JetPack 6.0: Cannot run a base CUDA Docker container

I have flashed JetPack 6.0 on a custom board and installed Docker-CE with the Docker convenience script. I then configured and restarted the daemon :

sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

However, after that, the test “sudo docker run --rm --runtime=nvidia --gpus all nvidia/cuda:11.6.2-base-ubuntu20.04 nvidia-smi” fails:

“NVIDIA-SMI couldn’t find libnvidia-ml.so library in your system. Please make sure that the NVIDIA Display Driver is properly installed and present in your system.
Please also try adding directory that contains libnvidia-ml.so to your system PATH.”

A few comments:
a) nvidia-smi at the OS level works just fine
b) libnvidia-ml.so is located in /usr/local/cuda-12.2/targets/aarch64-linux/lib/stubs/libnvidia-ml.so. However, putting it somewhere in the path (/usr/bin or /usr/lib) does not seem to help.
c) Never had any problems with this under Jetpack 5.

Any ideas how to fix this? Your help is really appreciated, thanks!

Hi,

Please refer jetson container.
The nvidia/cuda:11.6.2-base-ubuntu20.04 image you pull is for x86 arch, which not support Jetson device.

Thanks

Hi David,
Thanks for your help! I ran the image “dustynv/cuda-python:r36.2.0” and it all worked with nvidia-smi showing all the info.
Now here’s my problem: That image has CUDA 12.2. What I need is the following:

a) Ubuntu 22.04 (so I assume Jetpack 6.0 or higher)
b) CUDA 11.7

Can I run CUDA 11.7 on Jetpack 6.0? Thanks!

Hi,

No. Please refer to related topics

Thanks

Thanks David, I appreciate the clarification!

1 Like