jmtomczak/intro_dgm

Jupyter notebook implementations covering ten major generative model architectures—from mixture models and autoregressive transformers to VAEs, normalizing flows (RealNVP), diffusion models, score-based models, GANs, and energy-based models—each designed to run in minutes on standard hardware. Built with PyTorch and scikit-learn, the code prioritizes clarity and reproducibility, allowing developers to trace every implementation line-by-line while studying foundational concepts before extending to production systems. Includes teaching materials with assignment examples and lecture figures for educational deployment.

1,295 stars. Actively maintained with 1 commit in the last 30 days.