UPop: Unified and Progressive Pruning for Compressing Vision-Language Transformers

TL;DR

UPop introduces a unified, progressive pruning framework for compressing vision-language Transformers, automatically optimizing subnet structures and ratios to achieve high compression without significant performance loss.

Contribution

It proposes a novel universal pruning method that combines continuous search and progressive retraining for multimodal Transformer compression.

Findings

Effective across various tasks and datasets

Achieves high compression ratios with maintained accuracy

Versatile for different model architectures

Abstract

Real-world data contains a vast amount of multimodal information, among which vision and language are the two most representative modalities. Moreover, increasingly heavier models, \textit{e}.\textit{g}., Transformers, have attracted the attention of researchers to model compression. However, how to compress multimodal models, especially vison-language Transformers, is still under-explored. This paper proposes the \textbf{U}nified and \textbf{P}r\textbf{o}gressive \textbf{P}runing (\textbf{\emph{UPop}}) as a universal vison-language Transformer compression framework, which incorporates 1) unifiedly searching multimodal subnets in a continuous optimization space from the original model, which enables automatic assignment of pruning ratios among compressible modalities and structures; 2) progressively searching and retraining the subnet, which maintains convergence between the search and retrain to attain higher compression ratios. Experiments on various tasks, datasets, and model architectures demonstrate the effectiveness and versatility of the proposed UPop framework. The code is available at https://github.com/sdc17/UPop.