In GAN implementations, BCE loss is used to express GAN loss.
As I know, GAN loss have two KLD terms (which is 2*JSD). then how can I lead torch.nn.KLDivLoss to GAN loss ?
