Saved in:
Bibliographic Details
Main Authors: Zhang, Hanxiao, Zhou, Yifan, Wang, Guo-Hua, Wu, Jianxin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.18708
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913286596853760
author Zhang, Hanxiao
Zhou, Yifan
Wang, Guo-Hua
Wu, Jianxin
author_facet Zhang, Hanxiao
Zhou, Yifan
Wang, Guo-Hua
Wu, Jianxin
contents Few-shot model compression aims to compress a large model into a more compact one with only a tiny training set (even without labels). Block-level pruning has recently emerged as a leading technique in achieving high accuracy and low latency in few-shot CNN compression. But, few-shot compression for Vision Transformers (ViT) remains largely unexplored, which presents a new challenge. In particular, the issue of sparse compression exists in traditional CNN few-shot methods, which can only produce very few compressed models of different model sizes. This paper proposes a novel framework for few-shot ViT compression named DC-ViT. Instead of dropping the entire block, DC-ViT selectively eliminates the attention module while retaining and reusing portions of the MLP module. DC-ViT enables dense compression, which outputs numerous compressed models that densely populate the range of model complexity. DC-ViT outperforms state-of-the-art few-shot compression methods by a significant margin of 10 percentage points, along with lower latency in the compression of ViT and its variants.
format Preprint
id arxiv_https___arxiv_org_abs_2403_18708
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Dense Vision Transformer Compression with Few Samples
Zhang, Hanxiao
Zhou, Yifan
Wang, Guo-Hua
Wu, Jianxin
Computer Vision and Pattern Recognition
Few-shot model compression aims to compress a large model into a more compact one with only a tiny training set (even without labels). Block-level pruning has recently emerged as a leading technique in achieving high accuracy and low latency in few-shot CNN compression. But, few-shot compression for Vision Transformers (ViT) remains largely unexplored, which presents a new challenge. In particular, the issue of sparse compression exists in traditional CNN few-shot methods, which can only produce very few compressed models of different model sizes. This paper proposes a novel framework for few-shot ViT compression named DC-ViT. Instead of dropping the entire block, DC-ViT selectively eliminates the attention module while retaining and reusing portions of the MLP module. DC-ViT enables dense compression, which outputs numerous compressed models that densely populate the range of model complexity. DC-ViT outperforms state-of-the-art few-shot compression methods by a significant margin of 10 percentage points, along with lower latency in the compression of ViT and its variants.
title Dense Vision Transformer Compression with Few Samples
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.18708