[pytorch] pytorch0.4.0 installation tutorial and GPU configuration collection (including test code)

windows Installation under system pytorch0.4.0, stay linux When installing under the system, you only need to set the... In the command "win_amd64" Switch to ''linux_x86_64'' that will do .（ Mostly , It depends on your system , But the method is the same ）

Python3.6+pip install cpu edition

pip install http://download.pytorch.org/whl/cpu/torch-0.4.0-cp36-cp36m-win_amd64.whl

 Python3.6+pip install gpu edition
at present gpu Versioning support cuda8.0,cuda9.0 and cuda9.1, Please select the corresponding version to download and install , Do not execute the following three commands at the same time

pip install http://download.pytorch.org/whl/cu80/torch-0.4.0-cp36-cp36m-win_amd64.whl 
pip install http://download.pytorch.org/whl/cu90/torch-0.4.0-cp36-cp36m-win_amd64.whl
pip install http://download.pytorch.org/whl/cu91/torch-0.4.0-cp36-cp36m-win_amd64.whl

1. see pytorch Whether it exists and pytorch Version of

import torch
print(torch.__version__)

 2. see cuda Is it available

print(torch.cuda.is_available())

3. see cuda edition

print(torch.version.cuda)

  

4. View current working GPU 

print(torch.cuda.current_device())

 5. see gpu Number of

print(torch.cuda.device_count())

6. Set up work GPU

print(torch.cuda.current_device())
torch.cuda.set_device(1)
print(torch.cuda.current_device())

Some methods say that the following methods can also realize switching GPU But this sentence needs to be written in "import torch" Before .

os.environ["CUDA_VISIBLE_DEVICES"] = '0'

7. see gpu name , Device index defaults from 0 Start ：

print(torch.cuda.get_device_name(0))
print(torch.cuda.get_device_name(1))