Denoising as Path Planning: Training-Free Acceleration of Diffusion Models with DPCache

2026-02-26

AI Summary

This work addresses the high computational cost of multi-step sampling in diffusion models, where existing caching methods often accumulate errors and introduce visual artifacts due to their neglect of the global structure of denoising trajectories. To overcome this limitation, the paper formulates sampling acceleration as a global path planning problem and introduces DPCache—a training-free framework that constructs a path-aware cost tensor and leverages dynamic programming to select an optimal sequence of key timesteps. By integrating feature caching with a prediction mechanism, DPCache significantly reduces computation while preserving trajectory fidelity. The method achieves up to 4.87× speedup on DiT, FLUX, and HunyuanVideo, and on FLUX it attains a 3.54× acceleration over the full-step baseline with a +0.028 ImageReward gain, substantially outperforming current state-of-the-art approaches.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Diffusion models have demonstrated remarkable success in image and video generation, yet their practical deployment remains hindered by the substantial computational overhead of multi-step iterative sampling. Among acceleration strategies, caching-based methods offer a training-free and effective solution by reusing or predicting features across timesteps. However, existing approaches rely on fixed or locally adaptive schedules without considering the global structure of the denoising trajectory, often leading to error accumulation and visual artifacts. To overcome this limitation, we propose DPCache, a novel training-free acceleration framework that formulates diffusion sampling acceleration as a global path planning problem. DPCache constructs a Path-Aware Cost Tensor from a small calibration set to quantify the path-dependent error of skipping timesteps conditioned on the preceding key timestep. Leveraging this tensor, DPCache employs dynamic programming to select an optimal sequence of key timesteps that minimizes the total path cost while preserving trajectory fidelity. During inference, the model performs full computations only at these key timesteps, while intermediate outputs are efficiently predicted using cached features. Extensive experiments on DiT, FLUX, and HunyuanVideo demonstrate that DPCache achieves strong acceleration with minimal quality loss, outperforming prior acceleration methods by $+$0.031 ImageReward at 4.87$\times$ speedup and even surpassing the full-step baseline by $+$0.028 ImageReward at 3.54$\times$ speedup on FLUX, validating the effectiveness of our path-aware global scheduling framework. Code will be released at https://github.com/argsss/DPCache.

Problem

• diffusion models
• sampling acceleration
• denoising trajectory
• training-free
• computational overhead

Innovation

• diffusion acceleration
• path planning
• training-free
• dynamic programming
• feature caching