Getting Started
Beginner
20 minIntro to PyTorch in DLWΛY
Get familiar with tensors, autograd and layers in PyTorch, running on a kernel attached to DLWΛY.
PyTorch runs on a Jupyter or Kaggle kernel that you attach to DLWΛY under Settings → Compute; the in-browser Python runtime is for NumPy, pandas and scikit-learn work. With a kernel attached, this tutorial walks you through the basics.
Tensors
python
import torch # Create tensorsx = torch.tensor([1.0, 2.0, 3.0])y = torch.zeros(3, 4)z = torch.randn(2, 3) print(x, y.shape, z)Autograd
python
x = torch.tensor(2.0, requires_grad=True)y = x ** 2 + 3 * x + 1y.backward()print(x.grad) # dy/dx = 2x + 3 = 7Building Layers
python
import torch.nn as nn layer = nn.Linear(10, 5)inp = torch.randn(32, 10)out = layer(inp)print(out.shape) # [32, 5]