CUDA unavailable with image built from Dockerfile.base

[philippewarren]

When running an image built using Dockerfile.base using the nvidia container runtime, CUDA is not available in /usr/local/cuda. In fact, there are no trace of cuda in /usr/local.
When building an image that extends the image built using Dockefile.base (with FROM) and using the nvidia container runtime as the default runtime, CUDA is not available. There are no trace of cuda in /usr/local.

Tested on L4T 32.7.1, using a Jetson Xavier NX

[hupaliot]

Hello @philippewarren,

Have you eventually found out what was going on ? I'm currently also trying to use one of these image in order to use CUDA and i'm facing the same issue.

I'm currently on L4T 32.5.0, also using a Jetson Xavier NX