AccuQuant: Simulating Multiple Denoising Steps for Quantizing Diffusion Models
This addresses the challenge of efficiently deploying diffusion models on resource-constrained devices by reducing memory complexity from O(n) to O(1), which is incremental but impactful for practical applications.
The paper tackles the problem of quantization errors accumulating over denoising steps in diffusion models by introducing AccuQuant, a post-training quantization method that minimizes discrepancies between full-precision and quantized models across multiple steps, achieving improved performance on standard benchmarks.
We present in this paper a novel post-training quantization (PTQ) method, dubbed AccuQuant, for diffusion models. We show analytically and empirically that quantization errors for diffusion models are accumulated over denoising steps in a sampling process. To alleviate the error accumulation problem, AccuQuant minimizes the discrepancies between outputs of a full-precision diffusion model and its quantized version within a couple of denoising steps. That is, it simulates multiple denoising steps of a diffusion sampling process explicitly for quantization, accounting the accumulated errors over multiple denoising steps, which is in contrast to previous approaches to imitating a training process of diffusion models, namely, minimizing the discrepancies independently for each step. We also present an efficient implementation technique for AccuQuant, together with a novel objective, which reduces a memory complexity significantly from $\mathcal{O}(n)$ to $\mathcal{O}(1)$, where $n$ is the number of denoising steps. We demonstrate the efficacy and efficiency of AccuQuant across various tasks and diffusion models on standard benchmarks.