in the NiN code, I did not see the output softmax function. Is this on purpose?
The loss function used in d2l.train_ch5 is gluon.loss.SoftmaxCrossEntropyLoss() which deals with softmax.
in the NiN code, I did not see the output softmax function. Is this on purpose?
The loss function used in d2l.train_ch5 is gluon.loss.SoftmaxCrossEntropyLoss() which deals with softmax.