AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Structural Pruning for Diffusion Models

Gongfan Fang, Xinyin Ma, Xinchao Wang

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 23 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
Use Diff-Pruning with T=0.05 instead of the strongest listed pruning baselineCIFAR-10 diffusion model · CIFAR-10FID: 5.48 -> 5.29Improved
Use Diff-Pruning with T=0.00 instead of the strongest listed pruning baselineCelebA-HQ diffusion model · CelebA-HQFID: 6.64 -> 6.24Improved
Prune the pretrained LSUN-Church model with Diff-PruningLSUN-Church DDPM · LSUN-ChurchFID: 10.6 -> 13.9Hurt
Prune the pretrained LSUN-Bedroom model with Diff-PruningLSUN-Bedroom DDPM · LSUN-BedroomFID: 6.9 -> 18.6Hurt
Use Diff-Pruning instead of Taylor pruningImageNet-1K conditional LDM · ImageNet-1K 256×256FID: 11.18 -> 9.16Improved
Add pruning to the pretrained CIFAR-10 model at 16% pruning (ablation)CIFAR-10 diffusion model · CIFAR-10FID: 4.19 -> 4.62Hurt
Increase pruning ratio from 44% to 70% (ablation)CIFAR-10 Diff-Pruning · CIFAR-10FID: 5.29 -> 9.33Hurt
Prune uninformative timesteps using threshold T=0.05 rather than using all timesteps (ablation)CIFAR-10 Diff-Pruning · CIFAR-10FID: 5.49 -> 5.29Improved
Increase the number of Taylor timesteps beyond about 250 (ablation)CIFAR-10 pruning importance estimation · CIFAR-10SSIMHurt
Use more Taylor timesteps for importance estimation (ablation)CelebA-HQ pruning importance estimation · CelebA-HQSSIMImproved
Train the compact model from scratch instead of pruning the pretrained modelCompact diffusion model · CIFAR-10 and CelebA-HQFIDHurt

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