DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification

TL;DR

DynamicViT introduces a dynamic token pruning method for vision transformers, significantly reducing computational costs while maintaining high accuracy by selectively focusing on the most informative tokens during inference.

Contribution

It proposes a novel end-to-end framework for hierarchical token sparsification in vision transformers, enabling efficient computation without sacrificing accuracy.

Findings

Prunes 66% of input tokens hierarchically

Reduces FLOPs by 31-37%

Increases throughput by over 40% with minimal accuracy loss

Abstract

Attention is sparse in vision transformers. We observe the final prediction in vision transformers is only based on a subset of most informative tokens, which is sufficient for accurate image recognition. Based on this observation, we propose a dynamic token sparsification framework to prune redundant tokens progressively and dynamically based on the input. Specifically, we devise a lightweight prediction module to estimate the importance score of each token given the current features. The module is added to different layers to prune redundant tokens hierarchically. To optimize the prediction module in an end-to-end manner, we propose an attention masking strategy to differentiably prune a token by blocking its interactions with other tokens. Benefiting from the nature of self-attention, the unstructured sparse tokens are still hardware friendly, which makes our framework easy to achieve actual speed-up. By hierarchically pruning 66% of the input tokens, our method greatly reduces 31%~37% FLOPs and improves the throughput by over 40% while the drop of accuracy is within 0.5% for various vision transformers. Equipped with the dynamic token sparsification framework, DynamicViT models can achieve very competitive complexity/accuracy trade-offs compared to state-of-the-art CNNs and vision transformers on ImageNet. Code is available at https://github.com/raoyongming/DynamicViT