Full installation of Cuda and Cudnn with Pytorch for all GPU's in Linux

hello everyone welcome back to my channel and today we will see how we can install properly Kuda toolkit drivers and cool DNN for PC and these libraries are readily required while developing programs which require High computational resources like in machine learning deep learning games development and in virtual reality and most of the Python packages like opencv Pi torch and tensorflow also use these kind of libraries to accelerate the process and to reduce the time for the computation and that's why these libraries are really useful and in order to install these libraries properly make sure you follow each step properly and because every phc has different kind of GPU and architecture so a different kind of gpus works with different versions of the Cuda libraries so make sure you watch this video Until end and follow each step properly to properly install install the libraries for your architecture and in the end we will see how we can use the python packages to work with these libraries and also we need the specific version of the Python packages that works with the specific version of the Cuda libraries so here we go first of all we will install some Cuda libraries and after that we will install cool DNN libraries and after that we will install python development package like pytorch and we will use Pi torch to load these libraries and to work with the skuda libraries and we will see how we can install the Cuda first so open your Chrome browser and type install Cuda and go on this website developer.nvideo.com and inside this we are going to install Cuda toolkit 11.8 that works properly with all type of gpus so select Linux I have a Linux operating system and I have this kind of architecture and distribution I have Ubuntu and I am using Ubuntu 20 and I want to install with the tap local thing and actually the steps are the same for each version of Ubuntu and and you can just follow the steps for your Ubuntu version so here it is written that it is for Ubuntu 2004 and here 2004 so you can simply download this copy here and open your terminal and paste it here so and the next thing you can just follow each step line by line copy and paste it here and you need to give the right password here and here because we are using the sudo command so you need to type your password and the next command is this and you need to type copy and paste so it is downloading the Cuda 11.8.0 uh for the ubunt220.04 and the next command we need is this so I am going to copy this command so I'm just fast forwarding this video for you uh so the downloading is done so the next command we need is this so I am going to paste it here and it's done so just copy and paste the next command and it will take some time here and now the we will paste the next command here and after that we will update our terminal and for this we will need to type sudo app get update and after updating we need to install the Cuda so simply you can call apt-get minus y install Cuda and you can copy and paste it here simply so I have already installed this Cuda so it will take some time for you so Cuda is already the newest version 11.8.0 so after installing make sure you have installed it properly so for this to check it you need to type Nvidia Simi and it will display a window like this in which you have the name of your GPU here and it will show the percentage how much GPU resources are used and it will show the memory for the GPU for example I have 4 gigabytes of GPU so it will show here and it means it is installed properly and the next thing we are going to install is the kudiann library uh to install Kodi and then simply go here and type install go DNN and go to this website docs and here open it and go down here you need to install some prerequisites for this first of all you need to install the Nvidia graphic drivers for your PC so simply go to the Nvidia download drivers and choose your product type and product series and product and operating system and download type and also the language and click on search so it will search the driver display driver for your architecture you can simply download it and install it I have already installed it so if I should install it again so it will main create some conflict with the old install libraries so you can simply download and and install it so this is it and the next thing we have already installed the Cuda toolkit for Linux and the next we need the jet lip libraries and for Ubuntu we need to use this command so so you can simply copy and paste it here so I have already installed this and the next thing this is where we can install the code DNN for the Linux and you need to go to the qdnn home page here and after this you need to click on download qdnn and here you need to agree and you need to install the latest version for the Korean and because it works with Cuda 11.x because we have installed the Cuda 11.8 so we need to install qdnet version 8.6.0 and this version also works here with Cuda 10.2 so if you need code DNN for older versions of Cuda so you can simply go here and check the older versions here for example go here and here is the list of all the versions that are compatible with the specific version of the Cuda so I am going to install this Korean in version so simply go here and choose install it as installer for Linux and we have this type of architecture you can simply click on it so it will download it for you so I'm going to again fast forward and that the download is complete so go back here and follow the next step here so we have already installed this from the home page so we need to click download and click the short survey and submit if you have service so please complete the survey and click on submit and accept the terms and conditions and select Korean and version that you want to install we have already a look on these things and the next thing how we can install it using Tor file I have downloaded here tor5 so to install the kodiannon with the Tor file we need to follow these steps and first of all we need to unzip the Tor file and for this go to your terminal and simply use this store ah minus xcv and paste it here and the name of your file is code DNA and press tab so it will auto complete for you so here XXX means the version of the code DNN and Cuda so we are using the code again and 8.6.0 163 and Cuda 11 so simply click it here so it will extract or unzip the tar file for you and the next thing we are going to use these commands and you need to Simply copy it and copy and these commands what actually do it will copy the following files into the Cuda directory so it will take some time I will fast forward it so uh next do you have to copy this command and paste this here so it will copy some files from the code DNN folder to the Cuda folder so here and the next command is this one so you need to select this command copy and paste it again here and press enter and now the last command is this so the installation of the code DNN is done here and yes if you install it with the Debian local installation then you need to find follow this procedure so this is the simplest procedures using the Tor file installation so now our Cuda encodnn is installed properly so next thing we are going to test this Cuda and kudiann with our python package for example with pi torch and to install the pi torch you need to again type here install Pi torch and you need to go here pytorch.org and you need to click on install and here you need to again select the proper architectures that has that works with your Cuda version so I am going to install in Linux and package if you want to install through conda so you can select conda so I am going to install it through pip so I can install so I can select here pip and Python and I have download Cuda 11.8 11.7 also works with Cuda 11.8 so you can simply select here to 11.7 and simply copy these commands and open your terminal and paste it here so yes so yes the libraries are downloaded for you so the next thing we are going to test so if the Cuda libraries are work with this Pi torch python package so for this you need to open your python you need to type python3 and now we are going to import the torch Pi torch for this you need to type import torch and to check home and to check how many gpus you have you need to type torch dot Cuda dot device count device counts and so I have one GPU in my PC so it will display one here so it means torch can be torched successfully to use the Cuda for the computation and this is it for today and next we will see how we can develop the neural network architectures using these libraries Pi torch tensorflow and how we can train our models on GPU and see you next time bye

Channel: Robotics Workshop

Views: 10,293

Keywords: cuda install ubuntu, cuda install linux, cuda install ubuntu 20.04, cuda 11.8, cuda 11.8 pytorch, cuda 11.8 cudnn, cuda 11.8 install, cuda and cudnn, install cuda and cudnn ubuntu, install pytorch with cuda, install cudnn, install cudnn ubuntu, cuda toolkit installation, cuda toolkit ubuntu, cuda enabled gpu, cuda enabled pytorch, cuda 11 supported gpu, cuda 11 install ubuntu, how to install cuda, how to install cuda and cudnn, how to install cuda for pytorch, cuda, cudnn

Id: EzI9vy9XmOQ

Length: 12min 6sec (726 seconds)

Published: Sat Nov 26 2022