This article first appeared in public No. "computer vision cv"
Pytorch advantage
Why talk Pytorch, personally feel Pytorch novice kinder than Tensorflow, and more, a great counter-attack Tensorflow trend now Pytorch used in academia. Top article will last two years, the code of multi-use ratio Tensorflow Pytorch. If after you might be interested in Tensorflow, can learn Pytorch continue to learn Tensorflow, the basic principles are interlinked. Let's start with a pleasant Pytorch open learning journey!
In the Ubuntu system, or install Pytorch windows system
First open Pytorch official website: https: //pytorch.org/. In the home, we can see a variety of configuration options, here we select the CPU or the GPU version can. Generally choose GPU version of Pytorch, running large programs such speed will be much faster. And you want to run the GPU Pytorch, you need a computer (either Ubuntu system or windows system) and install the appropriate CUDA9 cudnn7, both to install sometimes encounter a lot of problems, so you need step by step to resolve. I wrote a blog recommended installation tutorial in the Ubuntu system, according to a blog written by a step by step can be friends. Blog link: HTTPS: //blog.csdn.net/xiewenrui1996/article/details/102736238 .
In fact, there is a more simple method is to install Anaconda , and then create a virtual environment Anaconda , in a virtual environment inside the installation GPU version of Pytorch , the system will automatically give you install the appropriate CUDA and cudnn, thus saving more than half time. Written procedures recommended by the compiler Pycharm, which integrates a number of modules required to write a program, simple and practical. Pycharm can be connected with the GPU version of the anaconda Pytorch in which it was created, to create virtual environments in Pycharm, the specific operation to add anaconda in the compiler which paths it.
The next update of content
The next series will update the content from scratch to learn Pytorch, such as to achieve linear regression, neural network framework code implementation, GAN etc., please sustained attention!
