Speed up your diffusion models—with evolutionary caching!
Proud to share @AnirudAgg’s first paper (mentored by Matt Gwilliam)! ECAD uses evolutionary search to find Pareto-optimal schedules that speed up diffusion & boost quality—no retraining or tuning.
Read the full 🧵below
🧵 Your DiT, faster
Introducing ECAD: we reframe diffusion model caching as multi-objective optimization and evolve Pareto-optimal schedules via a genetic algorithm—achieving 4.47 FID gain at 2.58× speedup, with no retraining or tuning.
🔗 aniaggarwal.github.io/ecad
#MachineLearning
