This repository implements the Gated PixelCNN architecture from the paper ‘Conditional Image Generation with PixelCNN Decoders’. The model uses gated convolutions to remember context and model complex interactions for autoregressive image generation. It supports unconditional generation, class-conditional generation, and can serve as a decoder replacement for deconvolution in autoencoders and GANs. The architecture includes vertical stacks with masking to handle blind spots and residual connections to improve performance.