FractalGen is a PyTorch implementation of the Fractal Generative Models paper, enabling pixel-by-pixel generation of high-resolution images. The repository includes a simple implementation of the fractal generative model architecture, pre-trained models for ImageNet 64x64 and 256x256 generation, and training/evaluation scripts using PyTorch DDP. The approach differs from standard diffusion models by generating images pixel-by-pixel rather than in a single forward pass.