arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2104.10853cs.LG
收藏

InstantNet:瞬时切换精度网络的自动生成与部署

InstantNet: Automated Generation and Deployment of Instantaneously Switchable-Precision Networks

Rice University(莱斯大学);University of Texas at Austin(德克萨斯大学奥斯汀分校);Sun Yat-sen University(中山大学)

Yonggan Fu, Zhongzhi Yu, Yongan Zhang, Yifan Jiang, Chaojian Li, Yongyuan Liang, Mingchao Jiang, Zhangyang Wang, Yingyan Celine Lin

AI总结:

本文提出InstantNet,自动生成并部署可变位宽的瞬时切换精度网络,以同时满足物联网设备对精度-效率权衡和高效数据流的需求,并在实验中持续超越现有设计。

摘要:

深度神经网络(DNN)驱动的物联网(IoT)设备的前景,催生了对自动化解决方案的巨大需求,以实现两类高效能力的快速开发与部署:(1)具备瞬时精度-效率权衡能力的DNN,以适应物联网设备上时变的资源;(2)用于优化DNN在不同设备上执行效率的数据流。为此,我们提出InstantNet,用于自动生成和部署可在可变位宽下运行的瞬时切换精度网络。大量实验表明,所提出的InstantNet持续优于现有最先进设计。

英文摘要:

The promise of Deep Neural Network (DNN) powered Internet of Thing (IoT) devices has motivated a tremendous demand for automated solutions to enable fast development and deployment of efficient (1) DNNs equipped with instantaneous accuracy-efficiency trade-off capability to accommodate the time-varying resources at IoT devices and (2) dataflows to optimize DNNs' execution efficiency on different devices. Therefore, we propose InstantNet to automatically generate and deploy instantaneously switchable-precision networks which operate at variable bit-widths. Extensive experiments show that the proposed InstantNet consistently outperforms state-of-the-art designs.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑