Anaconda creates a new environment and installs the GPU version of pytorch

1. Why create a new environment

Recent deep learning uses GPU Faster training , But when the program runs, it is found that the previous installation is CPU Version of pytorch. Originally wanted to put CPU Version of pytorch Delete , Then put on a new one GPU edition pytorch. But a search , It's a lot of trouble , Afraid of problems in the process , And some code written before is also CPU Version of , I'm afraid only GPU There will be problems with the version . therefore , I just thought of Anaconda I'll build a new one gpu Version environment , And then install it inside GPU Version of pytorch.
I found various schemes on the Internet , It doesn't work well in my environment , It took a lot of synthesis to finally solve , It's recorded here .

2. install CUDA

This reference, this link , It's very good , You can go and praise the big guy . Click here to view the

3. Create a new environment

Step1： Error introduction
The previous steps are OK according to the above link , But when I create the environment locally , It's all kinds of error reporting , Error is as follows ：

solve ：
I found many on the Internet , There may be a problem with the installation environment configuration . Last , It took several explorations to succeed . First , Find... In the following figure .condarc :

Then replace the contents with the following code , Remember to https Change it to http, Or you'll make another mistake ：

channels:
  - defaults
show_channel_urls: true
channel_alias: https://mirrors.tuna.tsinghua.edu.cn/anaconda
default_channels:
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/r
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/pro
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/msys2
custom_channels:
  conda-forge: http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  msys2: http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  bioconda: http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  menpo: http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  pytorch: http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  simpleitk: http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud

Step2： Create an environment
Go to the search box and find " Anaconda Prompt ", Then turn it on ：

Then enter the code in it ：

conda create -n pytorch_gpu python=3.9
# pytroch_gpu  It's the environment name 
# python=3.9  It's what you want to install in the new environment python edition 

Pictured ：

After success , Enter the command as shown in the figure , Check whether the new environment is completed ：

4. install GPU Version of pytorch

First , Click the link to select the corresponding pytorch, Click to get pytorch, Check it as shown in the figure below, and the installation command will appear ：

secondly , open Anaconda Prompt terminal , Activate to the new environment , Then enter the command above , that will do （ You can run the command before , Turn off the computer firewall ）：

Last , Test installed GPU Version of pytorch Can I call GPU, Enter the following commands in sequence ：

If the display True, It means success ; conversely , You don't succeed , Then you can only go and see what's wrong .

5. Postscript

Environmental installation is really a very metaphysical thing , Be patient to solve ！！！