AI Paper Atlas

Research map / Architectures and efficiency / Quantization

PackQViT: Faster Sub-8-bit Vision Transformers via Full and Packed Quantization on the Mobile

Peiyan Dong, LEI LU, Chao Wu, Cheng Lyu, Geng Yuan, Hao Tang, Yanzhi Wang

NeurIPS 2023 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization · 5 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 log2 attention quantization (ablation)Q-ViT baseline · ImageNet DeiT-T 8-bittop-1: 73.6 -> 74.1Improved
add log2 attention quantization (ablation)Q-ViT baseline · ImageNet DeiT-T 4-bittop-1: 72.1 -> 72.4Improved
add outlier-aware training (ablation)Q-ViT baseline · ImageNet DeiT-T 8-bittop-1: 73.6 -> 74.2Improved
add outlier-aware training (ablation)Q-ViT baseline · ImageNet DeiT-T 4-bittop-1: 72.1 -> 72.5Improved
add Int-2n-Softmax (integer softmax) (ablation)Q-ViT baseline · ImageNet DeiT-T 8-bittop-1: 73.6 -> 73.5Hurt
add Int-2n-Softmax (integer softmax) (ablation)Q-ViT baseline · ImageNet DeiT-T 4-bittop-1: 72.1 -> 72.0Hurt
add Int-LayerNorm (ablation)Q-ViT baseline · ImageNet DeiT-T 8-bittop-1: 73.6 -> 73.6No clear effect
combine all components with integer-only flow (ablation)Q-ViT baseline · ImageNet DeiT-T 8-bittop-1: 73.6 -> 74.5Improved
combine all components with integer-only flow (ablation)Q-ViT baseline · ImageNet DeiT-T 4-bittop-1: 72.1 -> 72.7Improved
full quantization of Softmax/LayerNorm/GELU at 4-bit (ablation)INT4 DeiT-T on mobile CPU · Speedup vs FP32speedup: 3.6x -> 4.9xImproved
full quantization at 4-bit (ablation)INT4 Swin-T on mobile CPU · Speedup vs FP32speedup: 2.8x -> 3.8xImproved
switch attention precision from 8 to 4 bitsPackQViT · ImageNet (avg)top-1: less than 0.1%No clear effect

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