CUDA and Pytorch on Ubuntu 21.10

CUDA Installation

The installation guide can be found here.Supported GPU is listed here.
NOTE: In the official installation, the supported Ubuntu is 20.04. However, it seems the current version of CUDA (11.5.1) is compatible with Ubuntu 21.10.

My PC and OS:

• Ubuntu 21.10
• GeForce GTX 1050

Check the version of Ubuntu

yang@yzubuntu:~$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 21.10
Release:        21.10
Codename:       impish

Check the version of Linux Kernel.

yang@yzubuntu:~$ uname -r
5.13.0-23-generic

check the GPU

yang@yzubuntu:~$ lspci | grep -i nvidia
01:00.0 VGA compatible controller: NVIDIA Corporation GP107 [GeForce GTX 1050] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GP107GL High Definition Audio Controller (rev a1)

Pre-installation

gcc

yang@yzubuntu:~$ gcc --version
gcc (Ubuntu 11.2.0-7ubuntu2) 11.2.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

kernel headers and development packages

yang@yzubuntu:~$ sudo apt-get install linux-headers-$(uname -r)

Install

Download the *.deb file and install

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.5.1/local_installers/cuda-repo-ubuntu2004-11-5-local_11.5.1-495.29.05-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2004-11-5-local_11.5.1-495.29.05-1_amd64.deb
sudo apt-key add /var/cuda-repo-ubuntu2004-11-5-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda

Post-installation

The following steps must be done to make sure Ubuntu can find the location of nvcc

yang@yzubuntu:~/Downloads$ echo 'export PATH=/usr/local/cuda-11.5/bin${PATH:+:${PATH}}' >> ~/.bashrc
yang@yzubuntu:~/Downloads$ echo 'export LD_LIBRARY_PATH=/usr/local/cuda-11.5/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}' >> ~/.bashrc

Check if the process is started.

yang@yzubuntu:~/Downloads$ systemctl status nvidia-persistenced

Check CUDA version

yang@yzubuntu:~/Downloads$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Thu_Nov_18_09:45:30_PST_2021
Cuda compilation tools, release 11.5, V11.5.119
Build cuda_11.5.r11.5/compiler.30672275_0

CUDA Example

One example of using CUDA can be found here.

Wayland disappears after CUDA installation

By default, Ubuntu 21.10 uses wayland instead of X11. However, after CUDA installation, there is no option to choose wayland when logging into the system. This needs to be checked.

Pytorch

The official installation guide can be found here
NOTE: It seems Pytorch supports CUDA 11.1. My CUDA version is 11.4. The installation process did not result into error. But it needs further test with running projects.

Make a virtual environment

yang@yzubuntu:~/Desktop/pytorch_test$ virtualenv --python=/usr/local/bin/python3.8 pytorch_test
created virtual environment CPython3.8.10.final.0-64 in 232ms
  creator CPython3Posix(dest=/home/yang/Desktop/pytorch_test/pytorch_test, clear=False, no_vcs_ignore=False, global=False)
  seeder FromAppData(download=False, pip=bundle, setuptools=bundle, wheel=bundle, via=copy, app_data_dir=/home/yang/.local/share/virtualenv)
    added seed packages: pip==21.1.2, setuptools==57.0.0, wheel==0.36.2
  activators BashActivator,CShellActivator,FishActivator,PowerShellActivator,PythonActivator,XonshActivator

Activate the virtual env

yang@yzubuntu:~/Desktop/pytorch_test$ source pytorch_test/bin/activate
(pytorch_test) yang@yzubuntu:~/Desktop/pytorch_test$ pip3 install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html

Test the Installation

Create a file test.py and add the following.

import torch
x = torch.rand(5, 3)
print(x)

(pytorch_test) yang@yzubuntu:~/Desktop/pytorch_test$ python test.py 
tensor([[0.2932, 0.9057, 0.3990],
        [0.6864, 0.6373, 0.3777],
        [0.4606, 0.1394, 0.6477],
        [0.0674, 0.8054, 0.0522],
        [0.9636, 0.1747, 0.5445]])

Check Pytorch Use CUDA

Create a file test2.py and add the following.

import os
import torch
device = 'cuda' if torch.cuda.is_available() else 'cpu'
print('Using {} device'.format(device))