FSampler: Training Free Acceleration of Diffusion Sampling via Epsilon Extrapolation

TL;DR

FSampler is a training-free acceleration method for diffusion sampling that reduces function evaluations by extrapolating denoising signals, maintaining high fidelity and broad compatibility with existing samplers.

Contribution

It introduces a novel epsilon extrapolation technique at the sampler level, enabling significant acceleration without retraining or altering core sampler formulas.

Findings

Reduces sampling time by up to 22% at high fidelity.

Cuts model calls by 15-25% without quality loss.

Achieves up to 50% reduction in model calls with lower fidelity.

Abstract

FSampler is a training free, sampler agnostic execution layer that accelerates diffusion sampling by reducing the number of function evaluations (NFE). FSampler maintains a short history of denoising signals (epsilon) from recent real model calls and extrapolates the next epsilon using finite difference predictors at second order, third order, or fourth order, falling back to lower order when history is insufficient. On selected steps the predicted epsilon substitutes the model call while keeping each sampler's update rule unchanged. Predicted epsilons are validated for finiteness and magnitude; a learning stabilizer rescales predictions on skipped steps to correct drift, and an optional gradient estimation stabilizer compensates local curvature. Protected windows, periodic anchors, and a cap on consecutive skips bound deviation over the trajectory. Operating at the sampler level, FSampler integrates with Euler/DDIM, DPM++ 2M/2S, LMS/AB2, and RES family exponential multistep methods and drops into standard workflows. FLUX.1 dev, Qwen Image, and Wan 2.2, FSampler reduces time by 8 to 22% and model calls by 15 to 25% at high fidelity (Structural Similarity Index (SSIM) 0.95 to 0.99), without altering sampler formulas. With an aggressive adaptive gate, reductions can reach 45 to 50% fewer model calls at lower fidelity (SSIM 0.73 to 0.74).