Table of contents
NVIDIA GPU driver
Check the computer graphics card type (Computer - Properties - Device Manager - Display Adapter)
Enter NVIDIA official website ( official driver | NVIDIA ) to download the corresponding driver,
Then install it by default.
After the installation is complete, check your driver version (NVIDIA Control Panel - System Information - Components)
CUDA ToolKit
Before installing CUDA, first look at
Before installing CUDA, first go to the pytorch official website ( Start Locally | PyTorch ) to check which versions of cuda Pytorch supports:
Select the version you want to install, and download the corresponding CUDA from: CUDA Toolkit Archive | NVIDIA Developer
Of course, this CUDA version not only needs to support pytorch, but also must support the graphics card driver, which can be viewed in CUDA 12.1 Update 1 Release Notes (nvidia.com)
Hidden
View the Cudnn corresponding to the current CUDA version, and decompress it:
Copy these three folders to the cuda installation directory
At the same time, modify the environment variable
Anaconda
Download URL ( Free Download | Anaconda )
It can be installed by default after downloading.
Create a virtual environment in Anaconda:
conda create -n pytorch python=3.9
Set the mirror source (Ali source), search for .condarc in the user interface of the C drive. The replacement content is:
channels:
- defaults
show_channel_urls: true
default_channels:
- http://mirrors.aliyun.com/anaconda/pkgs/main
- http://mirrors.aliyun.com/anaconda/pkgs/r
- http://mirrors.aliyun.com/anaconda/pkgs/msys2
custom_channels:
conda-forge: http://mirrors.aliyun.com/anaconda/cloud
msys2: http://mirrors.aliyun.com/anaconda/cloud
bioconda: http://mirrors.aliyun.com/anaconda/cloud
menpo: http://mirrors.aliyun.com/anaconda/cloud
pytorch: http://mirrors.aliyun.com/anaconda/cloud
simpleitk: http://mirrors.aliyun.com/anaconda/cloud
Pytorch
Go to the pytorch official website ( Start Locally | PyTorch ), find the corresponding installation version, enter the conda virtual environment, and install pytorch
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
Verify that the installation is complete