Installing the GPU version of TensorFlow for making use of your CUDA GPU

Installing the GPU version of TensorFlow for making use of your CUDA GPU [Collection] Welcome to part nine of the Deep Learning with Neural Networks and TensorFlow tutorials. If you are going to realistically continue with deep learning, you're going to need to start using a GPU. While there exists demo data that, like the MNIST sample we used, you can successfully work with, it is not going to prepare you for many of the hardships that large datasets come with, and you wont be able to try out many of the more interesting examples of what neural networks are capable of. Thus, in this tutorial, we're going to be covering the GPU version of TensorFlow. In order to use the GPU version of TensorFlow, you will need an NVIDIA GPU with a compute capability greater than 3.0. I had been using a couple GTX 980s, which had been relatively decent, but I was not able to create models to the size that I wanted so I have bought a GTX Titan X instead, which is much more enjoyable to work with, so pay close attention to VRAM on the card. 4 GB is what I had on my 980s, which gives you 8 GB total, but 4+4 in SLI is not the same as 8GB on one card, for example. You can use two cards at once, but this is not ideal. Buy the single best GPU that you can. https://pythonprogramming.net https://twitter.com/sentdex https://www.facebook.com/pythonprogramming.net/ https://plus.google.com/+sentdex