What’s the difference between torch.stack() and torch.cat() functions?

stack

Concatenates sequence of tensors along a new dimension.

cat

Concatenates the given sequence of seq tensors in the given dimension.

So if A and B are of shape (3, 4), torch.cat([A, B], dim=0) will be of shape (6, 4) and torch.stack([A, B], dim=0) will be of shape (2, 3, 4).