The GAN loss problem

[Linfengscat]

The GAN loss binary_cross_entropy, I think, should take logits and labels both in (0,1), Maybe the author forget to add sigmoid activation in the last dense layer of Discriminator? ? ?

[forever208]

following..