Scaling Laws in Patchification: An Image Is Worth 50,176 Tokens And More

TL;DR

This paper investigates how decreasing patch size in vision transformers improves performance, revealing a scaling law that extends to pixel-level tokenization and enables processing extremely long sequences with high accuracy.

Contribution

The study uncovers a scaling law in patchification, demonstrating that smaller patches consistently enhance model performance across tasks and architectures, and scales sequences up to 50,176 tokens.

Findings

Smaller patches lead to better predictive accuracy.

Models benefit from decreased patch sizes until pixel-level tokenization.

Achieved 84.6% accuracy on ImageNet-1k with 50,176 tokens.

Abstract

Since the introduction of Vision Transformer (ViT), patchification has long been regarded as a de facto image tokenization approach for plain visual architectures. By compressing the spatial size of images, this approach can effectively shorten the token sequence and reduce the computational cost of ViT-like plain architectures. In this work, we aim to thoroughly examine the information loss caused by this patchification-based compressive encoding paradigm and how it affects visual understanding. We conduct extensive patch size scaling experiments and excitedly observe an intriguing scaling law in patchification: the models can consistently benefit from decreased patch sizes and attain improved predictive performance, until it reaches the minimum patch size of 1x1, i.e., pixel tokenization. This conclusion is broadly applicable across different vision tasks, various input scales, and diverse architectures such as ViT and the recent Mamba models. Moreover, as a by-product, we discover that with smaller patches, task-specific decoder heads become less critical for dense prediction. In the experiments, we successfully scale up the visual sequence to an exceptional length of 50,176 tokens, achieving a competitive test accuracy of 84.6% with a base-sized model on the ImageNet-1k benchmark. We hope this study can provide insights and theoretical foundations for future works of building non-compressive vision models. Code is available at https://github.com/wangf3014/Patch_Scaling.