Cheat Sheet

(under construction)

Pytorch Cheat Sheet

127KB

PyTorch Cheat Sheet — PyTorch Tutorials 1.1.0 documentation.pdf

pdf

See also : numpy cheat sheet

Numpy <-> Torch Tensor

Tensors are similar to NumPy’s ndarrays, with the addition being that Tensors can also be used on a GPU to accelerate computing.

Converting a Torch Tensor to a NumPy Array

a = torch.ones(5)
print(a)
b = a.numpy()
print(b)

Converting NumPy Array to Torch Tensor

See how changing the np array changed the Torch Tensor automatically

import numpy as np
a = np.ones(5)
b = torch.from_numpy(a)
np.add(a, 1, out=a)
print(a)
print(b)

Datasets

There are several options for dataset

• load data into a numpy array, then convert this array into a torch.*Tensor

• For images, packages such as Pillow, OpenCV are useful

• (Recommend) For vision, we have created a package called torchvision

  • data loaders for common datasets such as Imagenet, CIFAR10, MNIST, etc.

Using GPU

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
net.to(device)

for epoch in range(epochs):
    for inputs, labels in train_loader:
        inputs, labels = inputs.to(device), labels.to(device)

# ...
# evaluation
with torch.no_grad():
    for inputs, labels in test_loader:
            inputs, labels = inputs.to(device), labels.to(device)
# ...

Visualizing Loss Curve

Evaluation

Model function

make it a function

Train function

Test function

Plot output image function