Fast Fourier transform (torch.fft.fft), How to mask/ignore some indices in tensor

1

I have input tensor :

x=torch.tensor( [ [5,3,4,2,1,4,0,0],
                  [2,3,4,2,1,0,0,0],
                  [8,3,4,0,0,0,0,0],
                  [1,3,4,2,3,1,2,0] ] )

Instead of zero , there can be any value. I want to ignore those indices.

I want to do torch.fft.fft( x, dim = -1 ) .
How can I apply a mask so It calculates before zero values or ignore those values/indices ? Should I use zeros or any other value or other solution ?
TORCH.FFT.FFT has a parameter n ( Signal length) but it trims for the whole dimension so it doesn’t work for me unless I do it row by row with zero padding.