{"id":"https://openalex.org/W4321466422","doi":"https://doi.org/10.1145/3572848.3577484","title":"Elastic Averaging for Efficient Pipelined DNN Training","display_name":"Elastic Averaging for Efficient Pipelined DNN Training","publication_year":2023,"publication_date":"2023-02-21","ids":{"openalex":"https://openalex.org/W4321466422","doi":"https://doi.org/10.1145/3572848.3577484"},"language":"en","primary_location":{"id":"doi:10.1145/3572848.3577484","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3572848.3577484","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 28th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100434177","display_name":"Zihao Chen","orcid":"https://orcid.org/0000-0002-9349-0349"},"institutions":[{"id":"https://openalex.org/I66867065","display_name":"East China Normal University","ror":"https://ror.org/02n96ep67","country_code":"CN","type":"education","lineage":["https://openalex.org/I66867065"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Zihao Chen","raw_affiliation_strings":["East China Normal University"],"affiliations":[{"raw_affiliation_string":"East China Normal University","institution_ids":["https://openalex.org/I66867065"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101755392","display_name":"Xu Chen","orcid":"https://orcid.org/0000-0003-0144-1775"},"institutions":[{"id":"https://openalex.org/I66867065","display_name":"East China Normal University","ror":"https://ror.org/02n96ep67","country_code":"CN","type":"education","lineage":["https://openalex.org/I66867065"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen Xu","raw_affiliation_strings":["East China Normal University"],"affiliations":[{"raw_affiliation_string":"East China Normal University","institution_ids":["https://openalex.org/I66867065"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089931216","display_name":"Weining Qian","orcid":"https://orcid.org/0000-0002-4132-8630"},"institutions":[{"id":"https://openalex.org/I66867065","display_name":"East China Normal University","ror":"https://ror.org/02n96ep67","country_code":"CN","type":"education","lineage":["https://openalex.org/I66867065"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weining Qian","raw_affiliation_strings":["East China Normal University"],"affiliations":[{"raw_affiliation_string":"East China Normal University","institution_ids":["https://openalex.org/I66867065"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101511862","display_name":"Aoying Zhou","orcid":"https://orcid.org/0000-0002-4665-7302"},"institutions":[{"id":"https://openalex.org/I66867065","display_name":"East China Normal University","ror":"https://ror.org/02n96ep67","country_code":"CN","type":"education","lineage":["https://openalex.org/I66867065"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Aoying Zhou","raw_affiliation_strings":["East China Normal University"],"affiliations":[{"raw_affiliation_string":"East China Normal University","institution_ids":["https://openalex.org/I66867065"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5100434177"],"corresponding_institution_ids":["https://openalex.org/I66867065"],"apc_list":null,"apc_paid":null,"fwci":1.2282,"has_fulltext":false,"cited_by_count":10,"citation_normalized_percentile":{"value":0.80827601,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"380","last_page":"391"},"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/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9972000122070312,"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"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.996999979019165,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8596632480621338},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.8185184001922607},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.7849409580230713},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.7391467094421387},{"id":"https://openalex.org/keywords/pipeline-transport","display_name":"Pipeline transport","score":0.7213236093521118},{"id":"https://openalex.org/keywords/data-parallelism","display_name":"Data parallelism","score":0.6341453790664673},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.561091959476471},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5349065065383911},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.4784300923347473},{"id":"https://openalex.org/keywords/partition","display_name":"Partition (number theory)","score":0.4775346517562866},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.36738577485084534},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.22407716512680054}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8596632480621338},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.8185184001922607},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.7849409580230713},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.7391467094421387},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.7213236093521118},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.6341453790664673},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.561091959476471},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5349065065383911},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4784300923347473},{"id":"https://openalex.org/C42812","wikidata":"https://www.wikidata.org/wiki/Q1082910","display_name":"Partition (number theory)","level":2,"score":0.4775346517562866},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.36738577485084534},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.22407716512680054},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.0},{"id":"https://openalex.org/C87717796","wikidata":"https://www.wikidata.org/wiki/Q146326","display_name":"Environmental engineering","level":1,"score":0.0},{"id":"https://openalex.org/C114614502","wikidata":"https://www.wikidata.org/wiki/Q76592","display_name":"Combinatorics","level":1,"score":0.0},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3572848.3577484","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3572848.3577484","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 28th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1299647234","display_name":null,"funder_award_id":"62272168","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":15,"referenced_works":["https://openalex.org/W1788418780","https://openalex.org/W2086161653","https://openalex.org/W2154987621","https://openalex.org/W2155893237","https://openalex.org/W2560674852","https://openalex.org/W2741609678","https://openalex.org/W2969388332","https://openalex.org/W2970139027","https://openalex.org/W2980268386","https://openalex.org/W2998508934","https://openalex.org/W3086105743","https://openalex.org/W3132107458","https://openalex.org/W3137550677","https://openalex.org/W3206832494","https://openalex.org/W4220967350"],"related_works":["https://openalex.org/W2950520577","https://openalex.org/W1501159154","https://openalex.org/W1554644772","https://openalex.org/W2003935582","https://openalex.org/W2494130044","https://openalex.org/W3170887803","https://openalex.org/W2963831937","https://openalex.org/W74409296","https://openalex.org/W3209384898","https://openalex.org/W4400951174"],"abstract_inverted_index":{"Nowadays,":[0],"the":[1,18,50,57,88,91,98,110,120,132],"size":[2],"of":[3,26,43,59,93,112,159],"DNN":[4],"models":[5],"has":[6],"grown":[7],"rapidly.":[8],"To":[9,86],"train":[10],"a":[11,36,103,125,139,153],"large":[12],"model,":[13],"pipeline":[14,32,61,94,160],"parallelism-based":[15],"frameworks":[16],"partition":[17],"model":[19],"across":[20],"GPUs":[21],"and":[22,39,115],"slice":[23],"each":[24],"batch":[25],"data":[27],"into":[28,138],"multiple":[29,83],"micro-batches.":[30],"However,":[31],"parallelism":[33,95,161],"suffers":[34],"from":[35],"bubble":[37],"issue":[38],"low":[40],"peak":[41],"utilization":[42],"GPUs.":[44],"Recent":[45],"work":[46],"tries":[47],"to":[48,55,81,119,129],"address":[49],"two":[51],"issues,":[52],"but":[53],"fails":[54],"exploit":[56,90],"benefit":[58],"vanilla":[60],"parallelism,":[62],"i.e.,":[63],"overlapping":[64],"communication":[65],"with":[66],"computation.":[67],"In":[68],"this":[69],"work,":[70],"we":[71,101,123],"employ":[72],"an":[73],"elastic":[74,79],"averaging-based":[75],"framework":[76,89,121],"which":[77],"explores":[78],"averaging":[80],"add":[82],"parallel":[84,113],"pipelines.":[85],"help":[87],"advantage":[92],"while":[96],"reducing":[97],"memory":[99],"footprints,":[100],"propose":[102,124],"schedule,":[104],"advance":[105],"forward":[106],"propagation.":[107],"Moreover,":[108],"since":[109],"numbers":[111],"pipelines":[114],"micro-batches":[116],"are":[117],"essential":[118],"performance,":[122],"profiling-based":[126],"tuning":[127],"method":[128],"automatically":[130],"determine":[131],"settings.":[133],"We":[134],"integrate":[135],"those":[136],"techniques":[137],"prototype":[140],"system,":[141],"namely":[142],"AvgPipe,":[143],"based":[144],"on":[145,162],"PyTorch.":[146],"Our":[147],"experiments":[148],"show":[149],"that":[150],"Avg-Pipe":[151],"achieves":[152],"1.7x":[154],"speedups":[155],"over":[156],"state-of-the-art":[157],"solutions":[158],"average.":[163]},"counts_by_year":[{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":5},{"year":2023,"cited_by_count":2}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}