Complete Guide for GPU setup on Ubuntu system for deep learning. Installation of NVIDIA-Driver,NVCC,CUDA, CUDNN.

Incase of any queries mail to machinelearningdeeplearning628@gmail.com

We have 4 major components to be completed for the installation.

1. Installation of NIVIDIA-DRIVER
2. Installation of NVCC
3. Installation Of CUDA
4. Installation of CUDNN

What is NVIDIA-DRIVER ??

• The NVIDIA Driver is the software driver for NVIDIA Graphics GPU installed on the PC.
• It is a program used to communicate from the OS to the device.
• This software is required in most cases for the hardware device to function properly.

What is NVCC ??

• The GPU code is implemented as a collection of functions in a language that is essentially C++.
• For Compilation we use NVCC. It is Nvidia Cuda Compiler.

What is Cuda ??

• Compute Unified Device Architecture proposed by NVIDIA for general purpose programming using GPU.
• It is a parallel computing platform and API from NVIDIA that lets developers use languages they are familiar with — C, C++ to build general purpose applications for graphics processing units.

What is CUDNN ??

• NVIDIA CUDA Deep Neural Network library (cuDNN) is a GPU-accelerated library of primitives for deep neural networks.
• It allows them to focus on training neural networks and developing software applications rather than spending time on low-level GPU performance

Major Doubt —

Nvidia-smi shows CUDA version, but CUDA is not installed ??

• Be aware that the CUDA VERSION displayed by nvidia-smi associated with newer drivers is the DRIVER API COMPATIBILITY VERSION.
• It does not indicate anything at all about what CUDA version is actually installed.
• For example: A 410.72 driver will display CUDA VERSION 10.0 even when no CUDA toolkit is installed.

1. For NVIDIA-DRIVER installation :

Command : sudo apt install nvidia-driver-455

Most of the cases it comes default with the installation of OS. If not we can also install from the Ubuntu Software manager . Generally its better install the latest version based on the compatibility. Once the driver is installed restart the system.

command : nvidia-smi

This command should work now.

2. For NVCC installation :

Command : sudo apt-get install nvidia-tool-kit

After the installation,

Comamand : nvcc — — version

• Above command is nvcc double hiphen.

Should work giving the version of the compiler being used.

3. Cuda installation:

• Navigate to https://developer.nvidia.com/cuda-downloads

• Select the parameters based on your PC and OS.
• In my case , instructions . Understand them before use
• Navigate to https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions
• While installation [x] means select. Use enter in key-board for this.
• Except the driver, install all if the driver is already installed.

• You will find these post installation instructions . Execute these commands.

CUDNN Installation :-

• Navigate to https://developer.nvidia.com/cudnn

• Click on Download CuDNN.
• Login to the page. Set the login things up . Those are simple. You can also login with google account.

• You will get this page . Click on the download you want. Iam using ubuntu 20 clicking on the first Download link.

• Download these three. These will change on your ubuntu version. Search based on your os version and install.
• sudo dpkg -i libcudnn8_8.1.1.33–1+cuda11.2_amd64.deb
• Run this command for the run time version for de-packaging debian file.
• The above command will change in your case based on the os version file you have installed.
• sudo dpkg -i libcudnn8-dev_8.1.1.33–1+cuda11.2_amd64.deb
• Run this command for the developer version for de-packaging debian file.
• The above command will change in your case based on the os version file you have installed.

Restart the system . Everything should work fine. Incase any queries send a email . Or post a response here.