EpaCache: Error-Propagation-Aware Caching for Accelerating Diffusion-Based Visual Generation
Artificial Intelligence
Summary
The authors propose EpaCache, a method to speed up image and video generation models without losing quality. Unlike earlier methods that reuse computations based only on short-term changes, EpaCache considers the overall impact of reusing these computations across the entire process. Their experiments show that EpaCache makes the models faster while maintaining or even improving image quality compared to previous caching methods. This approach requires no additional training and works well on multiple datasets.
Authors
Yuhan Liu, Zongwei Hong, Jinglun Li, Linze Li, Shen Zhang, Yao Tang
Abstract
Diffusion-based visual generative models deliver strong image and video synthesis quality but incur high inference costs because sequential samplers repeatedly evaluate large networks. Caching-based methods reduce inference latency by reusing intermediate computations across adjacent timesteps. However, existing cache controllers rely primarily on local temporal variation and overlook the trajectory-level consequences of cache reuse. We introduce Error-Propagation-Aware Cache (EpaCache), a training-free caching policy that adaptively allocates the reuse budget on timesteps with lower downstream impact. Experiments on image and video synthesis models demonstrate that EpaCache consistently improves the latency--fidelity trade-off over existing caching methods. On FLUX.1-dev, EpaCache outperforms the prior state-of-the-art caching method in both latency and fidelity, reducing inference time from $11.7$ s to $11.3$ s while improving PSNR from $21.4$ to $22.8$. On HunyuanVideo, EpaCache achieves a $2.63\times$ speedup over uncached inference and improves SSIM from $0.891$ to $0.905$ over the prior state-of-the-art method at matched latency.