Pytorch tensor.unfold

[ptrblck]

Yes, a nested loop could create a large overhead compared to a single operation.