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

TL;DR

InstantNet is a novel framework that automatically generates switchable-precision neural networks, enabling IoT devices to adapt their accuracy and efficiency dynamically, thus optimizing deployment across diverse hardware constraints.

Contribution

It introduces an automated method for creating switchable-precision DNNs that can be deployed instantly, addressing the need for adaptable IoT neural network solutions.

Findings

InstantNet outperforms existing designs in efficiency and accuracy.

The generated networks support real-time precision switching.

Extensive experiments validate the effectiveness of InstantNet.

Abstract

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.