Just curious whether the softmax in a Fully Convolutional Network equals to Softmax2D ? As in the source code it seems to be true. But the current version of Pytorch document does not mention anything about 2D softmax.
ywu36 (Ywu36) #1
i think it is just missing in documentation.
ywu36 (Ywu36) #3
Agreed. Adding this to the document will reduce ambiguities.
added in now. sorry for the delay.