Pytorch implementation for our DivCo. We propose a simple yet effective regularization term named latent-augmented contrastive loss that can be applied to arbitrary conditional generative adversarial ...