How to install nvidia-l4t-core in docker? (original) (raw)

April 10, 2025, 12:31pm 1

I’m using Jetson Orin AGX. [ jetpack 6.2 ( l4t 36.4.3 )]
And I’m using dustynv/zed:r35.4.1 docker container.
dustynv/zed:r35.4.1
I was trying to using zed camera. But I don’t no how to solve this.
Could you help me?

root@ubuntu:~/Downloads/docker# sudo dpkg -i stereolabs-zedlink-duo_1.3.0-LI-MAX96712-ALL-L4T35.4.1_arm64.deb
dpkg: regarding stereolabs-zedlink-duo_1.3.0-LI-MAX96712-ALL-L4T35.4.1_arm64.deb containing stereolabs-zedlink-duo, pre-dependency problem:
stereolabs-zedlink-duo pre-depends on nvidia-l4t-core (>> 35.4-0)
nvidia-l4t-core is not installed.
dpkg: error processing archive stereolabs-zedlink-duo_1.3.0-LI-MAX96712-ALL-L4T35.4.1_arm64.deb (–install):
pre-dependency problem - not installing stereolabs-zedlink-duo
Errors were encountered while processing:
stereolabs-zedlink-duo_1.3.0-LI-MAX96712-ALL-L4T35.4.1_arm64.deb

I’m don’t know how to use jetson-containers run to use
nvcr.io/nvidia/l4t-jetpack:r36.4.0 as the base image and to then build zed.

This would possibly work for you.

docker pull stereolabs/zed:4.2-tools-devel-l4t-r36.4

docker run --runtime nvidia -it --network=host --privileged stereolabs/zed:4.2-tools-devel-l4t-r36.4

The –priviledged is important to get the camera into the container.

AastaLLL April 11, 2025, 3:11am 4

Hi,

nvidia-l4t-core package can be found in the below link:

Could you share more info about your use case?
It looks like you try to run a r35 container on the r36 environment.
Due to the driver’s dependencies, you will need to use an r36-based image on the JetPack 6.2 setup.

Thansk.

Oh, your right.
I was trying to run an r35.4.1 L4T-based Docker container on an r36.4.3 (JetPack 6.2) host system because I need to use both ROS 1 (Noetic) and ROS 2 (Humble).
To make it work, I attempted to manually install nvidia-l4t-core_35.4.1 by downloading the .deb file based on the official L4T sources you provided.

However, I encountered the following error during installation:

root@ubuntu:~/Downloads/docker# sudo dpkg -i nvidia-l4t-core_35.4.1-20230801124926_arm64.deb
Selecting previously unselected package nvidia-l4t-core.
(Reading database … 49522 files and directories currently installed.)
Preparing to unpack nvidia-l4t-core_35.4.1-20230801124926_arm64.deb …
/var/lib/dpkg/tmp.ci/preinst: line 59: /proc/device-tree/compatible: No such file or directory
dpkg: error processing archive nvidia-l4t-core_35.4.1-20230801124926_arm64.deb (–install):
new nvidia-l4t-core package pre-installation script subprocess returned error exit status 1
Errors were encountered while processing:
nvidia-l4t-core_35.4.1-20230801124926_arm64.deb

Or do you have any alternative solutions?

AastaLLL April 14, 2025, 5:42am 6

Hi,

As there are some drivers mounted into the container from the Jetson host.
You will need to set up the environment with JetPack 5 for the container.

The driver list can be found in the below link:

/etc/nvidia-container-runtime/host-files-for-container.d/drivers.csv

Thanks.

Here’s a Dockerfile Dockerfile ros/noetic ros2/humble l4t-jetpack r36-4-0 to run ros and ros2 on one container that runs on AGX Orin 32gb.

It includes functional ros1_bridge but you could edit that out if not needed.