{"id":"https://openalex.org/W7125953006","doi":"https://doi.org/10.1145/3774934.3786417","title":"HelixPipe: Efficient Distributed Training of Long Sequence Transformers with Attention Parallel Pipeline Parallelism","display_name":"HelixPipe: Efficient Distributed Training of Long Sequence Transformers with Attention Parallel Pipeline Parallelism","publication_year":2026,"publication_date":"2026-01-28","ids":{"openalex":"https://openalex.org/W7125953006","doi":"https://doi.org/10.1145/3774934.3786417"},"language":null,"primary_location":{"id":"doi:10.1145/3774934.3786417","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3774934.3786417","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st 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/A5124092797","display_name":"Geng Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":true,"raw_author_name":"Geng Zhang","raw_affiliation_strings":["National University of Singapore, Singapore, Singapore"],"raw_orcid":"https://orcid.org/0000-0001-8898-4355","affiliations":[{"raw_affiliation_string":"National University of Singapore, Singapore, Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037038779","display_name":"Shenggan Cheng","orcid":"https://orcid.org/0000-0002-7966-2941"},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Shenggan Cheng","raw_affiliation_strings":["National University of Singapore, Singapore, Singapore"],"raw_orcid":"https://orcid.org/0000-0002-7966-2941","affiliations":[{"raw_affiliation_string":"National University of Singapore, Singapore, Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102796963","display_name":"Xuanlei Zhao","orcid":"https://orcid.org/0009-0000-4877-3115"},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Xuanlei Zhao","raw_affiliation_strings":["National University of Singapore, Singapore, Singapore"],"raw_orcid":"https://orcid.org/0009-0000-4877-3115","affiliations":[{"raw_affiliation_string":"National University of Singapore, Singapore, Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101924544","display_name":"Ziming Liu","orcid":"https://orcid.org/0009-0009-3355-6770"},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Ziming Liu","raw_affiliation_strings":["National University of Singapore, Singapore, Singapore"],"raw_orcid":"https://orcid.org/0009-0009-3355-6770","affiliations":[{"raw_affiliation_string":"National University of Singapore, Singapore, Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5124143247","display_name":"Yang You","orcid":null},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Yang You","raw_affiliation_strings":["National University of Singapore, Singapore, Singapore"],"raw_orcid":"https://orcid.org/0000-0003-2816-4384","affiliations":[{"raw_affiliation_string":"National University of Singapore, Singapore, Singapore","institution_ids":["https://openalex.org/I165932596"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5124092797"],"corresponding_institution_ids":["https://openalex.org/I165932596"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.13157239,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"413","last_page":"424"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.5526000261306763,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.5526000261306763,"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/T10363","display_name":"Low-power high-performance VLSI design","score":0.05869999900460243,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.046799998730421066,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.699400007724762},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6761000156402588},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.6488000154495239},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6355000138282776},{"id":"https://openalex.org/keywords/pipeline-transport","display_name":"Pipeline transport","score":0.47769999504089355},{"id":"https://openalex.org/keywords/schedule","display_name":"Schedule","score":0.42309999465942383},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.38359999656677246}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8069999814033508},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.699400007724762},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6801000237464905},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6761000156402588},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6488000154495239},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6355000138282776},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.47769999504089355},{"id":"https://openalex.org/C68387754","wikidata":"https://www.wikidata.org/wiki/Q7271585","display_name":"Schedule","level":2,"score":0.42309999465942383},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.38359999656677246},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.3617999851703644},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.33329999446868896},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.325300008058548},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.32440000772476196},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.2842999994754791},{"id":"https://openalex.org/C83283714","wikidata":"https://www.wikidata.org/wiki/Q121117","display_name":"Supercomputer","level":2,"score":0.2727999985218048},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.26739999651908875},{"id":"https://openalex.org/C134765980","wikidata":"https://www.wikidata.org/wiki/Q879126","display_name":"Bitwise operation","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.25839999318122864}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3774934.3786417","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3774934.3786417","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7","score":0.4494032561779022}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":13,"referenced_works":["https://openalex.org/W2119391823","https://openalex.org/W3129831491","https://openalex.org/W3132107458","https://openalex.org/W3204998121","https://openalex.org/W3206832494","https://openalex.org/W4385571616","https://openalex.org/W4386348101","https://openalex.org/W4394564255","https://openalex.org/W4394998892","https://openalex.org/W4395117922","https://openalex.org/W4399452094","https://openalex.org/W4407197296","https://openalex.org/W4408029480"],"related_works":[],"abstract_inverted_index":{"As":[0],"transformer":[1,36],"sequence":[2,35,104,139],"lengths":[3],"grow,":[4],"existing":[5,108],"pipeline":[6,31,54,59,116],"parallelisms":[7],"incur":[8],"suboptimal":[9],"performance":[10],"due":[11],"to":[12,70,88],"the":[13,18],"quadratic":[14],"attention":[15,41,46,84],"computation":[16],"and":[17,74,85,91,106,112,120],"substantial":[19],"memory":[20,72],"overhead.":[21],"To":[22],"relieve":[23],"these":[24],"challenges,":[25],"we":[26],"propose":[27],"HelixPipe,":[28],"a":[29,64,126,134],"novel":[30],"parallelism":[32],"for":[33],"long":[34],"training.":[37],"First,":[38],"HelixPipe":[39,80,98],"introduces":[40],"parallel":[42],"partition,":[43],"which":[44],"schedules":[45],"computations":[47],"of":[48],"different":[49,53],"micro":[50,67],"batches":[51],"across":[52,114],"stages":[55],"in":[56,110],"parallel,":[57],"reducing":[58],"bubbles.":[60],"Second,":[61],"it":[62,124],"employs":[63],"two-fold":[65],"first-in-last-out":[66],"batch":[68],"schedule":[69],"balance":[71],"usage":[73],"overlap":[75],"communication":[76],"with":[77,102,137],"computation.":[78],"Additionally,":[79],"utilizes":[81],"recomputation":[82],"without":[83],"chunked":[86],"MLP":[87],"mitigate":[89],"fragmentation":[90],"enable":[92],"longer":[93,103],"sequences.":[94],"Experiments":[95],"demonstrate":[96],"that":[97],"gains":[99],"increasing":[100],"advantages":[101],"lengths,":[105],"outperforms":[107],"methods":[109,131],"throughput":[111],"scalability":[113],"varying":[115],"sizes,":[117,119],"model":[118,136],"cluster":[121],"configurations.":[122],"Notably,":[123],"achieves":[125],"26%":[127],"speedup":[128],"over":[129],"baseline":[130],"when":[132],"training":[133],"7B":[135],"128k":[138],"length":[140],"on":[141],"64":[142],"H20":[143],"GPUs.":[144],"Code":[145],"is":[146],"available":[147],"at":[148],"https://github.com/zxgx/Megatron-LM.":[149]},"counts_by_year":[],"updated_date":"2026-01-29T23:17:01.242718","created_date":"2026-01-29T00:00:00"}