Skip to content

Instantly share code, notes, and snippets.

@jbohnslav
Created November 22, 2022 02:57
Show Gist options
  • Save jbohnslav/1867e9e3b78a192e4ebc3654929e2aa5 to your computer and use it in GitHub Desktop.
Save jbohnslav/1867e9e3b78a192e4ebc3654929e2aa5 to your computer and use it in GitHub Desktop.
Reinstall Nvidia drivers and CUDA on Ubuntu
# Uninstall all nvidia packages, old drivers, etc.
# list all the packages with nvidia in the name
dpkg -l | grep -i nvidia
sudo apt-get remove --purge nvidia*
sudo apt-get remove --purge cuda*
sudo apt-get remove --purge libnccl*
sudo apt-get remove --purge libnvidia*
sudo apt remove --purge libcudnn*
# keep going until you find no nvidia packages with dpkg -l
# Install drivers and cuda together!
# Just go to https://developer.nvidia.com/cuda-downloads
# Make sure it's the version you want
# Linux, x86_64, ubuntu, 18.04 (or the correct one for your machine), deb (network)
# Download it, cd to downloads, and type the instructions on the website
# example:
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb
sudo dpkg -i cuda-keyring_1.0-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda
# Change these lines in your .bashrc to the appropriate version
export PATH=/usr/local/cuda-11.3/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-11.3/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
# Reboot
# Verify drivers with
nvidia-smi
# Verify cuda with
nvcc --version
# Install CUDNN
# DON'T USE THE DEB FILE! Or if you do, you must install the DEVELOPER library AFTERWARDS!!
# First runtime, then developer, then code samples
https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html
# Download the code samples, and follow the post installation verification
# Involves running MNISTCudnn
# Do the post-installation actions:
# Involves adding two lines to the ~/.bashrc file
https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html
# Install NCCL: OPTIONAL
https://docs.nvidia.com/deeplearning/sdk/nccl-install-guide/index.html
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment