AI Dynamics

Global AI News Aggregator

About

DeepCache accelerates SDXL generation to 19 iterations per second

Faster vanilla SDXL with DeepCache in diffusers is interesting. @lucataco93 has got this up on Replicate: https://
replicate.com/lucataco/sdxl-
deepcache
… 1024×1024 at 25 steps seems to take ~3s on A100 (19it/s). More on DeepCache: https://
huggingface.co/papers/2312.00
858

→ View original post on X — @fofrai