AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

MixDQ: Memory-Efficient Few-Step Text-to-Image Diffusion Models with Metric-Decoupled Mixed Precision Quantization

Tianchen Zhao, Xuefei Ning, Tongcheng Fang, Enshu Liu, Guyue Huang, Zinan Lin, Shengen Yan, Guohao Dai, Yu Wang

ECCV 2024 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 28 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add BOS-aware quantization of text embeddings (ablation)MixDQ W8A8 on SDXL-turbo · COCO2014FID: 103.95 -> 31.65 (FID); CLIP 0.1478 -> 0.2652Improved
add mixed precision using SQNR-only sensitivity (ablation)BOS-aware W8A8 SDXL-turbo · COCO2014FID: 31.65 -> 37.35 (FID); CLIP 0.2652 -> 0.2624Hurt
add metric-decoupled sensitivity (SSIM for content, SQNR for quality) to mixed precision (ablation)MixDQ W8A8 on SDXL-turbo · COCO2014FID: 37.35 -> 17.03 (FID); CLIP 0.2624 -> 0.2703Improved
use MixDQ instead of Q-Diffusion at W4A8SDXL-turbo 1 step · COCO2014FID: 118.93 -> 17.68Improved
use MixDQ instead of Q-Diffusion at W8A8SDXL-turbo 1 step · COCO2014FID: 76.18 -> 17.03Improved
use FP8 non-uniform quantization instead of MixDQSDXL-turbo W8A8 · COCO subset 1024 promptsFID: 83.39 -> 101.73Hurt
use EDA-DM layer-wise reconstruction over Q-DiffusionSDXL-turbo W8A8 · COCO subset 1024 promptsFID: 149.15 -> 137.98No clear effect
run the 1-step model instead of 2-step under quantizationSDXL-turbo · COCO promptsquantization degradationHurt
use integer programming allocation vs random or naive sorting (ablation)MixDQ mixed precision · SDXL-turbo Pareto frontierCLIP Score vs bit-widthImproved
W4A16 vs FP16 latencyMixDQ · RTX 4080latencyNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy