Optimizing Grouped Convolutions on Edge Devices

Perry Gibson, Jose Cano, Jack Turner, Elliot J Crowley, Michael F P O'Boyle, Amos J Storkey

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

When deploying a deep neural network on con-strained hardware, it is possible to replace the network’s standard convolutions with grouped convolutions. This allows for substantial memory savings with minimal loss of accuracy. However, current implementations of grouped convolutions in modern deep learning frameworks are far from performing optimally in terms of speed. In this paper we propose Grouped Spatial Pack Convolutions (GSPC), a new implementation of grouped convolutions that outperforms existing solutions. We implement GSPC in TVM, which provides state-of-the-art performance on edge devices. We analyze a set of networks utilizing different types of grouped convolutions and evaluate their performance in terms of inference time on several edge devices. We observe that our new implementation scales well with the number of groups and provides the best inference times in all settings, improving the existing implementations of grouped convolutions in TVM, PyTorch and TensorFlow Lite by 3.4×, 8× and 4× on average respectively. Code is available at https://github.com/gecLAB/tvm-GSPC/
Original languageEnglish
Title of host publication2020 IEEE 31st International Conference on Application-specific Systems, Architectures and Processors (ASAP)
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Pages189 - 196
Number of pages8
ISBN (Electronic)978-1-7281-7147-0
ISBN (Print)978-1-7281-7279-8
DOIs
Publication statusPublished - 31 Jul 2020
Event31st IEEE International Conference on Application-specific Systems, Architectures and Processors - The University of Manchester, Machester, United Kingdom
Duration: 6 Jul 20208 Jul 2020
https://asap2020.cs.manchester.ac.uk/index.php

Publication series

Name
PublisherIEEE
ISSN (Print)2160-0511
ISSN (Electronic)2160-052X

Conference

Conference31st IEEE International Conference on Application-specific Systems, Architectures and Processors
Abbreviated titleASAP 2020
Country/TerritoryUnited Kingdom
CityMachester
Period6/07/208/07/20
Internet address

Fingerprint

Dive into the research topics of 'Optimizing Grouped Convolutions on Edge Devices'. Together they form a unique fingerprint.

Cite this