Parallel Diffusion Solver via Residual Dirichlet Policy Optimization

TL;DR

This paper introduces EPD-Solver, a parallel ODE solver for diffusion models that reduces sampling errors and latency by leveraging multiple gradient evaluations and a novel optimization framework, improving image generation quality.

Contribution

The paper presents a novel parallel ODE solver with a two-stage optimization and RL fine-tuning, enhancing diffusion model sampling efficiency and quality.

Findings

EPD-Solver reduces sampling errors in diffusion models.

The RL fine-tuning improves text-to-image generation performance.

The method maintains low-latency sampling while enhancing quality.

Abstract

Diffusion models (DMs) have achieved state-of-the-art generative performance but suffer from high sampling latency due to their sequential denoising nature. Existing solver-based acceleration methods often face significant image quality degradation under a low-latency budget, primarily due to accumulated truncation errors arising from the inability to capture high-curvature trajectory segments. In this paper, we propose the Ensemble Parallel Direction solver (dubbed as EPD-Solver), a novel ODE solver that mitigates these errors by incorporating multiple parallel gradient evaluations in each step. Motivated by the geometric insight that sampling trajectories are largely confined to a low-dimensional manifold, EPD-Solver leverages the Mean Value Theorem for vector-valued functions to approximate the integral solution more accurately. Importantly, since the additional gradient computations are independent, they can be fully parallelized, preserving low-latency sampling nature. We introduce a two-stage optimization framework. Initially, EPD-Solver optimizes a small set of learnable parameters via a distillation-based approach. We further propose a parameter-efficient Reinforcement Learning (RL) fine-tuning scheme that reformulates the solver as a stochastic Dirichlet policy. Unlike traditional methods that fine-tune the massive backbone, our RL approach operates strictly within the low-dimensional solver space, effectively mitigating reward hacking while enhancing performance in complex text-to-image (T2I) generation tasks. In addition, our method is flexible and can serve as a plugin (EPD-Plugin) to improve existing ODE samplers.