AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Q-DM: An Efficient Low-bit Quantized Diffusion Model

Yanjing Li, Sheng Xu, Xianbin Cao, Xiao Sun, Baochang Zhang

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Fast diffusion ODE solvers · 0 citations (OpenAlex)

Read the paper

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 timestep-aware quantization (TaQ) (ablation)Quantized DM baseline (LSQ) 4-bit · CIFAR-10 DDIM 50 stepsFID: 10.22 -> 9.25Improved
Add noise-estimating mimicking (NeM) (ablation)Quantized DM baseline (LSQ) 4-bit · CIFAR-10 DDIM 50 stepsFID: 10.22 -> 8.98Improved
Add TaQ and NeM together (Q-DM) (ablation)Quantized DM baseline (LSQ) 4-bit · CIFAR-10 DDIM 50 stepsFID: 10.22 -> 6.89Improved
Add TaQ (ablation)Quantized DM baseline 3-bit · CIFAR-10 DDIM 50 stepsFID: 13.24 -> 11.19Improved
Add NeM (ablation)Quantized DM baseline 3-bit · CIFAR-10 DDIM 50 stepsFID: 13.24 -> 11.02Improved
Add TaQ and NeM together (ablation)Quantized DM baseline 3-bit · CIFAR-10 DDIM 50 stepsFID: 13.24 -> 9.07Improved
Add TaQ (ablation)Quantized DM baseline 2-bit · CIFAR-10 DDIM 50 stepsFID: 18.74 -> 16.83Improved
Add NeM (ablation)Quantized DM baseline 2-bit · CIFAR-10 DDIM 50 stepsFID: 18.74 -> 16.97Improved
Add TaQ and NeM together (ablation)Quantized DM baseline 2-bit · CIFAR-10 DDIM 50 stepsFID: 18.74 -> 15.26Improved
Q-DM versus LSQ baseline at 4-bit with 1000-step DDPMQuantized DM · CIFAR-10 DDPM 1000 stepsFID: 9.11 -> 5.17Improved
Q-DM versus LSQ baseline at 4-bit with 50-step DDIMQuantized DM · ImageNet 64x64FID: 24.78 -> 20.02Improved
4-bit Q-DM versus 8-bit PTQ4DMQuantized DM · CIFAR-10 DDIM 50 stepsFID: 18.02 -> 6.89Improved

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