python-pytorchHow do I install PyTorch on a Windows computer?

1. Install the PyTorch pre-compiled binary wheel according to your system configuration, e.g. for CPU-only Windows 10 and Python 3.7:

pip install torch==1.4.0+cpu torchvision==0.5.0+cpu -f https://download.pytorch.org/whl/torch_stable.html

2. Verify the installation by running a small test program such as:

import torch
x = torch.rand(5, 3)
print(x)

Output example

tensor([[0.3236, 0.9200, 0.8447],
        [0.9221, 0.7366, 0.3277],
        [0.1781, 0.9082, 0.4459],
        [0.5090, 0.9583, 0.4251],
        [0.7274, 0.9019, 0.6324]])

3. If you have a GPU, you can install the GPU version of PyTorch, e.g. for CUDA 10.1 and Python 3.7:

pip install torch===1.4.0+cu101 torchvision===0.5.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html

4. After installation, you can check if your GPU is detected by PyTorch by running the following code:

import torch
print(torch.cuda.is_available())

Output example

True

5. To ensure that PyTorch is using the GPU, you can run the following code:

import torch
x = torch.rand(5, 3)
print(x)
if torch.cuda.is_available():
  device = torch.device("cuda")
  y = torch.ones_like(x, device=device)
  x = x.to(device)
  z = x + y
  print(z)

Output example

tensor([[0.7717, 0.9653, 0.8269],
        [0.7152, 0.8271, 0.5182],
        [0.9807, 0.2540, 0.7920],
        [0.7888, 0.1350, 0.6153],
        [0.0356, 0.7261, 0.9131]])
tensor([[1.7717, 1.9653, 1.8269],
        [1.7152, 1.8271, 1.5182],
        [1.9807, 1.2540, 1.7920],
        [1.7888, 1.1350, 1.6153],
        [1.0356, 1.7261, 1.9131]], device='cuda:0')

6. To learn more about installing PyTorch on Windows, please refer to the official documentation.

7. For more detailed instructions on how to install PyTorch, please refer to the tutorials.

Edit this code on GitHub