Using Batch Normalization in discriminator is making my DC GAN model to not work. It just outputs garbage images. Both generator and discriminator loss are zero and their accuracies are 1. Any idea why? If I remove Batch Normalization layers from Discriminator, my model starts working perfectly