{"id":"https://openalex.org/W4411725045","doi":"https://doi.org/10.1109/tpds.2025.3583983","title":"PIPEMESH: Achieving Memory-Efficient Computation-Communication Overlap for Training Large Language Models","display_name":"PIPEMESH: Achieving Memory-Efficient Computation-Communication Overlap for Training Large Language Models","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4411725045","doi":"https://doi.org/10.1109/tpds.2025.3583983"},"language":"en","primary_location":{"id":"doi:10.1109/tpds.2025.3583983","is_oa":true,"landing_page_url":"https://doi.org/10.1109/tpds.2025.3583983","pdf_url":null,"source":{"id":"https://openalex.org/S97130795","display_name":"IEEE Transactions on Parallel and Distributed Systems","issn_l":"1045-9219","issn":["1045-9219","1558-2183","2161-9883"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Parallel and Distributed Systems","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1109/tpds.2025.3583983","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5016904555","display_name":"Fanxin Li","orcid":"https://orcid.org/0000-0003-2268-3036"},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":true,"raw_author_name":"Fanxin Li","raw_affiliation_strings":["Department of Computer Science, The University of Hong Kong, Hong Kong, China","Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, The University of Hong Kong, Hong Kong, China","institution_ids":["https://openalex.org/I889458895"]},{"raw_affiliation_string":"Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088629918","display_name":"Shixiong Zhao","orcid":"https://orcid.org/0000-0002-1643-2583"},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Shixiong Zhao","raw_affiliation_strings":["Department of Computer Science, The University of Hong Kong, Hong Kong, China","Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, The University of Hong Kong, Hong Kong, China","institution_ids":["https://openalex.org/I889458895"]},{"raw_affiliation_string":"Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083329618","display_name":"Yuhao Qing","orcid":"https://orcid.org/0009-0009-1500-2192"},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Yuhao Qing","raw_affiliation_strings":["Department of Computer Science, The University of Hong Kong, Hong Kong, China","Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, The University of Hong Kong, Hong Kong, China","institution_ids":["https://openalex.org/I889458895"]},{"raw_affiliation_string":"Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091837703","display_name":"Jianyu Jiang","orcid":"https://orcid.org/0000-0002-8684-8509"},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Jianyu Jiang","raw_affiliation_strings":["Department of Computer Science, The University of Hong Kong, Hong Kong, China","Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, The University of Hong Kong, Hong Kong, China","institution_ids":["https://openalex.org/I889458895"]},{"raw_affiliation_string":"Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101527007","display_name":"Xusheng Chen","orcid":"https://orcid.org/0000-0002-2807-9780"},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xusheng Chen","raw_affiliation_strings":["Department of Computer Science, The University of Hong Kong, Hong Kong, China","Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, The University of Hong Kong, Hong Kong, China","institution_ids":["https://openalex.org/I889458895"]},{"raw_affiliation_string":"Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I889458895"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5015360935","display_name":"Heming Cui","orcid":"https://orcid.org/0000-0001-7746-440X"},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Heming Cui","raw_affiliation_strings":["Department of Computer Science, The University of Hong Kong, Hong Kong, China","Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, The University of Hong Kong, Hong Kong, China","institution_ids":["https://openalex.org/I889458895"]},{"raw_affiliation_string":"Department of Computer Science, University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I889458895"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5016904555"],"corresponding_institution_ids":["https://openalex.org/I889458895"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.08158226,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"18"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":1.0,"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"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":1.0,"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"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9998000264167786,"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"}},{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9958999752998352,"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":0.8685190677642822},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5803239345550537},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.566449761390686},{"id":"https://openalex.org/keywords/distributed-computing","display_name":"Distributed computing","score":0.3337436616420746},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3302891254425049},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.3264106512069702},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.32219767570495605},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.1991293728351593}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8685190677642822},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5803239345550537},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.566449761390686},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3337436616420746},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3302891254425049},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3264106512069702},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.32219767570495605},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.1991293728351593},{"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":"doi:10.1109/tpds.2025.3583983","is_oa":true,"landing_page_url":"https://doi.org/10.1109/tpds.2025.3583983","pdf_url":null,"source":{"id":"https://openalex.org/S97130795","display_name":"IEEE Transactions on Parallel and Distributed Systems","issn_l":"1045-9219","issn":["1045-9219","1558-2183","2161-9883"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Parallel and Distributed Systems","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1109/tpds.2025.3583983","is_oa":true,"landing_page_url":"https://doi.org/10.1109/tpds.2025.3583983","pdf_url":null,"source":{"id":"https://openalex.org/S97130795","display_name":"IEEE Transactions on Parallel and Distributed Systems","issn_l":"1045-9219","issn":["1045-9219","1558-2183","2161-9883"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Parallel and Distributed Systems","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":44,"referenced_works":["https://openalex.org/W1522301498","https://openalex.org/W2045271686","https://openalex.org/W2338908902","https://openalex.org/W2763421725","https://openalex.org/W2899663614","https://openalex.org/W2903697572","https://openalex.org/W2962834855","https://openalex.org/W2969388332","https://openalex.org/W2973727699","https://openalex.org/W2975712713","https://openalex.org/W2991040477","https://openalex.org/W3001279689","https://openalex.org/W3129831491","https://openalex.org/W3132107458","https://openalex.org/W3168867926","https://openalex.org/W3188065709","https://openalex.org/W3188270315","https://openalex.org/W3204998121","https://openalex.org/W3205068155","https://openalex.org/W3205949070","https://openalex.org/W3206832494","https://openalex.org/W4220703317","https://openalex.org/W4226479682","https://openalex.org/W4280611847","https://openalex.org/W4286987939","https://openalex.org/W4287756266","https://openalex.org/W4292779060","https://openalex.org/W4297097464","https://openalex.org/W4301239768","https://openalex.org/W4301581299","https://openalex.org/W4311642023","https://openalex.org/W4327909439","https://openalex.org/W4384648639","https://openalex.org/W4384918448","https://openalex.org/W4385245566","https://openalex.org/W4386768656","https://openalex.org/W4387356039","https://openalex.org/W4389157038","https://openalex.org/W4389518760","https://openalex.org/W4389776490","https://openalex.org/W4391125288","https://openalex.org/W4392223539","https://openalex.org/W4395117348","https://openalex.org/W4395117922"],"related_works":["https://openalex.org/W230091440","https://openalex.org/W2233261550","https://openalex.org/W2810751659","https://openalex.org/W258997015","https://openalex.org/W2997094352","https://openalex.org/W4412817058","https://openalex.org/W3216976533","https://openalex.org/W100620283","https://openalex.org/W2495260952","https://openalex.org/W4366179611"],"abstract_inverted_index":{"Efficiently":[0],"training":[1,23,61,169,179],"large":[2],"language":[3],"models":[4],"(LLMs)":[5],"on":[6,29,181],"commodity":[7,182],"cloud":[8],"resources":[9],"remains":[10],"challenging":[11],"due":[12],"to":[13,53,118,139,172,186],"limitations":[14],"in":[15],"network":[16],"bandwidth":[17],"and":[18,49,77,105,122,133],"accelerator":[19],"memory":[20,41,75,106,124,141],"capacity.":[21],"Existing":[22],"systems":[24,37],"can":[25],"be":[26],"categorized":[27],"based":[28],"their":[30],"pipeline":[31,93],"schedules.":[32],"Depth-first":[33],"scheduling,":[34],"employed":[35],"by":[36,80,158,184],"like":[38],"Megatron,":[39],"prioritizes":[40],"efficiency":[42],"but":[43,69],"restricts":[44],"the":[45,60,100,111,119,123,152,168],"overlap":[46,68,104],"between":[47,102],"communication":[48,67,103,120],"computation,":[50],"causing":[51],"accelerators":[52],"remain":[54],"idle":[55,155],"for":[56,163],"over":[57,99],"20%":[58],"of":[59,113,151,167],"time.":[62,170],"Conversely,":[63],"breadth-first":[64],"scheduling":[65],"maximizes":[66],"generates":[70],"excessive":[71],"intermediate":[72],"activations,":[73],"exceeding":[74],"capacity":[76],"slowing":[78],"computation":[79],"more":[81],"than":[82,165],"34%.":[83],"To":[84],"address":[85],"these":[86],"limitations,":[87],"we":[88,127],"propose":[89],"a":[90,129,134],"novel":[91],"elastic":[92],"schedule":[94],"that":[95,146],"enables":[96],"fine-grained":[97],"control":[98],"trade-off":[101],"consumption.":[107],"Our":[108],"approach":[109],"determines":[110],"number":[112],"micro-batches":[114],"scheduled":[115],"together":[116],"according":[117],"time":[121,156],"available.":[125],"Furthermore,":[126],"introduce":[128],"mixed":[130],"sharding":[131],"strategy":[132],"pipeline-aware":[135],"selective":[136],"recomputation":[137,161],"technique":[138],"reduce":[140],"usage.":[142],"Experimental":[143],"results":[144],"demonstrate":[145],"our":[147],"system":[148],"eliminates":[149],"most":[150],"28%":[153],"all-accelerator":[154],"caused":[157],"communication,":[159],"with":[160],"accounting":[162],"less":[164],"1.9%":[166],"Compared":[171],"existing":[173],"baselines,":[174],"<sc":[175],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[176],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">PIPEMESH</small>":[177],"improves":[178],"throughput":[180],"clouds":[183],"20.1%":[185],"33.8%.":[187]},"counts_by_year":[],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}