Install pytoch+anaconda+cuda+cudnn

1.anaconda(5.2.0 edition )

Tsinghua mirror （ Domestic first choice ）： https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/

Official website image ：https://repo.continuum.io/archive/

After installed ,windows+r key entry cmd

Set up scikit-image=0.14.0

 

 

Check the updated package

 

 

2.Cuda+cudnn install

Check the computer graphics card configuration , Right click to enter nvidia Control panel , Check the corresponding version of cuda, Enter Google search cuda9.1, Select the download page to download

 

Google search cudnn, Select the corresponding cudnn Download the corresponding version on the download page cudnn, And copy the downloaded four folders to the following path （ Don't skip , Do not replace ）

 

windows+r key entry cmd, Input nvcc -V See if it works correctly

3. Create a file called Pytorch Environment

 

 

Input ：conda activate PyTorch Activate the environment ,

Go to Pytorch Download the required version on the official website , The installation package of the following version is downloaded here （ Key Ctrl+F You can search in the web ）

 

Then install by , Because of the network, several downloads failed , Re Download , You can enter tor Then press Tab Key auto lock

 

 

The two packages in the display environment are successfully installed

4. install opencv-python==3.4.2

 

 

Installation successful

5.pycharm To configure

Select... In the configuration environment Python.exe

 

6. Verification test

 

 

end

PS: Tsinghua mirror （ Domestic first choice ）： 

https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/

Official website image ：Index of /

conda create -n PyTorch python=3.6.5

conda info --envs

conda activate PyTorch

conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=9.0 -c pytorch

Index of /anaconda/pkgs/main/ | Tsinghua University open source software image station | Tsinghua Open Source Mirror

pip install opencv-contrib-python==3.4.2.16

cv2. __version__