Jul 2026
OnlineCache: Learning Dynamic Caching Policies with Error Correction for Efficient Diffusion Inference
This paper proposes OnlineCache, a dynamic caching framework that jointly learns when to cache and how to correct approximation errors, and leverages policy gradient to train a lightweight network for adaptive speed-quality trade-offs, and incorporate a learnable corrector to mitigate caching-induced errors.
Zhikang Xie, Xichen Ye, Yifan Wu et al.
· arXiv.org · 0 citations