PaperlayerView exact revision on arXiv

Source-linked AI summary

EpaCache: Error-Propagation-Aware Caching for Accelerating Diffusion-Based Visual Generation

Yuhan Liu, Zongwei Hong, Jinglun Li, Linze Li, Shen Zhang, Yao Tang

arXiv:2608.29264v1cs.AI

TL;DR

Sequential denoising makes diffusion-based visual generation expensive, while existing cache controllers largely overlook how reuse errors propagate along the trajectory. EpaCache is a training-free policy that calibrates timestep-specific downstream impact and reallocates reuse tolerance accordingly, consistently improving latency–fidelity trade-offs across image and video models. Its scope includes reporting discrete timestep indices for both diffusion and flow-matching trajectories.

  • Problem

    Existing cache controllers rely mainly on local temporal variation and do not explicitly account for timestep-dependent propagation of approximation errors through the remaining sampling trajectory.

    Source: Introduction · p. 1

  • Method

    EpaCache uses isolated cache-reuse perturbations to calibrate downstream impact, then redistributes a global mean tolerance across timesteps.

    Source: Introduction · p. 2, EpaCache: Error-Propagation-Aware Cache · p. 4

  • Results

    EpaCache consistently improves latency–quality trade-offs over existing caching baselines across FLUX.1-dev, Wan2.1, and HunyuanVideo under comparable realized computation budgets.

    Source: Conclusion · p. 7

  • Takeaways & Limitations

    Error-propagation-aware cache allocation provides an effective strategy for assigning conservative refreshes to sensitive timesteps and aggressive reuse elsewhere.

    Source: Conclusion · p. 7

    The paper reports discrete timestep indices along the sampling trajectory for ease of interpretation across diffusion and flow-matching parameterizations.

    Source: Preliminary · p. 3

Abstract

from arXiv · show

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.

Introduction

Diffusion generation quality has improved, but sequential denoising makes high-resolution image and video synthesis costly. EpaCache addresses cache-reuse errors by allocating reuse tolerance according to timestep-dependent downstream impact.

  • Sequential multi-step sampling repeatedly evaluates large networks, making latency, memory use, and computation costly for high-resolution generation.Source: Introduction · p. 1
  • Caching reduces inference cost by reusing outputs or intermediate computations across adjacent timesteps without modifying model weights.Source: Introduction · p. 1
  • Cache errors can propagate through later denoising steps, with earlier or particularly sensitive timesteps producing larger final deviations.Source: Introduction · p. 1, Figure 2 · p. 2
  • EpaCache calibrates each timestep’s downstream impact using isolated reuse perturbations, then assigns smaller tolerances to high-impact steps and larger tolerances to low-impact steps.Source: Introduction · p. 2
  • Experiments show improved latency–fidelity trade-offs across acceleration budgets and generative models compared with prior caching baselines.Source: Introduction · p. 2

Related Work

Diffusion acceleration methods reduce sampling steps, evaluation cost, or redundant computation. Cache-based approaches increasingly use lightweight temporal proxies, but differ in representation and granularity.

Methods

EpaCache combines timestep-specific error-propagation-aware thresholds with a lightweight local-variation proxy for online residual reuse. It calibrates downstream impact, allocates a fixed mean tolerance across timesteps, and refreshes computations when accumulated variation exceeds the applicable threshold.

Error-Propagation-Aware Cache Budget Allocation

  • EpaCache calibrates each cacheable timestep’s downstream impact by injecting residual reuse and measuring final-state deviation against a full-computation reference.The calibration replaces one timestep’s computation while retaining later full computations, then averages relative L1 deviation over prompts.Source: EpaCache: Error-Propagation-Aware Cache · pp. 3–4
  • The first sampling step is excluded because no preceding residual exists, and the last is fully computed to prevent severe quality degradation.Source: EpaCache: Error-Propagation-Aware Cache · p. 3
  • High-impact timesteps receive smaller reuse thresholds, while low-impact timesteps receive larger thresholds under the same global mean threshold.The inverse-normalized allocation weights preserve the mean reuse threshold and discriminate more finely among low-deviation timesteps.Source: EpaCache: Error-Propagation-Aware Cache · p. 4

Online Cache Decision

Local Variation Proxy and Residual Reuse

  • The timestep-modulated proxy represents the noisy transformer input and estimates local temporal variation with relative ℓ1 distance.This inexpensive signal measures how much the model state changes between consecutive diffusion iterations.Source: EpaCache: Error-Propagation-Aware Cache · p. 4

Experiments

Experiments across image and video models show that EpaCache improves the efficiency–fidelity trade-off over caching baselines, while ablations confirm the value of propagation-aware allocation and local variation signals.

Text-to-image generation

  • EpaCache achieves the lowest latency while maintaining the best overall fidelity on FLUX.1-dev at approximately 70% skip rate.At approximately 50% skip rate, it attains the best PSNR, tied-best SSIM, and comparable LPIPS with latency comparable to SeaCache.Source: Quantitative Comparison · p. 5

Text-to-video generation

  • 30.47 PSNR at 212.88 s demonstrates EpaCache’s strong Wan2.1-1.3B efficiency–fidelity trade-off.It outperforms SeaCache in fidelity at similar latency and NaviCache in PSNR despite slightly lower latency.Source: Quantitative Comparison · p. 5
  • 286.11 s versus 290.69 s shows EpaCache is faster than SeaCache on HunyuanVideo while improving PSNR, SSIM, and LPIPS.The overall comparison covers both Wan2.1-1.3B and HunyuanVideo.Source: Quantitative Comparison · p. 5

Qualitative comparison

  • Qualitative comparisons show EpaCache better preserves image semantics and perceptual quality than competing methods at matched skip rates.On video, baseline methods exhibit facial-expression deviations or semantic changes in representative frames.Source: Qualitative Comparison · pp. 5–6

Ablation study

  • Minimum 50% Lt substantially degrades performance, showing that local variation is needed to identify stale cached features.Reverse Lt also performs worse, while Slash remains inferior to the error-propagation-aware threshold strategy.Source: Additional Analysis · p. 7

Calibration sensitivity

  • N = 5 substantially degrades reconstruction fidelity, while N = 1 can be competitive but produces a highly prompt-dependent impact profile.The default calibration setting uses 20 prompts.Source: Additional Analysis · p. 7, Table 5 · p. 7

Cache decision visualization

  • EpaCache reaches 3.82× speedup at a 74.7% skip rate by distributing refreshes across multiple influential regions.TeaCache uses 3.21× speedup and 70.0% skip rate, while SeaCache uses 3.68× and 74.0%.Source: Additional Analysis · p. 7, Figure 7 · p. 7

Conclusion

EpaCache is a training-free caching policy that calibrates timestep-wise error propagation and adapts reuse thresholds across diffusion trajectories. Experiments on FLUX.1-dev, Wan2.1, and HunyuanVideo show improved latency–quality trade-offs without retraining or architectural modifications.

Conclusion

  • EpaCache calibrates timestep-wise propagation impacts through isolated cache-reuse perturbations and converts them into an adaptive threshold schedule.Sensitive timesteps receive conservative refreshes, while less impactful timesteps allow more aggressive reuse.Source: Conclusion · p. 7
  • Experiments across FLUX.1-dev, Wan2.1, and HunyuanVideo demonstrate consistently improved latency–quality trade-offs over caching baselines.The comparisons use comparable realized computation budgets.Source: Conclusion · p. 7
  • Error-propagation-aware cache allocation can integrate into existing cache controllers without model retraining or architectural modifications.Source: Conclusion · p. 7

A Overhead of Error-Propagation Profiling

EpaCache profiles downstream cache-reuse effects from a reference trajectory and uses the resulting timestep impacts to construct its reuse schedule. Profiling excludes the first and last denoising steps from cacheable timesteps and can be parallelized across prompts and timesteps.

Profiling procedure

Cacheable timesteps

Cost model

Calibration overhead

  • 24.52 standard-inference equivalents are required for one profiling prompt, while N = 20 requires approximately 490.4 equivalents.Profiling runs are independent across timesteps and prompts and can be parallelized.Source: A Overhead of Error-Propagation… · p. 10

B Optimization: Reducing Profiling Overhead with Sparse-Step Interpolation

EpaCache’s offline error-propagation profiling adds practical overhead because it requires timestep-wise cache-injection runs. Sparse-step calibration reduces this burden by interpolating propagation profiles for uncalibrated steps while preserving stable latency–quality trade-offs.

C Dimension-Level VBench Evaluation on HunyuanVideo

On HunyuanVideo, EpaCache remains competitive across VBench semantic and quality dimensions. It ranks strongly among caching methods, especially on quality-related measures and several individual dimensions.

Aggregate VBench Results

Semantic Dimensions

Quality Dimensions

Rankings

D Effectiveness of the Dynamic Degree

Qualitative examples indicate that explicitly accounting for error propagation helps EpaCache reduce visible structural drift under cache reuse.

Metric

Dynamic Degree measures noticeable motion, but it does not reliably capture semantic preservation or visual fidelity. The evaluation therefore compares it with reference-fidelity and other quality metrics.

  • Dynamic Degree measures whether generated video contains sufficiently noticeable motion.Source: Metric · p. 13
  • Cache-induced errors may cause temporal flicker, semantic drift, or perceptual degradation that Dynamic Degree does not consistently reflect.Source: Metric · p. 13
  • EpaCache scores lower on Dynamic Degree but higher on PSNR, SSIM, LPIPS, Subject Consistency, and Motion Smoothness.Source: Results on the Current Evaluation Set · p. 13
  • Dynamic Agreement indicates that EpaCache better preserves the uncached reference’s motion decisions.Source: Results on the Current Evaluation Set · p. 13
  • Qualitative comparisons report better overall perceptual quality and closer appearance to the uncached reference for EpaCache than TeaCache.Source: Qualitative Visualization · p. 13

E Supplementary Visualization and Case Study

The supplementary visualization compares EpaCache with the uncached baseline and existing caching methods on FLUX.1-dev, while Table 10 compares Dynamic Degree with other quality metrics.

Qualitative comparison

Metric comparison

  • Table 10 compares Dynamic Degree with other quality metrics.Source: Table 10 · p. 14

Qualitative comparison

Related papers

Loading 2608.29264v1…