I’m trying to set up and use ramalama with cuda support. I’m following the instructions I read when I invoke man ramalama-cuda but one line had me stumped. The instructions there say to test the installation with podman run --rm --device=nvidia.com/gpu=all fedora nvidia-smi but when I do that I get Error: crun: executable file nvidia-smi not found in $PATH: No such file or directory: OCI runtime attempted to invoke a command that was not found
I’m on Fedora 44 and have a NVIDIA GeForce GTX 1060 6GB and it took me a little while, but I finally figured out that I needed to install the CUDA driver tools specifically for the driver for this card. In my case, I did dnf list --installed *nvidia* and noted that I have akmod-nvidia-580xx.x86_64 installed, so the corresponding CUDA driver tools are in xorg-x11-drv-nvidia-580xx-cuda.
After further searching, I found the COPR nvidia-container-toolkit which is quite useful, but where it says to install xorg-x11-drv-nvidia-cuda, if you’re running an older card like mine, you may have to install the corresponding CUDA tools rather than the latest generic one which has dropped support for some of the older cards.
I post this here in case it might be of use to others encountering this problem.