Closed guidopetri closed 4 years ago
L2 loss is being left to the optimizer, since it would require unpacking the parameters of the network every time (as far as I know). I really wanted to make it part of the loss function itself, but I can't find a way to do it otherwise.
L2 loss is being left to the optimizer, since it would require unpacking the parameters of the network every time (as far as I know). I really wanted to make it part of the loss function itself, but I can't find a way to do it otherwise.