1. Environmentconstruction
1.1 Install Anaconda _
Anaconda is a Python distribution for scientific computing , supports Linux , Mac , Windows , and contains many popular Python packages for scientific computing and data analysis .
Anaconda download address:
Tsinghua mirror : https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/ _ _ _ _ _ _ _ _ _ _ _ _ _
Official website mirror : https://repo.anaconda.com/archive/ _ _ _ _ _ _
Here we choose Python 3.6.5 version as an example, select the corresponding Anaconda version to download, here select Anaconda 3-
5.2.0 - Windows - x86_64.exe . _ _ Specific version correspondence reference blog: blog https://blog.csdn.net/heivy/article/details/92992887?spm=1001.2101.3001.6650.2&utm_medium=distribute.pc_relevant.none-task-blog-2~default~CTRLIST ~default-2.pc_relevant_default&depth_1-utm_source=distribute.pc_relevant.none-task-blog-2~default~CTRLIST~default-2.pc_relevant_default&utm_relevant_index=5
During the installation, check the Add path option.
If the host does not have a graphics card , skip steps 1.2 , 1.3 , and 1.6. In step 1.5 , select the CPU version when installing Pytorch , and check the graphics card type. You can open it in the control panel, or find it under the windows startup item.
1.2 Check the graphics carddriver
Enter the NVIDIA control panel , check the system information under the help tab, and find your CUDA version number. If the version number is too low, update the graphics card driver first.
1.3 Download and installation of CUDA and Cudnn
Here you can choose to download CU DA 10.2. You can choose the corresponding configuration according to your current system and download it. After downloading, choose custom installation, and check all the next steps.
Then download Cudnn , address : Cudnn official website ,
Here you need to register before downloading, select the Cudnn version corresponding to CUDA 10.2 .
After the download is complete, unzip the Cudnn package, copy the contents of the folder, and paste it to the C :\ Program Files \ NVIDIA GPU Computing Toolkit \ CUDA \ v 10.1 directory.
1.4 Create PyTorch environment
Different projects require different virtual environments, which can handle incompatibility between different versions of projects .
Enter the Anaconda Prompt command window
Open the anaconda prompt tool and enter the following: conda create -n PyTorch python =3.6
As shown in the picture, mine has already been installed. If it is not installed, select y, and if it is installed, select n. PyTorch is the name of the virtual environment (--name) - n is the abbreviation, and 3.6 is the python version. Then press y to continue installing various dependent packages.
After the creation is successful, enter the following command :
conda info -- envs
You can see all your environments, including the PyTorch environment you just created .
To configure Tsinghua TUNA image source, enter in the Anaconda Prompt command window: conda config -- set show_channel_urls yes
Then you can see the .condarc file under C :\ User \ XXX , open the .condarc file with Notepad , and rewrite it as the following:
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 channels: - conda - forge - defaults |
You can add the Anaconda Python free warehouse.
1.5 Install Pytorch _
Enter PyTorch official website: PyTorch official website , the official website will automatically display the version that can be installed according to your computer, and give commands .
CUDA11.6 installation command:
conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia
CPU installation command:
conda install pytorch torchvision torchaudio cpuonly -c pytorch
Choose according to your own system. Copy the above command, open the Anaconda Prompt command window, enter the environment just created (the one created above is called PyTorch), and activate it through conda activate PyTorch.
Paste the command you just copied to download and install.
1.6 Test
Open the Anaconda Prompt command window, activate the environment, and enter the python development environment .
conda activate PyTorch
python
import torch
torch.cuda.is_available()
If it is true and the environment is normal, if it is false, you need to check the system and version.
At this point, the data can be cleaned up and ready to train your model.