Use Cuda 12.2 in a container

Autonomous Machines Jetson & Embedded Systems Jetson Orin NX
,
1

I have a jetson orin nx that I’ve setup using Jetpack 5.1.2. I followed these instructions to install cuda 12.2.2 1. CUDA for Tegra — CUDA for Tegra 12.3 documentation.

With this I am able to use the 12.2.2 driver and tools like nvcc. I would like to move my code to run in a container, but I have been unsuccessful. I started with the docker hub nvidia/cuda images without success. I switched to using the nvcr.io/nvidia/l4t-base images. With those images, I am able to install cuda 11.4, but I can’t get the update to 12.2.

Is there any way for me to use the 12.2 driver from within a docker container?

1 Like
3

Hi,

Please use the below instructions to install newer CUDA within the container:

Thanks.

4

I have the same problem for some reason I just can’t get cuda to update to 12.2, it works on the board directly but inside docker it simply doesn’t.
I am using the nvcr.io/nvidia/l4t-jetpack:r35.4.1 image. I tried the l4t-base docker image and the l4t-cuda docker image and none of them worked.

1 Like
5

Thank you for the suggestion, but that doesn’t work inside of the container. I am running docker run -it --rm --runtime nvidia --gpus all nvcr.io/nvidia/l4t-base:35.4.1 bash to start up the container. When I follow the linked instructions, it fails on the apt-get install cuda with:

The following packages have unmet dependencies:
 cuda : Depends: cuda-12-2 (>= 12.2.2) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.
1 Like
6

Exact same error here. I hope we can find a fix.

1 Like
7

Hi,

Thank you both for the update.
We will give it a check and update.

Thanks.

2 Likes
8

Hi,

Confirmed that we can reproduce the same error.

Our internal team will check for this.
Will update more information later.

Thanks.

1 Like
10

Tested every version available for Jetson: 11.8 , 12.0, 12.2, 12.3 all of them give the exact same error.

11

Hi,

Please check the below link to get more info:

Currently, upgrading CUDA is not working inside the container.
We recommend waiting for our next JetPack 6 release which will support CUDA 12.2.

Thanks.

12

Hi, thank you for your answers. Is there a planned dev release for JetPack 6 or any way to access the alpha/beta version?

13

Hi

Please find our release schedule below:

Thanks.

14

@jsallay @mihnea.bastea see here for a workaround to try for this:

1 Like
16

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.