Lossless Token Merging Even Without Fine-Tuning in Vision Transformers

TL;DR

This paper introduces Adaptive Token Merging (ATM), a lossless, training-free method for reducing tokens in Vision Transformers, significantly decreasing computational costs without sacrificing accuracy.

Contribution

ATM is a novel, adaptive token merging technique that prevents information loss and eliminates the need for fine-tuning in Vision Transformers.

Findings

ATM outperforms existing training-free methods.

Achieves over 30% FLOPs reduction without accuracy loss.

Surpasses many training-based approaches without additional training.

Abstract

Although Vision Transformers (ViTs) have become the standard architecture in computer vision, their massive sizes lead to significant computational overhead. Token compression techniques have attracted considerable attention to address this issue, but they often suffer from severe information loss, requiring extensive additional training to achieve practical performance. In this paper, we propose Adaptive Token Merging (ATM), a novel method that ensures lossless token merging, eliminating the need for fine-tuning while maintaining competitive performance. ATM adaptively reduces tokens across layers and batches by carefully adjusting layer-specific similarity thresholds, thereby preventing the undesirable merging of less similar tokens with respect to each layer. Furthermore, ATM introduces a novel token matching technique that considers not only similarity but also merging sizes, particularly for the final layers, to minimize the information loss incurred from each merging operation. We empirically validate our method across a wide range of pretrained models, demonstrating that ATM not only outperforms all existing training-free methods but also surpasses most training-intensive approaches, even without additional training. Remarkably, training-free ATM achieves over a 30% reduction in FLOPs for the DeiT-T and DeiT-S models without any drop in their original accuracy.