Hi, am using pytorch lightning to train some model
and i use torch.pow function in my loss function…
then it keeps giving NAN , then i found when computing the gradients of torch.pow() function,
if the base is non-positive, the gradients its nan which makes sense for base 0, but not sure for negative values.
It can be reproduced using this:
a = torch.tensor([-2.011, -0.000001,0,1,200], requires_grad=True)
I wonder in this case, how do we deal with it?
Anybody has ideas? Thanks @ptrblck