What type of study is this?

This is a Experimental Study study.

September 17, 2025

I&S-ViT: An Inclusive & Stable Method for Post-Training ViTs Quantization

Key Points

I&S-ViT elevates the performance of W3A3 ViT-B by 50.68% in low-bit scenarios, showcasing its effectiveness.
The method introduces a shift-uniform-log2 quantizer that improves quantization and domain representation.
I&S-ViT employs a three-stage optimization strategy that stabilizes learning during the quantization process.
Evaluations across various vision tasks demonstrate I&S-ViT's superiority over existing post-training quantization methods.

Abstract

Albeit the scalable performance of vision transformers (ViTs), the dense computational costs undermine their position in industrial applications. Post-training quantization (PTQ), tuning ViTs with a tiny dataset and running in a low-bit format, well addresses the cost issue but unluckily bears more performance drops in lower-bit cases. In this paper, we introduce I (2) Rugged and magnified loss landscape in coarse-grained quantization granularity for post-LayerNorm activations. Then, I (2) A three-stage smooth optimization strategy (SOS) that amalgamates the strengths of channel-wise and layer-wise quantization to enable stable learning. Comprehensive evaluations across diverse vision tasks validate I&S-ViT's superiority over existing PTQ of ViTs methods, particularly in low-bit scenarios. For instance, I&S-ViT elevates the performance of W3A3 ViT-B by an impressive 50.68%. Our code is available at https://github.com/zysxmu/IaS-ViT.

Demander à l'IA

Bookmark

View Full Paper