Token Merging: Your ViT But Faster

TL;DR

Token Merging (ToMe) is a lightweight, training-free method that doubles the throughput of Vision Transformer models across images, videos, and audio with minimal accuracy loss.

Contribution

ToMe introduces a simple, fast token merging algorithm that enhances existing ViT models' speed without retraining, applicable during inference and training.

Findings

2x throughput on image models with 0.2-0.3% accuracy drop

2.2x throughput on video models with minimal accuracy loss

Effective merging of object parts into single tokens across frames

Abstract

We introduce Token Merging (ToMe), a simple method to increase the throughput of existing ViT models without needing to train. ToMe gradually combines similar tokens in a transformer using a general and light-weight matching algorithm that is as fast as pruning while being more accurate. Off-the-shelf, ToMe can 2x the throughput of state-of-the-art ViT-L @ 512 and ViT-H @ 518 models on images and 2.2x the throughput of ViT-L on video with only a 0.2-0.3% accuracy drop in each case. ToMe can also easily be applied during training, improving in practice training speed up to 2x for MAE fine-tuning on video. Training with ToMe further minimizes accuracy drop, leading to 2x the throughput of ViT-B on audio for only a 0.4% mAP drop. Qualitatively, we find that ToMe merges object parts into one token, even over multiple frames of video. Overall, ToMe's accuracy and speed are competitive with state-of-the-art on images, video, and audio.