Discovering Multi-Hardware Mobile Models via Architecture Search

Hardware-aware neural architecture designs have been predominantly focusing on optimizing model performance on single hardware and model development complexity, where another important factor, model deployment complexity, has been largely ignored. In this paper, we argue that, for applications that...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Chu, Grace, Arikan, Okan, Bender, Gabriel, Wang, Weijun, Brighton, Achille, Kindermans, Pieter-Jan, Liu, Hanxiao, Akin, Berkin, Gupta, Suyog, Howard, Andrew
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Chu, Grace
Arikan, Okan
Bender, Gabriel
Wang, Weijun
Brighton, Achille
Kindermans, Pieter-Jan
Liu, Hanxiao
Akin, Berkin
Gupta, Suyog
Howard, Andrew
description Hardware-aware neural architecture designs have been predominantly focusing on optimizing model performance on single hardware and model development complexity, where another important factor, model deployment complexity, has been largely ignored. In this paper, we argue that, for applications that may be deployed on multiple hardware, having different single-hardware models across the deployed hardware makes it hard to guarantee consistent outputs across hardware and duplicates engineering work for debugging and fixing. To minimize such deployment cost, we propose an alternative solution, multi-hardware models, where a single architecture is developed for multiple hardware. With thoughtful search space design and incorporating the proposed multi-hardware metrics in neural architecture search, we discover multi-hardware models that give state-of-the-art (SoTA) performance across multiple hardware in both average and worse case scenarios. For performance on individual hardware, the single multi-hardware model yields similar or better results than SoTA performance on accelerators like GPU, DSP and EdgeTPU which was achieved by different models, while having similar performance with MobilenetV3 Large Minimalistic model on mobile CPU.
doi_str_mv 10.48550/arxiv.2008.08178
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2008_08178</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2008_08178</sourcerecordid><originalsourceid>FETCH-LOGICAL-a678-d78d6f111d108f9896af3c6f4d66e371a2abcd3dce7057eebd017822b17543aa3</originalsourceid><addsrcrecordid>eNotj8FuwjAQRH3hgIAP4ER-IKkdJ7a5gYAWJBAHuEcb7xosBaicENq_J9CenkYjjeYxNhY8yUye8w8IP75NUs5Nwo3Qps9mS1_bW0vBX0_R7l41Pl5DwAcEina30lcvIFV11HqI5sGefUO2uXf1gaCLQ9ZzUNU0-ueAHT9Xx8U63u6_Nov5NgalTYzaoHJCCBTcuKmZKnDSKpehUiS1gBRKixItaZ5rohJ5dy9NS6HzTALIAZv8zb4Viu_gLxB-i5dK8VaRTyBmRAY</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Discovering Multi-Hardware Mobile Models via Architecture Search</title><source>arXiv.org</source><creator>Chu, Grace ; Arikan, Okan ; Bender, Gabriel ; Wang, Weijun ; Brighton, Achille ; Kindermans, Pieter-Jan ; Liu, Hanxiao ; Akin, Berkin ; Gupta, Suyog ; Howard, Andrew</creator><creatorcontrib>Chu, Grace ; Arikan, Okan ; Bender, Gabriel ; Wang, Weijun ; Brighton, Achille ; Kindermans, Pieter-Jan ; Liu, Hanxiao ; Akin, Berkin ; Gupta, Suyog ; Howard, Andrew</creatorcontrib><description>Hardware-aware neural architecture designs have been predominantly focusing on optimizing model performance on single hardware and model development complexity, where another important factor, model deployment complexity, has been largely ignored. In this paper, we argue that, for applications that may be deployed on multiple hardware, having different single-hardware models across the deployed hardware makes it hard to guarantee consistent outputs across hardware and duplicates engineering work for debugging and fixing. To minimize such deployment cost, we propose an alternative solution, multi-hardware models, where a single architecture is developed for multiple hardware. With thoughtful search space design and incorporating the proposed multi-hardware metrics in neural architecture search, we discover multi-hardware models that give state-of-the-art (SoTA) performance across multiple hardware in both average and worse case scenarios. For performance on individual hardware, the single multi-hardware model yields similar or better results than SoTA performance on accelerators like GPU, DSP and EdgeTPU which was achieved by different models, while having similar performance with MobilenetV3 Large Minimalistic model on mobile CPU.</description><identifier>DOI: 10.48550/arxiv.2008.08178</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition</subject><creationdate>2020-08</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,777,882</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2008.08178$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2008.08178$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Chu, Grace</creatorcontrib><creatorcontrib>Arikan, Okan</creatorcontrib><creatorcontrib>Bender, Gabriel</creatorcontrib><creatorcontrib>Wang, Weijun</creatorcontrib><creatorcontrib>Brighton, Achille</creatorcontrib><creatorcontrib>Kindermans, Pieter-Jan</creatorcontrib><creatorcontrib>Liu, Hanxiao</creatorcontrib><creatorcontrib>Akin, Berkin</creatorcontrib><creatorcontrib>Gupta, Suyog</creatorcontrib><creatorcontrib>Howard, Andrew</creatorcontrib><title>Discovering Multi-Hardware Mobile Models via Architecture Search</title><description>Hardware-aware neural architecture designs have been predominantly focusing on optimizing model performance on single hardware and model development complexity, where another important factor, model deployment complexity, has been largely ignored. In this paper, we argue that, for applications that may be deployed on multiple hardware, having different single-hardware models across the deployed hardware makes it hard to guarantee consistent outputs across hardware and duplicates engineering work for debugging and fixing. To minimize such deployment cost, we propose an alternative solution, multi-hardware models, where a single architecture is developed for multiple hardware. With thoughtful search space design and incorporating the proposed multi-hardware metrics in neural architecture search, we discover multi-hardware models that give state-of-the-art (SoTA) performance across multiple hardware in both average and worse case scenarios. For performance on individual hardware, the single multi-hardware model yields similar or better results than SoTA performance on accelerators like GPU, DSP and EdgeTPU which was achieved by different models, while having similar performance with MobilenetV3 Large Minimalistic model on mobile CPU.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8FuwjAQRH3hgIAP4ER-IKkdJ7a5gYAWJBAHuEcb7xosBaicENq_J9CenkYjjeYxNhY8yUye8w8IP75NUs5Nwo3Qps9mS1_bW0vBX0_R7l41Pl5DwAcEina30lcvIFV11HqI5sGefUO2uXf1gaCLQ9ZzUNU0-ueAHT9Xx8U63u6_Nov5NgalTYzaoHJCCBTcuKmZKnDSKpehUiS1gBRKixItaZ5rohJ5dy9NS6HzTALIAZv8zb4Viu_gLxB-i5dK8VaRTyBmRAY</recordid><startdate>20200818</startdate><enddate>20200818</enddate><creator>Chu, Grace</creator><creator>Arikan, Okan</creator><creator>Bender, Gabriel</creator><creator>Wang, Weijun</creator><creator>Brighton, Achille</creator><creator>Kindermans, Pieter-Jan</creator><creator>Liu, Hanxiao</creator><creator>Akin, Berkin</creator><creator>Gupta, Suyog</creator><creator>Howard, Andrew</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20200818</creationdate><title>Discovering Multi-Hardware Mobile Models via Architecture Search</title><author>Chu, Grace ; Arikan, Okan ; Bender, Gabriel ; Wang, Weijun ; Brighton, Achille ; Kindermans, Pieter-Jan ; Liu, Hanxiao ; Akin, Berkin ; Gupta, Suyog ; Howard, Andrew</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a678-d78d6f111d108f9896af3c6f4d66e371a2abcd3dce7057eebd017822b17543aa3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>Chu, Grace</creatorcontrib><creatorcontrib>Arikan, Okan</creatorcontrib><creatorcontrib>Bender, Gabriel</creatorcontrib><creatorcontrib>Wang, Weijun</creatorcontrib><creatorcontrib>Brighton, Achille</creatorcontrib><creatorcontrib>Kindermans, Pieter-Jan</creatorcontrib><creatorcontrib>Liu, Hanxiao</creatorcontrib><creatorcontrib>Akin, Berkin</creatorcontrib><creatorcontrib>Gupta, Suyog</creatorcontrib><creatorcontrib>Howard, Andrew</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Chu, Grace</au><au>Arikan, Okan</au><au>Bender, Gabriel</au><au>Wang, Weijun</au><au>Brighton, Achille</au><au>Kindermans, Pieter-Jan</au><au>Liu, Hanxiao</au><au>Akin, Berkin</au><au>Gupta, Suyog</au><au>Howard, Andrew</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Discovering Multi-Hardware Mobile Models via Architecture Search</atitle><date>2020-08-18</date><risdate>2020</risdate><abstract>Hardware-aware neural architecture designs have been predominantly focusing on optimizing model performance on single hardware and model development complexity, where another important factor, model deployment complexity, has been largely ignored. In this paper, we argue that, for applications that may be deployed on multiple hardware, having different single-hardware models across the deployed hardware makes it hard to guarantee consistent outputs across hardware and duplicates engineering work for debugging and fixing. To minimize such deployment cost, we propose an alternative solution, multi-hardware models, where a single architecture is developed for multiple hardware. With thoughtful search space design and incorporating the proposed multi-hardware metrics in neural architecture search, we discover multi-hardware models that give state-of-the-art (SoTA) performance across multiple hardware in both average and worse case scenarios. For performance on individual hardware, the single multi-hardware model yields similar or better results than SoTA performance on accelerators like GPU, DSP and EdgeTPU which was achieved by different models, while having similar performance with MobilenetV3 Large Minimalistic model on mobile CPU.</abstract><doi>10.48550/arxiv.2008.08178</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2008.08178
ispartof
issn
language eng
recordid cdi_arxiv_primary_2008_08178
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
title Discovering Multi-Hardware Mobile Models via Architecture Search
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-18T06%3A12%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Discovering%20Multi-Hardware%20Mobile%20Models%20via%20Architecture%20Search&rft.au=Chu,%20Grace&rft.date=2020-08-18&rft_id=info:doi/10.48550/arxiv.2008.08178&rft_dat=%3Carxiv_GOX%3E2008_08178%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true