TUTORIAL · NOVICE · SKILLS.READING ≈ B1

PyTorch — a one-evening tour

slug: pytorch · 11 cards · ~25 min. gemini-3.1-pro outline · gemini-3.1-flash-lite bodies · pyodide sandboxes · CPU only

Generated Cached 30d

CARD 01 · CONCEPT

A tensor is a NumPy array that can live on a GPU and remembers how to differentiate itself. Today we just need the CPU one.

CARD 02 · SANDBOX

pyodide · seed.py

// output appears here

CARD 03 · CONCEPT

If you want PyTorch to track gradients, mark the tensor with requires_grad=True. Otherwise it's just a number.

CARD 04 · QUIZ

Which tensor below will accumulate gradients?

A.torch.tensor([1., 2., 3.])

B.torch.tensor([1., 2., 3.], requires_grad=True)

C.torch.zeros(3)

CARD 05 · SANDBOX

pyodide · seed.py

// output appears here

NEXT 6 CARDS WILL LOAD AS YOU SCROLL

card 6 · sandbox · "Backprop through a tiny MLP" — then concept → quiz → sandbox → checkpoint.