before installation
First look at what version should be installed, you can use the command line to see
Open CMD, enter
nvidia-smi
I saw another method mentioned by others (I tried it after installing it):
first search for NVIDIA under win
to open
this, click the system information below
to see the version
1. Install CUDA
The tutorial I used
[NVIDIA] Win10 + CUDA10 + cuDNN installation tutorial
But I am a little different from this one:
the folder when decompressing is given by default, because this folder will be automatically deleted after the installation is complete. So when you can choose which one to
install, there are three things to install: development, documentation, samples. I created a folder on the D drive to hand in CUDA10.2, and then created three folders in the folder called: Development , Documentation, Samples, and then put the loaded things into these three folders.
Problems that arise:
Windows 10 -64 encountered when installing tensorflow: the installation file directory cannot be found after cuda installation
After the installation is complete, configure the environment variables
2. Install CUDNN
Follow the above tutorial to install.
After installation, copy the CUDA10.2 file to the D drive
without configuring environment variables.
3.Anaconda
I installed it before, I don't remember how
4.pytorch installation
After reading the tutorial on the official website of pytorch
, I don’t know where to enter the command line given by the official website of pytorch. I found a tutorial written in great detail about the
configuration and installation of the PyTorch environment
. Don’t forget to add parentheses after entering the command!
Ahhh I succeeded! ! ! ! ! ! ! ! ! ! ! !
Installing such a bunch should be so that machine learning can be performed on the GPU.