I saw that creating a Gaussian window works for 1D tensors with scipy.signal.window.gaussian, but how would it work for 3D tensors? I expect for example a tensor of shape (6
