Distillation-Based Methods & Learning Fast Generators from Scratch

This session covers two complementary routes toward fast diffusion-based generators. The first distills a pretrained diffusion model into a few-step sampler through distribution-based and progressive distillation. The second learns fast generators from scratch, building up consistency models in discrete and continuous time, consistency trajectory models, and mean flow.

Chapters 10 or 11 of The Principles of Diffusion Models.