DRIFT: Harnessing Inherent Fault Tolerance for Efficient and Reliable Diffusion Model Inference

TL;DR

DRIFT is a co-optimization framework that leverages the inherent fault tolerance of diffusion models to improve inference efficiency and reliability, reducing energy consumption and latency.

Contribution

It introduces a resilience-aware DVFS strategy and an adaptive rollback fault tolerance mechanism tailored for diffusion models.

Findings

Achieves 36% energy savings through voltage underscaling.

Attains 1.7x speedup via overclocking while maintaining quality.

Provides a comprehensive resilience analysis of diffusion models.

Abstract

Diffusion model deployment has been suffering from high energy consumption and inference latency despite its superior performance in visual generation tasks. Dynamic voltage and frequency scaling (DVFS) offers a promising solution to exploit the potential of the underlying accelerators. However, existing approaches often lead to either limited efficiency gains or degraded output quality because they overlook the inherent fault tolerance of the diffusion model. Therefore, in this paper, we propose DRIFT, a novel algorithmarchitecture co-optimization framework that harnesses the fault tolerance for efficient and reliable diffusion model inference. We first perform a comprehensive resilience analysis on representative diffusion models. Building on these observations, we introduce a fine-grained, resilience-aware DVFS strategy that selectively protects error-sensitive network blocks and timesteps, and a rollback algorithm-based fault tolerance (ABFT) mechanism that adaptively corrects only critical errors by reverting to previous timesteps. We further optimize offloading intervals and reorganize data layouts to reduce memory overhead. Experiments across diverse models and datasets show that DRIFT can achieve on average 36% energy savings through voltage underscaling or 1.7x speedup via overclocking while maintaining generation quality.