We attempt to reduce the computational costs in vision transformers (ViTs), which increase quadratically in the token number. We present a novel training paradigm that trains only one ViT model at a time, but is capable of providing improved image recognition performance with various computational costs. Here, the trained ViT model, termed super vision transformer (SuperViT), is empowered with the versatile ability to solve incoming patches of multiple sizes as well as preserve informative tokens with multiple keeping rates (the ratio of keeping tokens) to achieve good hardware efficiency for inference, given that the available hardware resources often change from time to time. Experimental results on ImageNet demonstrate that our SuperViT can considerably reduce the computational costs of ViT models with even performance increase. For example, we reduce 2x FLOPs of DeiT-S while increasing the Top-1 accuracy by 0.2% and 0.7% for 1.5x reduction. Also, our SuperViT significantly outperforms existing studies on efficient vision transformers. For example, when consuming the same amount of FLOPs, our SuperViT surpasses the recent state-of-the-art (SoTA) EViT by 1.1% when using DeiT-S as their backbones. The project of this work is made publicly available at https://github.com/lmbxmu/SuperViT.

翻译：我们试图降低视觉变压器的计算成本,这些变压器的计算成本在象征性数字中增加了四倍。 我们展示了一个新的培训模式,每次只培训一个 Vit 模型,但能够以各种计算成本提供更好的图像识别性能。 这里, 受过训练的 Vit 模型, 称为超级视觉变压器( SuperViT), 具有解决多尺寸进取补补补丁的多功能能力, 并保存信息信号, 并具有多种保存率( 保持标语的比例), 以实现高硬件的推断效率, 因为可用的硬件资源经常随时间变化。 图像网络的实验结果表明, 我们的 SuperViViT 能够大幅降低 ViT 模型的计算成本, 甚至提高性能。 例如, 我们将 DeiT- S 的2x FLOPs 降低2 % 和 0. 0. 0. 0. 0. 和 0. 7 % 。 此外, 我们的超ViT 大大超过目前对高效变压器的研究。 例如, 我们的SUViT 超越了最近的一个 State-Fib/SVIA项目。