How can I access cuda from a docker container other than Ubuntu 18?

I have the Nvidia Docker runtime set up but I’m still hazy on how it works. Can I use, for instance, a kali linux image and container with cuda accelerated programs?
Thanks!

Hi @jcwscience, on Jetson the containers should be based on l4t-base (or another container that’s derived/based on l4t-base) to use CUDA/GPU. You should also run the container with --runtime nvidia

I’m not familiar with the Kali container, it seems likely it would need rebuilt for ARM (aarch64) and with CUDA support. When doing that, you would probably want to change it’s base container to l4t-base.

You should use the container tag of l4t-base which matches you version of JetPack-L4T (i.e. l4t-base:r32.5.0 for L4T R32.5.0). The exception is that JetPack 4.5.1 (L4T R32.5.1) uses l4t-base:r32.5.0 because the BSP didn’t change with that update.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.