Conda+pytorch environment tutorial

Preface

conda Able to manage different Python Environmental Science , Different environments are isolated from each other , Not affecting each other . To avoid mistakes caused by the messy original environment .
That is, to the original environment copy, Is a copy of the original environment .
It can support multiple environments on one computer .

conda It is a configuration isolation python Tools for the environment

Because sometimes we need different versions python Or different versions of pip modular （ For example, you need to run two from github Up and down code , their tensorflow Version one requirement 1.0 One requirement 2.0）

conda It can be realized pip Different versions of modules coexist , It can also be realized python Different versions coexist

conda It's just a tool , It has two distributions , Namely Anaconda and Miniconda

anaconda It is a large and comprehensive set of tools , It includes many commonly used pip modular （ Such as numpy、pandas、scipy、matplotlib etc. ）, It also has graphical management tools

miniconda It's a simple conda Tools , Only bring a few python Necessary bags , Clean and pure , You can build any environment according to your own needs

miniconda Than anaconda Very small , It can realize all daily needs , Just do it yourself

Default conda Would be base Environmental Science , Of course you can put any bag you need here , If you want to create multiple isolated virtual environments , You also need to master conda Creation and switching of environment .

One 、 newly build conda Environmental Science

newly build conda Environmental Science

conda create --name yolov5 python=3.9

name Is the name of the newly created virtual environment
python=3.9 yes python Version number of , Version can be specified .
View existing environment

conde info -e

The name of my new environment is yolov5

Two 、 Configure Tsinghua TUNA Mirror source

1. Generate .condarc file

stay Anaconda prompt Command window , Input in ：

conda config --set show_channel_urls yes

After that, you can go to C:\Users\xxx see .condarc file

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

You can add Anaconda Python Free warehouse .

stay Anaconda prompt Command window to run conda clean -i Clear index cache , Make sure you use the index provided by the mirror station

3、 ... and 、 install pytorch

After activating the environment , You can install pytorch. Get into pytorch Official website selection PyTorch Official website
Web page drop-down , You can see the following figure , The official website will automatically according to your computer , What is displayed is what you can install CUDA edition , And give the installation command

Copy the commands provided by the web page

open Anaconda prompt Command window , Enter the environment you just created （ My environment name is yolov5）

conda activate yolov5

Copy the long string of commands copied from the official website

The download process will take a lot of time , And the network speed is better

Four 、 test

open Anaconda prompt Command window , Activate the environment , Input python, Get into python In the development environment

Enter the following command

import torch
torch.cuda.is_available()

Results output True success ！！

5、 ... and 、conda command

newly build conda Environmental Science

conda create --name yourEnv python=3.9

yourEnv Is the name of the newly created virtual environment , Created , It can hold anaconda Found in the directory envs/yourEnv Catalog
python=3.9： yes python Version number of . You can also specify 【python=3.6】, If not specified , The default is to install anaconda when python Version of
If you want to install... While creating the environment python Some of my bags ：

conda create -n yourEnv python=3.6 numpy pandas

Activate conda Environmental Science

windows ==> activate yourEnv
linux/mac ==> source  yourEnv

List all current environments

conda info -e
 perhaps 
conda env list

Check the package and version of the current environment

conda list

All packages for a specific environment

conda list -n env_name
install numpy sklearn package

conda install numpy scikit-learn

Delete your environment

conda env remove -n yourEnv --all

Quit current conda Environmental Science

conda deactivate

Environment switching

conda activate YourEnvs

Clone virtual environment

conda create -n newYourEnvs --clone oldYourEnvs

adopt export Command export environment configuration Will generate a .yml file

conda env export > environment.yml

open environment.yml You can see the environmental information , The following command can reinstall the same environment

conda env create -f environment.yml

Reference link

PyTorch The latest installation tutorial
About conda Configuration of the environment , Just read this one