System and method for deep learning application model offloading for wearable devices

Abstract

The present invention provides a deep learning application model offloading system and method for wearable devices, including: a prediction model module, used to predict the time and power required to run each layer of the deep learning application model as prediction data; a developer interface The module is used as an interface for introducing model offloading algorithms and data into deep learning applications; the status information collection module is used to collect hardware information and the runtime status of application models as the basis for offloading; the model offloading decision module is used to load models The load sharing algorithm and data are used to obtain all possible load sharing situations, and then search for the optimal load sharing method according to the load sharing basis and forecast data; according to the optimal load sharing method, the deep learning application model is divided into two sub- Model. As an automated deep learning operation and optimization framework, its core idea is to find the optimal model offloading method according to the current runtime environment, and offload part of the sub-model calculations to the paired mobile phone.

Description

technical field [0001] The present invention relates to the field of software technology, in particular to artificial intelligence, and specifically to a system and method for optimizing deep learning applications running on wearable devices through adaptive model offloading, which is suitable for development, deployment and operation on wearable devices deep learning applications. Background technique [0002] Deep learning is a machine learning algorithm and an important branch of artificial intelligence. From rapid development to practical application, in just a few years, deep learning has subverted the algorithm design ideas in many fields such as speech recognition, image classification, and text understanding, and gradually formed an end-to-end ( end-to-end) model, and then directly output a new model to get the final result. On wearable devices, deep learning technology has also been widely used. For example, relying on deep learning in smart glasses for high-accu...

AI Technical Summary

Problems solved by technology

[0010] 1) Computational offloading needs a faster network connection as support, but this premise is difficult to meet in real conditions;

[0011] 2) Computing offloading requires data to be transmitted to a remote server through a wireless network, which may lead to leakage of private data

Images