Pytorch installs and uses GPU acceleration

In the previous study , We came into contact with pytorch, At that time, bloggers considered their own graphics card performance and CUDA Version and other issues , So the installation is cpu Version of , Later, bloggers need to install gpu Version of , Therefore, you can only install it again .
Bloggers are using conda install , Because use pip It doesn't seem to support
Since the blogger has already installed and configured Anoconda, So bloggers only need to create a virtual space and install it

Activate the virtual environment

conda activate your_env_name( Virtual environment name )

pytorch Installation of each version

https://pytorch.org/get-started/previous-versions/

Bloggers use cuda8, This corresponds to the following installation statement

conda install pytorch==1.0.0 torchvision==0.2.1 cuda80 -c pytorch

Installation successful

We go back to our pycharm Take a look in the middle ：

pytorch Judge gpu Is it available , Test it ：

import torch
print('CUDA edition :',torch.version.cuda)
print('Pytorch edition :',torch.__version__)
print(' Whether the graphics card is available :',' You can use ' if(torch.cuda.is_available()) else ' Unavailable ')
print(' Number of graphics cards :',torch.cuda.device_count())
print(' Of the current graphics card CUDA Calculate the force :',torch.cuda.get_device_capability(0))
print(' Current graphics card model :',torch.cuda.get_device_name(0))

Although the graphics card is very stretched , But after all, it still works perfectly