Some experience of using D2L package and related environment configuration

One introduction

written :

Beginners of deep learning
Students who lack experience in environmental error troubleshooting
Teacher Li Mu B Station machine learning course ,dive into deep learning Website Students who study for textbooks

following :

All the thunder I stepped on , Also spent a lot of time in troubleshooting mistakes , Gained some experience
I use cpu Version of pytorch
Many pits have been sorted out , Also added some own analysis , I hope I can help you

Two About anaconda Configuration of

1 anaconda Installation process

Be careful
1 It's best to install Do not install directly to c disc ( If c If the disk space is not rich )
Because it will get bigger , By the time I reinstalled the system 50g It's big
2 When installing, add a to the environment variable (add to path) You'd better check , Otherwise, it may be added to the environment variable later , A little bit of trouble

2 d2l Bao He pytorch Package download

1 Create a new environment

It's best not to put all the necessary bags directly in base In the environment
You can create a new environment
Create command :
conda create -n newd2l python=3.7
newd2l : The name of this environment is newd2l , Readers can change
py=3.7 : Created python The version is 3.7
d2l : Here is l It's lowercase l , No 1

Tips : python It is best to 3.7 edition , I installed 3.6 Version of , Some code doesn't work ( Because there are updates ) , debug Long time
  example 1 I am 3.6 Received in grad can be implicitly created only for scalar outputs The error of , however 3.7 Can run
  example 2 pip install d2l To search the d2l yes 0.17.0 Version of , Not what you need 0.17.5 Version of , There are some incompatibilities

2 Add image source

Why add an image source ?
anaconda Our server is out of the country , The download speed in China is very slow and even the download chain is broken
Using the image source can download faster and more stable

1 . to conda Add image source

According to the path , find This computer - C disc - user - Your account name - .condarc file
If not , Create a new one yourself .condarc file , The file name is empty , Just one .condarc suffix

Open it in Notepad mode
Copy and paste the following code to set up the image source

channels:
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
  - http://pypi.douban.com/simple/
  - defaults
show_channel_urls: true

Here added 3 Image source , You can find another image source and paste it in the above format

If you can't change the suffix
Try the following

The above settings are applicable conda install command To download the package

2 . to pip Add image source

stay This computer - C disc - user - Your account name Create a folder pip Folder
stay pip Create in a folder pip.ini file

Notepad mode opens
Assign and paste the following code

[global]
index-url = https://pypi.mirrors.ustc.edu.cn/simple/
extra-index-url = https://pypi.mirrors.ustc.edu.cn/simple/
		https://mirrors.aliyun.com/pypi/simple/
		https://pypi.tuna.tsinghua.edu.cn/simple/
		http://pypi.mirrors.ustc.edu.cn/simple/
		https://pypi.org/simple/
trusted-host = pypi.mirrors.ustc.edu.cn
		pypi.mirrors.ustc.edu.cn
		mirrors.aliyun.com
		pypi.tuna.tsinghua.edu.cn
		pypi.mirrors.ustc.edu.cn
		pypi.org

index-url Is to set the image source Multiple image sources are set above
trusted-host Set up trust for these websites

The above settings are applicable pip install command To download the package

3 . If you are surfing the Internet in science

You can choose Temporarily turn off scientific Internet , If you insist
It also needs to be Setting up a proxy server

1 stay .condarc file Add to

proxy_servers:
  http: http://127.0.0.1:7893
  https: http://127.0.0.1:7893
ssl_verify: false

Here 7893 Change it to your port name

2 stay pip.info file Add to

proxy    = http://127.0.0.1:7893

As shown in the figure below

3 d2l Bao He pytorch Package download

1 open anaconda prompt

2 Switch to directory 2 .2.1 The new environment you created in

3 Input pip install d2l

And then wait , If you encounter a choice yes or no , Input y , Hit enter
Wait for it to download and install

Now? d2l 0.17.5 It's already installed

4 pytorch Installation
If you are installing cpu edition Of torch , Input
conda install pytorch torchvision torchaudio cpuonly -c pytorch
that will do

If you are installing gpu edition Of torch , Step link
https://zh-v2.d2l.ai/chapter_installation/index.html
If you have any questions, please check the comment area and search by yourself
( recommend : gpu More appropriate
I am shy in my pocket , Unable to replace the old computer , No, NVIDIA The graphics card , Helpless use cpu edition )

3 kaggle

Treasure website that provides free computing power
The code on my computer appears due to the version bug But I can't find the reason ,
I temporarily used this website for code running training

How to be in kaggle How to import the required packages ? for example d2l?
Just type in
!pip install d2l
Click on the run , You can do it d2l Download this package

If you are a computer with low configuration , You can also run the code here