Error Propagation Mechanisms and Compensation Strategies for Quantized Diffusion
By: Songwei Liu , Hong Liu , Fangmin Chen and more
Potential Business Impact:
Makes AI images faster without losing quality.
Diffusion models have transformed image synthesis by establishing unprecedented quality and creativity benchmarks. Nevertheless, their large-scale deployment faces challenges due to computationally intensive iterative denoising processes. Although post-training quantization(PTQ) provides an effective pathway for accelerating sampling, the iterative nature of diffusion models causes stepwise quantization errors to accumulate progressively during generation, inevitably compromising output fidelity. To address this challenge, we develop a theoretical framework that mathematically formulates error propagation in Diffusion Models (DMs), deriving per-step quantization error propagation equations and establishing the first closed-form solution for cumulative error. Building on this theoretical foundation, we propose a timestep-aware cumulative error compensation scheme. Extensive experiments across multiple image datasets demonstrate that our compensation strategy effectively mitigates error propagation, significantly enhancing existing PTQ methods to achieve state-of-the-art(SOTA) performance on low-precision diffusion models.
Similar Papers
AccuQuant: Simulating Multiple Denoising Steps for Quantizing Diffusion Models
CV and Pattern Recognition
Makes AI image generators faster and smaller.
Quantization Error Propagation: Revisiting Layer-Wise Post-Training Quantization
Machine Learning (CS)
Makes AI models smaller and faster.
Diffusion Model Quantization: A Review
CV and Pattern Recognition
Makes AI art generators run on phones.