{"id":"https://openalex.org/W2969766737","doi":"https://doi.org/10.1109/mm.2019.2935967","title":"Optimizing Multi-GPU Parallelization Strategies for Deep Learning Training","display_name":"Optimizing Multi-GPU Parallelization Strategies for Deep Learning Training","publication_year":2019,"publication_date":"2019-07-30","ids":{"openalex":"https://openalex.org/W2969766737","doi":"https://doi.org/10.1109/mm.2019.2935967","mag":"2969766737"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:1907.13257","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1907.13257","pdf_url":"https://arxiv.org/pdf/1907.13257","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"type":"article","indexed_in":["arxiv"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/1907.13257","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5036103216","display_name":"Saptadeep Pal","orcid":"https://orcid.org/0000-0002-8777-8573"},"institutions":[{"id":"https://openalex.org/I2803209242","display_name":"University of California System","ror":"https://ror.org/00pjdza24","country_code":"US","type":"education","lineage":["https://openalex.org/I2803209242"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Pal, Saptadeep","raw_affiliation_strings":["University of California"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California","institution_ids":["https://openalex.org/I2803209242"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089789557","display_name":"Eiman Ebrahimi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ebrahimi, Eiman","raw_affiliation_strings":["NVIDIA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065112065","display_name":"Arslan Zulfiqar","orcid":"https://orcid.org/0009-0003-6240-5900"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zulfiqar, Arslan","raw_affiliation_strings":["NVIDIA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010195820","display_name":"Yaosheng Fu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fu, Yaosheng","raw_affiliation_strings":["NVIDIA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101765640","display_name":"Victor Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Victor","raw_affiliation_strings":["NVIDIA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031316782","display_name":"Szymon Migacz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Migacz, Szymon","raw_affiliation_strings":["NVIDIA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031781240","display_name":"David Nellans","orcid":"https://orcid.org/0000-0001-5203-8367"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nellans, David","raw_affiliation_strings":["NVIDIA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5084229134","display_name":"Puneet Gupta","orcid":"https://orcid.org/0000-0002-6188-1134"},"institutions":[{"id":"https://openalex.org/I2803209242","display_name":"University of California System","ror":"https://ror.org/00pjdza24","country_code":"US","type":"education","lineage":["https://openalex.org/I2803209242"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Gupta, Puneet","raw_affiliation_strings":["University of California"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California","institution_ids":["https://openalex.org/I2803209242"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5036103216"],"corresponding_institution_ids":["https://openalex.org/I2803209242"],"apc_list":null,"apc_paid":null,"fwci":3.8802,"has_fulltext":false,"cited_by_count":72,"citation_normalized_percentile":{"value":0.94928667,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":100},"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998000264167786,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998000264167786,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9998000264167786,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9977999925613403,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.9119777679443359},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.8708691596984863},{"id":"https://openalex.org/keywords/dataflow","display_name":"Dataflow","score":0.8586966395378113},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.7335657477378845},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5879821181297302},{"id":"https://openalex.org/keywords/data-parallelism","display_name":"Data parallelism","score":0.5285263061523438},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.45476725697517395},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.4125509560108185},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.37640509009361267},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.3139733076095581}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.9119777679443359},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.8708691596984863},{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.8586966395378113},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.7335657477378845},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5879821181297302},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.5285263061523438},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.45476725697517395},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.4125509560108185},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37640509009361267},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.3139733076095581},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:arXiv.org:1907.13257","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1907.13257","pdf_url":"https://arxiv.org/pdf/1907.13257","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:1907.13257","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1907.13257","pdf_url":"https://arxiv.org/pdf/1907.13257","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":20,"referenced_works":["https://openalex.org/W1547840952","https://openalex.org/W2168231600","https://openalex.org/W2183341477","https://openalex.org/W2259472270","https://openalex.org/W2287011250","https://openalex.org/W2336650964","https://openalex.org/W2525778437","https://openalex.org/W2617242334","https://openalex.org/W2622263826","https://openalex.org/W2785105386","https://openalex.org/W2785452945","https://openalex.org/W2786414509","https://openalex.org/W2807147113","https://openalex.org/W2901299405","https://openalex.org/W2963959597","https://openalex.org/W2963959650","https://openalex.org/W2969766737","https://openalex.org/W2991040477","https://openalex.org/W4293775467","https://openalex.org/W4298312531"],"related_works":["https://openalex.org/W2187181201","https://openalex.org/W1966837078","https://openalex.org/W1554644772","https://openalex.org/W2950520577","https://openalex.org/W2494130044","https://openalex.org/W2003935582","https://openalex.org/W2177838837","https://openalex.org/W2963831937","https://openalex.org/W3209384898","https://openalex.org/W74409296"],"abstract_inverted_index":{"Deploying":[0],"deep":[1],"learning":[2],"(DL)":[3],"models":[4,14],"across":[5,139],"multiple":[6],"compute":[7],"devices":[8,45],"to":[9,16,67,79,81,104,109,195],"train":[10],"large":[11],"and":[12,26,91,176,191],"complex":[13],"continues":[15],"grow":[17],"in":[18,46],"importance":[19],"because":[20],"of":[21,44,75,96,134,186],"the":[22,34,42,53,98,141,178],"demand":[23],"for":[24,173],"faster":[25],"more":[27,135,159],"frequent":[28],"training.":[29],"Data":[30],"parallelism":[31,120],"(DP)":[32],"is":[33,132,146],"most":[35],"widely":[36],"used":[37],"parallelization":[38],"strategy,":[39],"but":[40],"as":[41],"number":[43,74,95],"data":[47,129],"parallel":[48,130],"training":[49,89,112,156,164,184],"grows,":[50],"so":[51],"does":[52],"communication":[54],"overhead":[55],"between":[56],"devices.":[57],"Additionally,":[58],"a":[59,72,82,93],"larger":[60,73],"aggregate":[61],"batch":[62],"size":[63],"per":[64],"step":[65,113],"leads":[66],"statistical":[68],"efficiency":[69],"loss,":[70],"i.e.,":[71],"epochs":[76],"are":[77],"required":[78],"converge":[80],"desired":[83],"accuracy.":[84],"These":[85],"factors":[86],"affect":[87],"overall":[88],"time":[90,165],"beyond":[92],"certain":[94],"devices,":[97],"speedup":[99,185],"from":[100],"leveraging":[101],"DP":[102,168,197],"begins":[103],"scale":[105],"poorly.":[106],"In":[107],"addition":[108],"DP,":[110],"each":[111,128],"can":[114,199],"be":[115,158],"accelerated":[116],"by":[117],"exploiting":[118,167],"model":[119,142],"(MP).":[121],"This":[122],"work":[123],"explores":[124],"hybrid":[125,155,179],"parallelization,":[126],"where":[127],"worker":[131],"comprised":[133],"than":[136,166],"one":[137],"device,":[138],"which":[140],"dataflow":[143],"graph":[144],"(DFG)":[145],"split":[147],"using":[148],"MP.":[149],"We":[150,170],"show":[151],"that":[152,172],"at":[153,161,187,201],"scale,":[154],"will":[157],"effective":[160],"minimizing":[162],"end-to-end":[163,183],"alone.":[169],"project":[171],"Inception-V3,":[174],"GNMT,":[175],"BigLSTM,":[177],"strategy":[180],"provides":[181],"an":[182],"least":[188],"26.5%,":[189],"8%,":[190],"22%":[192],"respectively":[193],"compared":[194],"what":[196],"alone":[198],"achieve":[200],"scale.":[202]},"counts_by_year":[{"year":2025,"cited_by_count":14},{"year":2024,"cited_by_count":8},{"year":2023,"cited_by_count":12},{"year":2022,"cited_by_count":9},{"year":2021,"cited_by_count":19},{"year":2020,"cited_by_count":7},{"year":2019,"cited_by_count":3}],"updated_date":"2026-04-28T14:05:53.105641","created_date":"2019-08-29T00:00:00"}