Scaling Vision Transformers

Citations Over TimeTop 1% of 2022 papers

Abstract

Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well for few-shot transfer, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class.

Related Papers

• → Corrections to scaling and finite size effects(1984)12 cited
• → On the parameters of the Dixon–Nagel scaling procedure(1998)10 cited
• → Observed Scaling Behavior(2013)1 cited
• → Inverse scaling can become U-shaped(2022)1 cited
• → Beyond Positive Scaling: How Negation Impacts Scaling Trends of Language Models(2023)