CUDA 11.2 Installation on Ubuntu 20.04

cuda_11.2_installation_on_Ubuntu_20.04

#!/bin/bash
## This gist contains instructions about cuda v11.2 and cudnn8.1 installation in Ubuntu 20.04 for Pytorch 1.8 & Tensorflow 2.7.0

### steps ####
# verify the system has a cuda-capable gpu
# download and install the nvidia cuda toolkit and cudnn
# setup environmental variables
# verify the installation
###

### If you have previous installation remove it first. 
sudo apt-get purge nvidia*
sudo apt remove nvidia-*
sudo rm /etc/apt/sources.list.d/cuda*
sudo apt-get autoremove && sudo apt-get autoclean
sudo rm -rf /usr/local/cuda*


### to verify your gpu is cuda enable check
lspci | grep -i nvidia

### gcc compiler is required for development using the cuda toolkit. to verify the version of gcc install enter
gcc --version

# system update
sudo apt-get update
sudo apt-get upgrade


# install other import packages
sudo apt-get install g++ freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libglu1-mesa libglu1-mesa-dev


# first get the PPA repository driver
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update

# install nvidia driver with dependencies
sudo apt install libnvidia-common-470
sudo apt install libnvidia-gl-470
sudo apt install nvidia-driver-470


sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/7fa2af80.pub
echo "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64 /" | sudo tee /etc/apt/sources.list.d/cuda.list
sudo apt-get update

 # installing CUDA-11.2
sudo apt install cuda-11-2 


# setup your paths
echo 'export PATH=/usr/local/cuda-11.2/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-11.2/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc
sudo ldconfig

# install cuDNN v8.1
# in order to download cuDNN you have to be regeistered here https://developer.nvidia.com/developer-program/signup
# then download cuDNN v8.1 form https://developer.nvidia.com/cudnn

CUDNN_TAR_FILE="cudnn-11.2-linux-x64-v8.1.1.33.tgz"
wget https://developer.nvidia.com/compute/machine-learning/cudnn/secure/8.1.1.33/11.2_20210301/cudnn-11.2-linux-x64-v8.1.1.33.tgz
tar -xzvf ${CUDNN_TAR_FILE}


# copy the following files into the cuda toolkit directory.
sudo cp -P cuda/include/cudnn.h /usr/local/cuda-11.2/include
sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda-11.2/lib64/
sudo chmod a+r /usr/local/cuda-11.2/lib64/libcudnn*

# Finally, to verify the installation, check
nvidia-smi
nvcc -V

# install Pytorch (an open source machine learning framework)
# I choose version 1.8.0 because it is stable and compatible with CUDA 11.2 Toolkit and cuDNN 8.1

pip3 install pytorch==1.8.0 torchvision==0.9.0

Thank you very much for the instructions !
The only thing that went wrong was that when I checked nvidia-smi it said it couldn't communicate with the driver. I checked the additional drivers and the 495 was installed. I manually installed 470. Is this ok?

Line 49 is not working
sudo apt install cuda-11-2

Line 49 is not working sudo apt install cuda-11-2

Change line 44 to:
sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub

Thanks, it worked

Hmm - even with line 44 changed, I get:

sudo apt install cuda-11-2
Reading package lists... Done
Building dependency tree
Reading state information... Done
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
cuda-11-2 : Depends: cuda-runtime-11-2 (>= 11.2.2) but it is not going to be installed
Depends: cuda-demo-suite-11-2 (>= 11.2.152) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

Can I use pip install torch==1.8.0 torchvision==0.9.0 instead of pip3 install pytorch==1.8.0 torchvision==0.9.0

The following packages have unmet dependencies:
cuda-11-2 : Depends: cuda-runtime-11-2 (>= 11.2.2) but it is not going to be installed
Depends: cuda-demo-suite-11-2 (>= 11.2.152) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

I am getting the same error.

Hmm - even with line 44 changed, I get:

sudo apt install cuda-11-2 Reading package lists... Done Building dependency tree Reading state information... Done Some packages could not be installed. This may mean that you have requested an impossible situation or if you are using the unstable distribution that some required packages have not yet been created or been moved out of Incoming. The following information may help to resolve the situation:

The following packages have unmet dependencies: cuda-11-2 : Depends: cuda-runtime-11-2 (>= 11.2.2) but it is not going to be installed Depends: cuda-demo-suite-11-2 (>= 11.2.152) but it is not going to be installed E: Unable to correct problems, you have held broken packages.

I had the same issue, check this page:
https://forums.developer.nvidia.com/t/cuda-install-unmet-dependencies-cuda-depends-cuda-10-0-10-0-130-but-it-is-not-going-to-be-installed/66488/5

and recheck the download link at:
https://developer.nvidia.com/cuda-11.2.0-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=2004&target_type=debnetwork

I only succeeded by installing the CUDA toolkit part from .tgz (and not installing the driver from that non-apt package)

I just discovered all these commands, so sorry that the script stopped working now. Apparently the links have changed. I will try to create an updated one when I install it in a new system.

Line 49 is not working sudo apt install cuda-11-2

Change line 44 to: sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub

its work for me

Line 64 gives error : gzip: stdin: not in gzip format
tar: Child returned status 1
tar: Error is not recoverable: exiting now

can any one help?

Line 64 gives error : gzip: stdin: not in gzip format tar: Child returned status 1 tar: Error is not recoverable: exiting now

can any one help?

This worked for me:

Using wget is not working since, you need to login to nvidia-developer to download the zip file from the archive.
Create an account in nvidia-developer and download manually.