Running Kubuntu 23.10 on a Ryzen-9 processor
Followed the instructions on the NVIDIA drivers page to get “NVIDIA-Linux-x86_64-535.146.02.run” and install it.
Followed the instructions on the CUDA pages to get a local “cuda-repo-ubuntu2204-12-3-local_12.3.1-545.23.08-1_amd64.deb” and add this to the package system APT. And then managed to install the “cuda-12.3” package
Then in python3 / Jupyter-Lab I run the following code
which throws some errors, the last of which says:
“NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.”
running the command “nvidia-smi” returns the same error
What do I need to do to fix this??
Good day. Try to use this ppa Proprietary GPU Drivers : “Graphics Drivers” team
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update && sudo apt upgrade
Went a different route :-)
used the “standard” NVIDIA-535 driver so that “nvidia-smi” works
This presents the correct information and tells me it wants the CUDA 12.2
So google a lot and work out how to do that
CUDA 12.2 managed to install
Then reinstalled TensorFlow, Keras and other Python bits
Might be useful when installing CUDA xx.x to include some simple binary to show that this works. Dig into the underlying GPU and give me some facts.
----- Just Saying -----
Depending on how you install cuda, it contains sample binaries like deviceQuery which does exactly that.