This repository offers configurable TensorFlow implementations for Generative Adversarial Networks, covering non-saturating GAN losses, WGAN, least-squares GAN, gradient penalties, spectral normalization, batch normalization, and layer normalization. It includes neural architectures such as BigGAN, ResNet, and DCGAN, along with evaluation metrics including FID score, Inception Score, precision-recall, and KID score. The code runs on GPU, TPU, and CPUs and has been used in multiple NeurIPS and ICML publications.