Pytorch installation (GPU) in Anaconda (step on pit + fill pit)

Computer hardware and software environment

Installation time ：2022 year 2 month 27 Japan

Hardware

GPU：NVIDIA GeForce RTX 2080Ti

Software version

operating system ：Windows 10, My computer user name is Fishc, Readers can change to their own user name
Graphics card driver version ：511.79
anaconda：4.9.2, The installation position is D:\anaconda3, This is a The first pit
python：3.8.5

Brief description

View version

How to view the version ： Open the command line
Shortcut key Win + r Turn on run , Input cmd Then click OK

C:\Users\Fishc>nvidia-smi
Sun Feb 27 21:42:48 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 511.79       Driver Version: 511.79       CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ... WDDM  | 00000000:01:00.0  On |                  N/A |
| 29%   32C    P8     3W / 300W |    869MiB / 11264MiB |      3%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A       952    C+G   ...2txyewy\TextInputHost.exe    N/A      |
|    0   N/A  N/A      1268    C+G                                   N/A      |
|    0   N/A  N/A      1408    C+G   ...kyb3d8bbwe\Calculator.exe    N/A      |
|    0   N/A  N/A      3848    C+G   ...in7x64\steamwebhelper.exe    N/A      |
|    0   N/A  N/A      4852    C+G   C:\Windows\explorer.exe         N/A      |
|    0   N/A  N/A      5632    C+G   ...4__8wekyb3d8bbwe\Time.exe    N/A      |
|    0   N/A  N/A      7756    C+G   ...me\Application\chrome.exe    N/A      |
|    0   N/A  N/A      9928    C+G   ...5n1h2txyewy\SearchApp.exe    N/A      |
|    0   N/A  N/A     11008    C+G   ...bbwe\Microsoft.Photos.exe    N/A      |
|    0   N/A  N/A     11388    C+G   ...ekyb3d8bbwe\YourPhone.exe    N/A      |
|    0   N/A  N/A     11712    C+G   ...cw5n1h2txyewy\LockApp.exe    N/A      |
|    0   N/A  N/A     13692    C+G   ...108.62\msedgewebview2.exe    N/A      |
|    0   N/A  N/A     13984    C+G   ...lPanel\SystemSettings.exe    N/A      |
|    0   N/A  N/A     14376    C+G   ...perience\NVIDIA Share.exe    N/A      |
|    0   N/A  N/A     15220    C+G   ...perience\NVIDIA Share.exe    N/A      |
|    0   N/A  N/A     16620    C+G   ...y\ShellExperienceHost.exe    N/A      |
|    0   N/A  N/A     18360    C+G   ...ge\Application\msedge.exe    N/A      |
+-----------------------------------------------------------------------------+

C:\Users\Fishc>conda activate

(base) C:\Users\Fishc>conda -V
conda 4.9.2

(base) C:\Users\Fishc>python -V
Python 3.8.5

Activate the environment

Activate anaconda The environment is by command conda activate, Then the front appears (base), In the default environment (base Environmental Science ) Run in , In the following, the command line is preceded by (base), The environment is activated by default , If you want to activate other environments

C:\Users\Fishc>conda activate envs_name

(envs_name)C:\Users\Fishc>

envs_name Is the name of the environment to be activated , Modify as needed

Start formal installation

Fill the first hole

If anaconda Installed in the C Disk other than disk , So congratulations , At the end of the installation, there will be an error of insufficient permission , So change the folder permissions before installing . My folder is D:\anaconda3

Right click after the folder is selected , Open properties , Click Security , Choose SYSTEM Then click OK .

anaconda Source change

Select one of method 1 and method 2

Method 1

On the command line, enter the following command

(base) C:\Users\Fishc>conda config --set show_channel_urls yes

After running, it can be in C:\Users\Fishc The next file found is .condarc, Open this file in Notepad , Replace the contents with

channels:
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
  - 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/cloud/pytorch/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/peterjc123/
ssl_verify: true
show_channel_urls: true

It's here The second pit

Method 2

This is followed by the command line above , Run the following command , The added commands in the middle have been listed separately later , It can be copied directly

(base) C:\Users\Fishc>conda config --set show_channel_urls yes

(base) C:\Users\Fishc>conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/

(base) C:\Users\Fishc>conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/

(base) C:\Users\Fishc>conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/

(base) C:\Users\Fishc>conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/

(base) C:\Users\Fishc>conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/peterjc123/

The order is as follows

conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/peterjc123/

It's here The second pit

Fill the second pit

Each of them http No https, If you don't change here , After that, an error message will be displayed during the installation after the certificate expires .

Download and install CUDA

I downloaded it CUDA 11.3, Because when I installed it pytorch The latest display on the official website is this
pytorch Official website

Baidu can be directly cuda11.3, The first one is

Or the following website
Official website CUDA11.3
This is my choice , Click download to download , After downloading, you can use the compact installation directly ( That is, the default installation ) Just fine , The default installation location is C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3 Next , If the installation position is changed , Remember your installation location

cudnn install

The official website is as follows
cudnn Official website
To download from the official website, you need to register an account , Fill out a questionnaire , It's more convenient . I found out he didn't 11.3, So choose 11.x The latest one is OK

After downloading, unzip to get the following 3 A folder

Find the installation location just remembered , Mine is C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3

The computer will automatically merge 2 A folder with the same name , So directly unzip the 3 A folder ( It's not a document ) Copy and paste here

test CUDA Is the installation complete

Command line run

C:\Users\Fishc>cd C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3\extras\demo_suite

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3\extras\demo_suite>bandwidthTest.exe
[CUDA Bandwidth Test] - Starting...
Running on...

 Device 0: NVIDIA GeForce RTX 2080 Ti
 Quick Mode

 Host to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     12801.7

 Device to Host Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     12682.5

 Device to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     520748.6

Result = PASS

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3\extras\demo_suite>

notice Result = PASS Description has been successfully installed

Configure environment variables

double-click bin, Then copy the path

Put the above folder ( namely C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3) Medium bin Add environment variables to the directory ( Click new and paste )

Create a new anaconda Environmental Science ( Optional )

Run... On the command line , Created an environment named pytorch_envs, Just make your own name

(base) C:\Users\Fishc>conda create --name pytorch_envs python=3.8

Activate this environment

C:\Users\Fishc>conda activate pytorch_envs

(pytorch_envs)C:\Users\Fishc>

The following text appears (pytorch_envs) By default, the environment has been activated

install pytorch

go back to pytorch Its official website , Select the appropriate condition
pytorch Official website

You can get the corresponding commands that need to be run on the command line , namely conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
The third pit Here we go ~~

Fill the third pit

If you really follow the official website , Then the speed will be very touching , The previous source changes have been done in vain
The correct operation should run the following code , Get rid of the back -c pytorch

(pytorch_envs) C:\Users\Fishc>conda install pytorch torchvision torchaudio cudatoolkit=11.3

-c pytorch Download from the official website , But the official website is very slow
If just now No, newly build anaconda Environmental Science , That should be

(base) C:\Users\Fishc>conda install pytorch torchvision torchaudio cudatoolkit=11.3

installation is complete

As the installation proceeds , There will be options for the installer to choose , Just go back , Run the following test

(pytorch_envs) C:\Users\Fishc>python
Python 3.8.12 (default, Oct 12 2021, 03:01:40) [MSC v.1916 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.__version__
'1.10.2'
>>>

No error message indicates that the installation has been completed