know what we're going to do

1. Install Anaconda
2. Configure Anaconda 's environment variables
3.  Setting up Jupyter Notebook
4. Anaconda virtual environment
5.   Install the GPU version of the PyTorch library

First of all, we need to know that when Anaconda is installed, the python interpreter can be installed, so there is no need to install the python interpreter separately.

python interpreter: The name of the Python interpreter consists of Python and the version number, such as Python3.9.0. The interpreter is a tool to explain the Python language to
the computer . Only by downloading the interpreter, the computer can use Python programming. Therefore, Python programming can be done in the command prompt (cmd) after downloading Python 3.

Table of contents

1. Install Anaconda first

1.1 Mirror source installation (fast)

1.2 Install Anaconda

2. Configure environment variables

3. Set up Jupyter Notebook

3.1 Add shortcut

method

3.2 The user name is Chinese

3.3 Modify the working path of Jupyter

 method

 4. Virtual environment (very important)

4.1 Get familiar with the basic commands first

4.2 Virtual environment link Jupter kernel

5. Install the GPU version of the PyTorch library

5.1 Install PyTorch

5.2 Inspection

1. Install Anaconda first

Anaconda has been installed before, no need to uninstall, you can skip this step . It doesn't matter which versions of the python interpreter are installed in the computer, just make sure that Anaconda can also create a virtual environment.

1.1 Mirror source installation (fast)

Considering that the virtual environment will be used later, the version of the Python interpreter in the new environment can be set when creating the virtual environment, so it does not matter which version of Anaconda to download here.

1.2 Install Anaconda

1. Just me and All Users , select Just me ;
2. Select the largest disk (usually D disk) for the installation path, and place it in the newly created [ D:\Anaconda ];
3. For the last fork, uncheck the first box . (Because the environment variable can be configured by itself, the first box does not need to check Add Anaconda3 to the system PATH environment variable)

waiting...

2. Configure environment variables

• Find the environment variable in the system and click "Edit System Environment Variable".
• Click "Environment Variables" in the pop-up "System Properties" window, then select path in the pop-up "Environment Variables" window, and click Edit.
• Through the "New" button on the right, you can create a new environment variable path, and add [ D:\Anaconda ], [D:\Anaconda\Scripts ] and [ D:\Anaconda\Library\bin ] to the environment variable.

pip uninstall pyzmq
pip install pyzmq==19.0.2

 method

1. New D:\Jupyter ;
2. Open the Prompt in the desktop shortcut ;
3. Enter the jupyter notebook --generate-config command and execute it;
4. Open the configuration file address generated in the previous step, that is, C:\Users\ username \.jupyter
5. Use Ctrl + F to find in jupyter_notebook_config.py (open as Notepad)
   And modify the following configuration items:
   1. Before modification: # c.NotebookApp.notebook_dir = ''
   2. After modification: c.NotebookApp.notebook_dir = 'D:\Jupyter'
   3. That is to say, delete the # comment in front, and enter the directory path to be set in the single quotation marks behind, pay attention,

      There can be no spaces in 'D:\Jupyter' , otherwise Jupyter will crash when it opens. Save and close.
6. Find the jupyter notebook shortcut icon on the desktop, reverse the mouse button >> properties >> shortcut >> target, delete the last "%USERPROFILE%/".

 3.4 Modify font

Open address: D:\Anaconda\Lib\site-packages\notebook\static\components\codemirror\lib

Search according to your specific address 

font-family: 'Fira Code Light','Consolas';

 4. Virtual environment (very important)

We install the libraries you need in the virtual environment. There can be many virtual environments, and different libraries can be installed in different virtual environments.

# 清屏
cls
# 在base环境中下的操作
    # 列出所有的虚拟环境
    conda env list
    # 创建名为“环境名”的虚拟环境，并指定 Python 解释器的版本
    conda create -n 环境名 python=3.9
    # 删除名为“环境名”的虚拟环境
    conda remove -n 环境名 --all
    # 进入名为“环境名”的虚拟环境
    conda activate 环境名
# 虚拟环境内的操作
    # 列出当前环境下的所有库
    conda list
    # 安装 NumPy 库，并指定版本 1.12.5
    pip install numpy==1.21.5 -i https://pypi.tuna.tsinghua.edu.cn/simple
    # 安装 Pandas 库，并指定版本 1.2.4
    pip install Pandas==1.2.4 -i https://pypi.tuna.tsinghua.edu.cn/simple
    # 安装 Matplotlib 库，并指定版本 3.5.1
    pip install Matplotlib==3.5.1 -i https://pypi.tuna.tsinghua.edu.cn/simple
    # 查看当前环境下某个库的版本（以 numpy 为例）
    pip show numpy
    # 退出虚拟环境
    conda deactivate

Create a virtual environment DL: conda create -n DL python=3.9

4.2 Virtual environment link Jupter kernel

# 列出 Jupyter 的内核列表
jupyter kernelspec list
# 安装 ipykernel
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple ipykernel
# 将虚拟环境导入 Jupyter 的 kernel 中
python -m ipykernel install --user --name=环境名
# 删除虚拟环境的 kernel 内核
jupyter kernelspec remove 环境名

5.1 Install PyTorch

In it, Ctrl + F search [ pip install torch==1.12.0 ], as shown in the figure.

 Copy that code from the web page, that is

pip install torch==1.12.0+cu113 torchvision==0.13.0+cu113 torchaudio==0.12.0 --extra-index-url https://download.pytorch.org/whl/cu113

It is considered successful if you see Successfully installed in the last few lines of code . 

If the network speed is not good, you can first download the wheel files ( torch, torchvision, torchaudio ) to the local, and use pip install to install them.

5.2 Inspection

Check in the python interpreter

Enter the DL virtual environment, enter python, and enter  import torch to import the torch library. If the installation is successful, no statement will be returned, and " >>> " will appear on the next line.

In the last step, enter torch.cuda.is_available() , and True will appear.

CPU users will display false

Testing in Jupyter

Switch to the DL kernel, enter  import torch and enter torch.cuda.is_available()