GPU not detected in custom devcontainer with nvcr.io/nvidia/pytorch:23.05-py3

[albert_kelo]

Hi,
I have created a custom container using the image: nvcr.io/nvidia/pytorch:23.05-py3.

The host computer is using Ubuntu 23.04, I have nvidia-docker2, and the nvidia drivers above 530 to be able to use CUDA. Here is the output of nvidia-smi:

The container runs well, but the drivers from the graphics card are not detected:

I have used the following devcontainer file:

and docker-compose:

Thank you for your help.
Best regards

[albert_kelo]

Just updating a bit, after some research. It may be that I need to use “sudo” command to build the image. However, how can I achieve the same without the need to use sudo?

[TomNVIDIA]

Hello,

This may be related to a reported outage with nvcr.io.

Please follow this thread: I can't pull nvidia images - #5 by TomNVIDIA

Thanks,
Tom

[albert_kelo]

Hello,
I have check and it is something related to the topic described in:

nvidia-docker seems unable to use GPU as non-root user

I have come across a potential rough edge with the nvidia docker runtime provided with Jetpack 4.2.1. All of he following is run on a TX2 module mounted on a Colorado Engineering XCarrier carrier board. I am working with a deviceQuery binary built locally from the Cuda samples provided in jetpack and I can run it successfully in any user account on the device itself. When I try to run it in a container under the root user e.g.: FROM nvcr.io/nvidia/l4t-base:r32.2 COPY deviceQuery . CMD ./de…

I will check the information from that topic to see if I can make it work.

Best regards,
Albert

[albert_kelo]

Well, I actually found that the issue is related with Docker Desktop, as shown in the link: Nvidia/cuda doesn't work on Docker Desktop but works on Docker Engine - #5 by rimelek - Docker Desktop for Linux - Docker Community Forums

Thank you.
Albert