Installing cuDNN and CUDA Toolkit on Ubuntu 20.04 for Machine Learning Tasks
It is always convoluted and challenging to install a CUDA toolkit and library that needs to interact with your NVIDIA GPU on an Ubuntu machine. However, if done right, the CUDA toolkit with your NVIDIA GPU can be a great tool that can harness the power of GPU to produce fast applications.
Basic Requirements
The basic requirement for following instructions in this article is a computer with Ubuntu 20.04 installed with an NVIDIA GPU. In my case, it was NVIDIA GeForce GTX 1650 Ti.
Further, at the time of writing this article, I installed the latest version of the CUDA toolkit which was CUDA Toolkit 11.3.
The other requirement is the NVIDIA developer account that can be obtained at https://developer.nvidia.com/login free of charge.
CUDA Toolkit Installation
I decided to install a .deb file of the CUDA toolkit following the instruction I reproduced below:
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/11.3.0/local_installers/cuda-repo-ubuntu2004-11-3-local_11.3.0-465.19.01-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2004-11-3-local_11.3.0-465.19.01-1_amd64.deb
sudo apt-key add /var/cuda-repo-ubuntu2004-11-3-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda
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}}A few checks can be performed once the CUDA toolkit is installed:
systemctl status nvidia-persistencedFrom its documentation, nvidia-persistenced is intended to be run as a daemon from system initialization and is generally designed as a tool for compute-only platforms where the NVIDIA device is not used to display a graphical user interface. If the daemon is not running, you can start/restart the daemon as follows
sudo systemctl enable nvidia-persistencedTo get the version of the NVIDIA driver, type
cat /proc/driver/nvidia/versionwhich, in my case, gave output as
NVRM version: NVIDIA UNIX x86_64 Kernel Module 465.19.01 Fri Mar 19 07:44:41 UTC 2021
GCC version: gcc version 9.3.0 (Ubuntu 9.3.0-17ubuntu1~20.04)Now, we move to download cuDNN, a deep-learning library for writing applications for ML task using NVIDIA GPU:
Installing cuDNN
In order to download cuDNN libraries, you need to go to https://developer.nvidia.com/cudnn and click on the Download cuDNN button. The webpage will ask you to login into the NVIDIA developer account. After logging in and accepting their terms and conditions, you should click on the following three links:
cuDNN Runtime Library for Ubuntu20.04 x86_64 (Deb)
cuDNN Developer Library for Ubuntu20.04 x86_64 (Deb)
cuDNN Code Samples and User Guide for Ubuntu20.04 x86_64 (Deb)
which is relevant to Ubuntu 20.04 LTS. After downloading, you should have the following three .deb files:
- libcudnn8-samples_8.2.0.53–1+cuda11.3_amd64.deb
- libcudnn8-dev_8.2.0.53–1+cuda11.3_amd64.deb
- libcudnn8_8.2.0.53–1+cuda11.3_amd64.deb
that you can install using dpkg commands:
sudo dpkg -i libcudnn8_8.2.0.53-1+cuda11.3_amd64.deb
sudo dpkg -i libcudnn8-dev_8.2.0.53-1+cuda11.3_amd64.deb
sudo dpkg -i libcudnn8-samples_8.2.0.53-1+cuda11.3_amd64.debNext, you should type the following to see if your cuDNN communicates with the NVIDIA driver:
nvidia-smiIf you get an error and a message saying unable to communicate, you probably have secure boot enabled that can be turned off by typing:
sudo mokutil --disable-validationTyping the above command will ask you to enter a password. Reboot. Then enter the password. And note down the password. Usually, the system will ask you to enter specific letters of your password. Once you successfully verify your password, the secure boot will be disabled. Now, reboot your computer and in the terminal, type as follows
nvidia-smiand you should see the following output:
Fri Apr 30 21:19:53 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 465.19.01 Driver Version: 465.19.01 CUDA Version: 11.3 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... On | 00000000:01:00.0 Off | N/A |
| N/A 41C P8 3W / N/A | 825MiB / 3914MiB | 7% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1084 G /usr/lib/xorg/Xorg 133MiB |
| 0 N/A N/A 1720 G /usr/lib/xorg/Xorg 339MiB |
| 0 N/A N/A 1892 G /usr/bin/gnome-shell 68MiB |
| 0 N/A N/A 3439 C+G ...R2021a/bin/glnxa64/MATLAB 226MiB |
| 0 N/A N/A 3720 G ...1531959360FF99E4245683A9B 5MiB |
| 0 N/A N/A 4966 G ...AAAAAAAAA= --shared-files 33MiB |
+-----------------------------------------------------------------------------+Testing the installation of cuDNN
To test the installation of cuDNN, copy cuDNN samples into your home directory,
cp -r /usr/src/cudnn_samples_v8/ $HOMEChange directory to mnist example, compile the code, and run.
cd $HOME/cudnn_samples_v8/mnistCUDNN/
make
./mnistCUDNNYou should some outputs with some texts at the end stating Tests passed.
You are now ready to write deep learning applications using NVIDIA and cuDNN.
