error in /opt/nvidia/nvidia_entrypoint.sh

[perler]

after building the container I get

/opt/nvidia/nvidia_entrypoint.sh: line 67: exec: .: cannot execute: Is a directory

Looks similar to https://github.com/PromtEngineer/localGPT/issues/507 but the solution doesn't work.

I realised that I built the image before I installed the CUDA drivers and toolkit. Is there a need (and a way) to completely restart the process or is there another problem here?

[spicedreams]

I had the same problem and I built the image after installing the CUDA drivers and toolkit. Turns out I had a trailing full stop on the line starting 'docker run'.