) to learn how you can build a Docker container and install NVIDIA SDKs inside the container.


Move Docker data directory to USB drive

You may want to expand your storage anyway if you have multiple container images on your Jetson.

One solution is to add a USB thumb drive to your Jetson and configure Docker daemon to use the USB drive as the data directory, to store all the overlay files.