YouTipThe output result is:
nonNaNSum of values: tensor(11.) dim=0 nonNaNSum of values: tensor([5., 1., 5.]) dim=1 nonNaNSum of values: tensor([3., 8.])Pytorch Torch Nansum
# Returns the sum of non-NaN values
print("nonNaNSum of values:", torch.nansum(x))
# Along dim=0 nonNaNSum of values
y = torch.tensor([[1.0,float('nan'),2.0],[4.0,1.0,3.0]])
print("dim=0 nonNaNSum of values:", torch.nansum(y, dim=0))
print("dim=1 nonNaNSum of values:", torch.nansum(y, dim=1))