InvarDiff: Cross-Scale Invariance Caching for Accelerated Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Wu, Zihao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908694415933440
author Wu, Zihao
author_facet Wu, Zihao
contents Diffusion models deliver high-fidelity synthesis but remain slow due to iterative sampling. We empirically observe there exists feature invariance in deterministic sampling, and present InvarDiff, a training-free acceleration method that exploits the relative temporal invariance across timestep-scale and layer-scale. From a few deterministic runs, we compute a per-timestep, per-layer, per-module binary cache plan matrix and use a re-sampling correction to avoid drift when consecutive caches occur. Using quantile-based change metrics, this matrix specifies which module at which step is reused rather than recomputed. The same invariance criterion is applied at the step scale to enable cross-timestep caching, deciding whether an entire step can reuse cached results. During inference, InvarDiff performs step-first and layer-wise caching guided by this matrix. When applied to DiT and FLUX, our approach reduces redundant compute while preserving fidelity. Experiments show that InvarDiff achieves $2$-$3\times$ end-to-end speed-ups with minimal impact on standard quality metrics. Qualitatively, we observe almost no degradation in visual quality compared with full computations.
format Preprint
id arxiv_https___arxiv_org_abs_2512_05134
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle InvarDiff: Cross-Scale Invariance Caching for Accelerated Diffusion Models
Wu, Zihao
Computer Vision and Pattern Recognition
Distributed, Parallel, and Cluster Computing
Machine Learning
Diffusion models deliver high-fidelity synthesis but remain slow due to iterative sampling. We empirically observe there exists feature invariance in deterministic sampling, and present InvarDiff, a training-free acceleration method that exploits the relative temporal invariance across timestep-scale and layer-scale. From a few deterministic runs, we compute a per-timestep, per-layer, per-module binary cache plan matrix and use a re-sampling correction to avoid drift when consecutive caches occur. Using quantile-based change metrics, this matrix specifies which module at which step is reused rather than recomputed. The same invariance criterion is applied at the step scale to enable cross-timestep caching, deciding whether an entire step can reuse cached results. During inference, InvarDiff performs step-first and layer-wise caching guided by this matrix. When applied to DiT and FLUX, our approach reduces redundant compute while preserving fidelity. Experiments show that InvarDiff achieves $2$-$3\times$ end-to-end speed-ups with minimal impact on standard quality metrics. Qualitatively, we observe almost no degradation in visual quality compared with full computations.
title InvarDiff: Cross-Scale Invariance Caching for Accelerated Diffusion Models
topic Computer Vision and Pattern Recognition
Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2512.05134