Suppose I have a tensor t of arbitrary ndim I want to pad (with zeroes) it such that a) I introduce the fewest possible # elements b) after padding, (t.numel(
t
(t.numel(