Autoregressive Pretraining with Mamba in Vision
South China University of Technology, Tsinghua University · University of California, Santa Cruz · University of Chinese Academy of Sciences · Tsinghua University, Tsinghua University · Alibaba · University of California, San Diego · Johns Hopkins University · ByteDance Inc. · Bytedance/TikTok · NVIDIA · UC Santa Cruz
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
The vision community has started to build with the recently developed state space model, Mamba, as the new backbone for a range of tasks. This paper shows that Mamba's visual capability can be significantly enhanced through autoregressive pretraining, a direction not previously explored. Efficiency-wise, the autoregressive nature can well capitalize on the Mamba's unidirectional recurrent structure, enabling faster overall training speed compared to other training strategies like mask modeling. Performance-wise, autoregressive pretraining equips the Mamba architecture with markedly higher accuracy over its supervised-trained counterparts and, more importantly, successfully unlocks its scaling potential to large and even huge model sizes. For example, with autoregressive pretraining, a base-size Mamba attains 83.2\% ImageNet accuracy, outperforming its supervised counterpart by 2.0\%; our huge-size Mamba, the largest Vision Mamba to date, attains 85.0\% ImageNet accuracy (85.5\% when finetuned with $384\times384$ inputs), notably surpassing all other Mamba variants in vision. The code is available at \url{https://github.com/OliverRensu/ARM}.