AI Paper Atlas

Research map / Architectures and efficiency / Quantization

I-ViT: Integer-only Quantization for Efficient Vision Transformer Inference

Zhikai Li, Qingyi Gu

ICCV 2023 · Architectures and efficiency › Quantization · approach: Post-training quantization of ViTs · 174 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 integer-only INT8 quantization with I-ViT instead of FP32DeiT-S · ImageNettop-1 accuracy: 79.85 -> 80.12Improved
use I-ViT INT8 instead of FP32Swin-S · ImageNettop-1 accuracy: 83.20 -> 83.01Hurt
use I-ViT instead of I-BERT polynomial approximation for non-linear opsSwin-S · ImageNettop-1 accuracy: 81.86 -> 83.01Improved
replace ShiftGELU with polynomial GELU (ablation)I-ViT DeiT-B · ImageNettop-1 accuracy: 81.74 -> 80.88Hurt
replace Shiftmax with polynomial softmax (ablation)I-ViT DeiT-B · ImageNettop-1 accuracy: 81.74 -> 81.62Hurt
replace Shiftmax with LIS (Log-Int-Softmax) (ablation)I-ViT DeiT-B · ImageNettop-1 accuracy: 81.74 -> 81.66Hurt
replace I-LayerNorm with L1 LayerNorm (ablation)I-ViT DeiT-B · ImageNettop-1 accuracy: 81.74 -> 79.25Hurt
replace I-LayerNorm with L1 LayerNorm (ablation)I-ViT DeiT-B · RTX 2080Ti latencylatency (ms): 7.93 -> 7.91No clear effect
use shifting-oriented Shiftmax/ShiftGELU instead of polynomial versions (ablation)I-ViT DeiT-B · RTX 2080Ti latencylatency (ms): 8.10 -> 7.93 (ShiftGELU vs Poly)Improved
integer-only inference instead of keeping non-linear ops in floating pointI-ViT vs FasterTransformer · RTX 2080Ti latency (DeiT-B)speedup: x3.72 -> x4.11Improved

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