Need to reinstall Drivers inside a Docker Container

Hi everyone, I’m very new to the CUDA world, so apologies in advance for anything extraordinarily dumb I’ll end up asking.
I’ve a bit of confusion regarding the usage of CUDA inside docker containers running on WSL2;
I’ve already followed the CUDA on WSL tutorial step by step, but I’m still using DockerDesktopWSL2 backend instead of Docker-CE on WSL: the tutorial containers work as expected, but I saw that, for example, the tensorflow one has a cuda-10.2 folder at /usr/local/. Does this mean the it installs its own version of the drivers, despite the 11.4 being used in the host machine?

I’m asking that since I’m trying to compile the darknet_ros ROS project, which searches for a CUDA package in its makefile and, since no cuda folder is generated by default, I was wondering if I had to reinstall the drivers inside the Image as well.

Thanks in advance!

1 Like