AI Paper Atlas

Research map / Architectures and efficiency / Quantization

PTQ4ViT: Post-Training Quantization for Vision Transformers with Twin Uniform Quantization

Zhihang Yuan, Chenhao Xue, Yiqi Chen, Qiang Wu, Guangyu Sun

ECCV 2022 · Architectures and efficiency › Quantization · approach: Post-training quantization of ViTs · 220 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 Hessian guided metric to base PTQ (ablation)ViT-S/224 W6A6 · ImageNet-1k valtop-1 accuracy: 70.24 -> 77.20Improved
add Hessian guided metric to base PTQ (ablation)ViT-S/224 W8A8 · ImageNet-1k valtop-1 accuracy: 80.47 -> 80.93Improved
add Hessian guided metric to base PTQ (ablation)ViT-B/384 W6A6 · ImageNet-1k valtop-1 accuracy: 46.89 -> 79.99Improved
add Hessian guided metric to twin quantization (ablation)ViT-S/224 W8A8 · ImageNet-1k valtop-1 accuracy: 79.25 -> 81.00Improved
add Hessian guided metric to twin quantization (ablation)ViT-S/224 W6A6 · ImageNet-1k valtop-1 accuracy: 74.07 -> 78.63Improved
add twin uniform quantization (softmax and GELU) on top of Hessian metric (ablation)ViT-S/224 W6A6 · ImageNet-1k valtop-1 accuracy: 77.20 -> 78.63Improved
add twin uniform quantization on top of Hessian metric (ablation)ViT-B/384 W6A6 · ImageNet-1k valtop-1 accuracy: 79.99 -> 83.19Improved
add twin uniform quantization without Hessian metric (ablation)ViT-S/224 W8A8 · ImageNet-1k valtop-1 accuracy: 80.47 -> 79.25Hurt
use PTQ4ViT instead of base PTQViT-B/384 W6A6 · ImageNet-1k valtop-1 accuracy: 46.88 -> 83.34Improved
use PTQ4ViT instead of Liu et al. (MP)DeiT-B W4A4 · ImageNet-1k valtop-1 accuracy: 75.94 -> 60.91Hurt
add bias correction to PTQ4ViTDeiT-B W4A4 · ImageNet-1k valtop-1 accuracy: 60.91 -> 64.39Improved
use PTQ4ViT instead of EasyQuant with fewer calibration images (32 vs 1024)DeiT-S W8A8 · ImageNet-1k valtop-1 accuracy: 76.59 -> 79.47Improved

Later papers that compared with this method

8 later accepted papers report PTQ4ViT as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.

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