Install Tensorflow-GPU in 5 mins - EASY!!

so today I'm going to show you how to install tensorflow GPU in under five minutes it's really straightforward and we're going to use anaconda to do it I'm going to show you how to create a new environment inside of anaconda how to install a new kernel to that environment and then we're going to use jeep the notebook to verify that our GPU is connected and working properly and that we can build and train a neural network let's have a look so as I said you obviously need anaconda if you don't have it I recommend you use it it's one of the best distributions for Python at the minute especially machine learning stuff I leave a link to this below go download the 3.7 version and then come back to this video so once you've downloaded it if it is your first time using it you're gonna need to run anaconda navigate that first because it does some configuration stuff but then close that and we're going to be using the anaconda prompt I've already got a prompt pinned to my taskbar so I'm going to just open that first thing we want to do is create a new environment now if you don't understand environments I've created another video here when I explain everything I'm not gonna go too in depth in this video because we're focusing on tensorflow but I do recommend you check that out first so what we want to do is create a new environment icon to create and we're going to call it GPU test just first tutorial followed by the version of Python that we want the environment to have three point seven in this case so that'll ask you you wanna install a bunch of stuff I'm gonna hit yes this can take a while depending on whether you have to download things or not but it's usually pretty quick there we go so now we want to activate that environment this tells the prompt that anything we execute it then inside that environment so we want to kondeh activate and our name was GPU test if you're using a Mac I believe you need to do source activate instead you might want to lock that up but it's relatively similar now that we've created the environment we need to install a kernel to the environment and this is a really important part so first we need to pip install a PI kernel and this is just basically the tool that allows us to do that okay so once that's done we no need to install the kernel and we can do this with a following command - - M and we're going to use a PI kernel make sure your Spelling's right this along command in store - - user - - name and this is where we give it the environment name so GPU test and we've also got to give it a display name I recommend keeping this the same to avoid any confusion down the road hit enter that should install the kernel there we go install kernel spec GPU test awesome so the next the next big step is to install Condor is to Condor install tensorflow GPU so we can do that that should install everything we need to run tensorflow GPU this one can take a while there's a lot that I need to install again just hit yes but I'll skip a bit ahead the video so you don't have to wait okay so that's 10th floor GPU installed now what I'm going to do is Condor install Jupiter because we're going to be using Jupiter notebook to do this I reckon I recommend you do the same just to check the GPUs working so I'm also going to be using care ass so I'm gonna pip install care apps it's important to pip install care out here because if you try and condor install it it will downgrade your tensorflow GPU and that can cause issues which a lot of people i think run into it sort pip install cameras once we've done that we want to launch keep the notebook that should open up any browser and he'll be like a essentially launches a server so we'll go back here we'll drag this over so what I've done is created a dupe the notebook that contains in GPU detection tests and a basic LS TM neural net so we can try and train it so I'll put a link to all these files in the description below so you can go download them and then what you want to do is drag the whole folder into your Jupiter directory so when you open YouTube the notebook you'll be greeted with a directory like this you want to go into anaconda 3m GPU test and basically you want to drop the file anywhere in here now you're gonna have to do it from the windows explorer obviously so again you just want to locate the same directory paste the file anywhere in here like I've already done it so they're already in there then come back to dupe the notebook and you should just be able to open that notebook inside of the server so if I colonel restart and clear out put now completely clear all the cells and first thing we want to check is in the top right is that we're using the correct colonel and this is where a lot of people go wrong so basically why we have to install a PI kernel and install kernel it used to be that Condor would default install a new kernel when you create an environment it doesn't do that anymore so a lot of people try and install tensorflow GPU without installing a kernel so what they end up doing is trying to run a TENS flow GPU code with a kernel that doesn't know about tens pee-yew so because that kernel is for a different environment so to make if you want to switch curls if you don't have the right one here you want to click kernel change kernel and you'll see all your available kernels there make sure you select the one that we created previously in my case GPU test now we're using the right kernel we can hit shift enter and we can execute cells but first I want to show you this tool so this is a sub GPU tweak and it basically shows us our GPU utilization in terms of clock speed and temperature and as you can see in a minute it's basically idling it's not being used so if we hit shift enter on our first code cell we can see that we've successfully detected the GPU we run intense flow version 1.1 3 and we're using default GPU device 0 and if you go back to anaconda prompt you should be able to see in here there we go so it's using device co which is my GeForce GTX 770 and I know I need to upgrade it's getting old but basically if we go back now to chip the notebook again you can you can see here we're idling again but it's just been used by tensorflow so now if we build this basic neural net underneath it's just an ALICE TM Network let's training on jokes that I created I'll do a tutorial on it soon so subscribe if you're interested this is where we're going to fit the model so we're going to start training it so if we open this we can see that way I'll in our GPU hit shift enter we should now see our neural net start to train and our GP you should start ramping up and there we go so we know a full clock speed phone memory clock speed and the temperature will start raising that's basically it it's that simple I mean there's there's a lot that can go wrong if you don't know what you're doing but hopefully this tutorial has been really helpful to you and now you can start taking over the world with your neural nets so have fun with it I hope I hope it's really helped you so before you go I just want to say thank you so much for watching I really do appreciate all the support if you found this video helpful please consider leaving a thumbs up and subscribing to my channel I'd really appreciate this since it helps my channel grow and I put a lot of effort into making these videos if you want to support me in other ways there's a link in the description below where you can do that but it's up to you that's about it from me I hope you all have an awesome day take it easy

Channel: thehardwareguy

Views: 90,658

Keywords: how to install tensorflow gpu, how to install tensorflow gpu on windows 10, install tensorflow gpu, install tensorflow gpu windows 10, tensorflow-gpu, tensorflow easy install, windows 10 tensorflow gpu, tensorflow gpu windows 10, jupyter notebook tensorflow gpu, tensorflow gpu installation, tensorflow gpu windows, tensorflow gpu install windows 10 anaconda, tensorflow gpu anaconda windows, tensorflow gpu python, how to install tensorflow in windows

Id: tPq6NIboLSc

Length: 8min 45sec (525 seconds)

Published: Sat Mar 23 2019